Mirantis OpenStack for Kubernetes Documentation¶
This documentation provides information on how to deploy and operate a Mirantis OpenStack for Kubernetes (MOSK) environment. The documentation is intended to help operators to understand the core concepts of the product. The documentation provides sufficient information to deploy and operate the solution.
The information provided in this documentation set is being constantly improved and amended based on the feedback and kind requests from the consumers of MOS.
The following table lists the guides included in the documentation set you are reading:
Guide |
Purpose |
|---|---|
Learn the fundamentals of MOSK reference architecture to appropriately plan your deployment |
|
Deploy a MOSK environment of a preferred configuration using supported deployment profiles tailored to the demands of specific business cases |
|
Operate your MOSK environment |
|
Learn about new features and bug fixes in the current MOSK version |
Intended audience¶
This documentation is intended for engineers who have the basic knowledge of Linux, virtualization and containerization technologies, Kubernetes API and CLI, Helm and Helm charts, Mirantis Kubernetes Engine (MKE), and OpenStack.
Documentation history¶
The following table contains the released revision of the documentation set you are reading.
Release date |
Release name |
|---|---|
August, 2023 |
MOSK 23.2 series |
Conventions¶
This documentation set uses the following conventions in the HTML format:
Convention |
Description |
|---|---|
boldface font |
Inline CLI tools and commands, titles of the procedures and system response examples, table titles |
|
Files names and paths, Helm charts parameters and their values, names of packages, nodes names and labels, and so on |
italic font |
Information that distinguishes some concept or term |
External links and cross-references, footnotes |
|
Main menu > menu item |
GUI elements that include any part of interactive user interface and menu navigation |
Superscript |
Some extra, brief information |
Note The Note block |
Messages of a generic meaning that may be useful for the user |
Caution The Caution block |
Information that prevents a user from mistakes and undesirable consequences when following the procedures |
Warning The Warning block |
Messages that include details that can be easily missed, but should not be ignored by the user and are valuable before proceeding |
See also The See also block |
List of references that may be helpful for understanding of some related tools, concepts, and so on |
Learn more The Learn more block |
Used in the Release Notes to wrap a list of internal references to the reference architecture, deployment and operation procedures specific to a newly implemented product feature |
Product Overview¶
Mirantis OpenStack for Kubernetes (MOSK) combines the power of Mirantis Container Cloud for delivering and managing Kubernetes clusters, with the industry standard OpenStack APIs, enabling you to build your own cloud infrastructure.
The advantages of running all of the OpenStack components as a Kubernetes application are multi-fold and include the following:
Zero downtime, non-disruptive updates
Fully automated Day-2 operations
Full-stack management from bare metal through the operating system and all the necessary components
The list of the most common use cases includes:
- Software-defined data center
The traditional data center requires multiple requests and interactions to deploy new services, by abstracting the data center functionality behind a standardized set of APIs service can be deployed faster and more efficiently. MOSK enables you to define all your data center resources behind the industry standard OpenStack APIs allowing you to automate the deployment of applications or simply request resources through the UI to quickly and efficiently provision virtual machines, storage, networking, and other resources.
- Virtual Network Functions (VNFs)
VNFs require high performance systems that can be accessed on demand in a standardized way, with assurances that they will have access to the necessary resources and performance guarantees when needed. MOSK provides extensive support for VNF workload enabling easy access to functionality such as Intel EPA (NUMA, CPU pinning, Huge Pages) as well as the consumption of specialized networking interfaces cards to support SR-IOV and DPDK. The centralized management model of MOSK and Mirantis Container Cloud also enables the easy management of multiple MOSK deployments with full lifecycle management.
- Legacy workload migration
With the industry moving toward cloud-native technologies many older or legacy applications are not able to be moved easily and often it does not make financial sense to transform the applications to cloud-native applications. MOSK provides a stable cloud platform that can cost-effectively host legacy applications whilst still providing the expected levels of control, customization, and uptime.
Reference Architecture¶
Mirantis OpenStack for Kubernetes (MOSK) is a virtualization platform that provides an infrastructure for cloud-ready applications, in combination with reliability and full control over the data.
MOSK combines OpenStack, an open-source cloud infrastructure software, with application management techniques used in the Kubernetes ecosystem that include container isolation, state enforcement, declarative definition of deployments, and others.
MOSK integrates with Mirantis Container Cloud to rely on its capabilities for bare-metal infrastructure provisioning, Kubernetes cluster management, and continuous delivery of the stack components.
MOSK simplifies the work of a cloud operator by automating all major cloud life cycle management routines including cluster updates and upgrades.
Deployment profiles¶
A Mirantis OpenStack for Kubernetes (MOSK) deployment profile is a thoroughly tested and officially supported reference architecture that is guaranteed to work at a specific scale and is tailored to the demands of a specific business case, such as generic IaaS cloud, Network Function Virtualisation infrastructure, Edge Computing, and others.
A deployment profile is defined as a combination of:
Services and features the cloud offers to its users.
Non-functional characteristics that users and operators should expect when running the profile on top of a reference hardware configuration. Including, but not limited to:
Performance characteristics, such as an average network throughput between VMs in the same virtual network.
Reliability characteristics, such as the cloud API error response rate when recovering a failed controller node.
Scalability characteristics, such as the total amount of virtual routers tenants can run simultaneously.
Hardware requirements - the specification of physical servers, and networking equipment required to run the profile in production.
Deployment parameters that an operator for the cloud can tweak within a certain range without being afraid of breaking the cloud or losing support.
In addition, the following items may be included in a definition:
Compliance-driven technical requirements, such as TLS encryption of all external API endpoints.
Foundation-level software components, such as Tungsten Fabric or Open vSwitch as a backend for the networking service.
Note
Mirantis reserves the right to revise the technical implementation of any profile at will while preserving its definition - the functional and non-functional characteristics that operators and users are known to rely on.
MOSK supports a huge list of different deployment profiles to address a wide variety of business tasks. The table below includes the profiles for the most common use cases.
Note
Some components of a MOSK cluster are mandatory and are being installed during the managed cluster deployment by Container Cloud regardless of the deployment profile in use. StackLight is one of the cluster components that are enabled by default. See Container Cloud Operations Guide for details.
Profile |
OpenStackDeployment CR Preset |
Description |
|---|---|---|
Cloud Provider Infrastructure (CPI) |
|
Provides the core set of the services an IaaS vendor would need including some extra functionality. The profile is designed to support up 50-70 compute nodes and a reasonable number of storage nodes. 0 The core set of services provided by the profile includes: |
CPI with Tungsten Fabric |
|
A variation of the CPI profile 1 with Tugsten Fabric as a backend for networking. |
- 0
The supported node count is approximate and may vary depending on the hardware, cloud configuration, and planned workload.
- 1(1,2)
Ironic is an optional component for the CPI profile. See Bare Metal service for details.
- 2
Ironic is not supported for the CPI with Tungsten Fabric profile. See Tungsten Fabric known limitations for details.
Components overview¶
Mirantis OpenStack for Kubernetes (MOSK) includes the following key design elements.
HelmBundle Operator¶
The HelmBundle Operator is the realization of the Kubernetes Operator
pattern that provides a Kubernetes custom resource of the HelmBundle
kind and code running inside a pod in Kubernetes. This code handles changes,
such as creation, update, and deletion, in the Kubernetes resources of this
kind by deploying, updating, and deleting groups of Helm releases from
specified Helm charts with specified values.
OpenStack¶
The OpenStack platform manages virtual infrastructure resources, including virtual servers, storage devices, networks, and networking services, such as load balancers, as well as provides management functions to the tenant users.
Various OpenStack services are running as pods in Kubernetes and are
represented as appropriate native Kubernetes resources, such as
Deployments, StatefulSets, and DaemonSets.
For a simple, resilient, and flexible deployment of OpenStack and related services on top of a Kubernetes cluster, MOSK uses OpenStack-Helm that provides a required collection of the Helm charts.
Also, MOSK uses OpenStack Controller (Rockoon) as the
realization of the Kubernetes Operator pattern. Rockoon provides a custom
Kubernetes resource of the OpenStackDeployment kind and code running
inside a pod in Kubernetes. This code handles changes such as creation,
update, and deletion in the Kubernetes resources of this kind by
deploying, updating, and deleting groups of the Helm releases.
Ceph¶
Ceph is a distributed storage platform that provides storage resources, such as objects and virtual block devices, to virtual and physical infrastructure.
MOSK uses Rook as the implementation of the
Kubernetes Operator pattern that manages resources of the CephCluster
kind to deploy and
manage Ceph services as pods on top of Kubernetes to provide Ceph-based
storage to the consumers, which include OpenStack services, such as Volume
and Image services, and underlying Kubernetes through Ceph CSI (Container
Storage Interface).
The Ceph Controller is the implementation of the Kubernetes Operator
pattern, that manages resources of the MiraCeph kind to simplify
management of the Rook-based Ceph clusters.
StackLight Logging, Monitoring, and Alerting¶
The StackLight component is responsible for collection, analysis, and visualization of critical monitoring data from physical and virtual infrastructure, as well as alerting and error notifications through a configured communication system, such as email. StackLight includes the following key sub-components:
Prometheus
OpenSearch
OpenSearch Dashboards
Fluentd
Requirements¶
MOSK cluster hardware requirements¶
This section provides hardware requirements for the Mirantis Container Cloud management cluster with a managed Mirantis OpenStack for Kubernetes (MOSK) cluster.
For installing MOSK, the Mirantis Container Cloud management cluster and managed cluster must be deployed with baremetal provider.
Important
A MOSK cluster is to be used for a deployment of an OpenStack cluster and its components. Deployment of third-party workloads on a MOSK cluster is neither allowed nor supported.
Note
One of the industry best practices is to verify every new update or configuration change in a non-customer-facing environment before applying it to production. Therefore, Mirantis recommends having a staging cloud, deployed and maintained along with the production clouds. The recommendation is especially applicable to the environments that:
Receive updates often and use continuous delivery. For example, any non-isolated deployment of Mirantis Container Cloud.
Have significant deviations from the reference architecture or third party extensions installed.
Are managed under the Mirantis OpsCare program.
Run business-critical workloads where even the slightest application downtime is unacceptable.
A typical staging cloud is a complete copy of the production environment including the hardware and software configurations, but with a bare minimum of compute and storage capacity.
The table below describes the node types the MOSK reference architecture includes.
Node type |
Description |
|---|---|
Mirantis Container Cloud management cluster nodes |
The Container Cloud management cluster architecture on bare metal requires three physical servers for manager nodes. On these hosts, we deploy a Kubernetes cluster with services that provide Container Cloud control plane functions. |
OpenStack control plane node and StackLight node |
Host OpenStack control plane services such as database, messaging, API, schedulers conductors, and L3 and L2 agents, as well as the StackLight components. Note MOSK enables the cloud operator to collocate the OpenStack control plane with the managed cluster master nodes on the OpenStack deployments of a small size. This capability is available as technical preview. Use such configuration for testing and evaluation purposes only. |
Tenant gateway node |
Optional. Hosts OpenStack gateway services including L2, L3, and DHCP agents. The tenant gateway nodes are combined with OpenStack control plane nodes. The strict requirement is a dedicated physical network (bond) for tenant network traffic. |
Tungsten Fabric control plane node |
Required only if Tungsten Fabric is enabled as a backend for the OpenStack networking. These nodes host the TF control plane services such as Cassandra database, messaging, API, control, and configuration services. |
Tungsten Fabric analytics node |
Required only if Tungsten Fabric is enabled as a backend for the OpenStack networking. These nodes host the TF analytics services such as Cassandra, ZooKeeper, and collector. |
Compute node |
Hosts the OpenStack Compute services such as QEMU, L2 agents, and others. |
Infrastructure nodes |
Runs underlying Kubernetes cluster management services. The MOSK reference configuration requires minimum three infrastructure nodes. |
The table below specifies the hardware resources the MOSK reference architecture recommends for each node type.
Node type |
# of servers |
CPU cores # per server |
RAM per server, GB |
Disk space per server, GB |
NICs # per server |
|---|---|---|---|---|---|
Mirantis Container Cloud management cluster node |
3 0 |
16 |
128 |
1 SSD x 960
1 SSD x 1900 1
|
3 2 |
OpenStack control plane, gateway 3, and StackLight nodes |
3 or more |
32 |
128 |
1 SSD x 500
2 SSD x 1000 6
|
5 |
Tenant gateway (optional) |
0-3 |
32 |
128 |
1 SSD x 500 |
5 |
Tungsten Fabric control plane nodes 4 |
3 |
16 |
64 |
1 SSD x 500 |
1 |
Tungsten Fabric analytics nodes 4 |
3 |
32 |
64 |
1 SSD x 1000 |
1 |
Compute node |
3 (varies) |
16 |
64 |
1 SSD x 500 7 |
5 |
Infrastructure node (Kubernetes cluster management) |
3 8 |
16 |
64 |
1 SSD x 500 |
5 |
Infrastructure node (Ceph) 5 |
3 |
16 |
64 |
1 SSD x 500
2 HDDs x 2000
|
5 |
Note
The exact hardware specifications and number of the control plane and gateway nodes depend on a cloud configuration and scaling needs. For example, for the clouds with more than 12,000 Neutron ports, Mirantis recommends increasing the number of gateway nodes.
- 0
Adding more than 3 nodes to a management cluster is not supported.
- 1
In total, at least 2 disks are required:
disk0- system storage, minimum 120 GB.disk1- Container Cloud services storage, not less than 110 GB. The exact capacity requirements depend on StackLight data retention period.
See Management cluster storage for details.
- 2
OOB management (IPMI) port is not included.
- 3
OpenStack gateway services can optionally be moved to separate nodes.
- 4(1,2)
TF control plane and analytics nodes can be combined with a respective addition of RAM, CPU, and disk space to the hardware hosts. Though, Mirantis does not recommend such configuration for production environments as the risk of the cluster downtime if one of the nodes unexpectedly fails increases.
- 5
A Ceph cluster with 3 Ceph nodes does not provide hardware fault tolerance and is not eligible for recovery operations, such as a disk or an entire node replacement. Therefore, a minimum of 5 Ceph nodes is recommended for production use.
A Ceph cluster uses the replication factor that equals 3. If the number of Ceph OSDs is less than 3, a Ceph cluster moves to the degraded state with the write operations restriction until the number of alive Ceph OSDs equals the replication factor again.
- 6
1 SSD x 500 for operating system
1 SSD x 1000 for OpenStack LVP
1 SSD x 1000 for StackLight LVP
- 7
When Nova is used with local folders, additional capacity is required depending on the VM images size.
- 8
For nodes hardware requirements, refer to Container Cloud Reference Architecture: Managed cluster hardware configuration.
Note
If you would like to evaluate the MOSK capabilities and do not have much hardware at your disposal, you can deploy it in a virtual environment. For example, on top of another OpenStack cloud using the sample Heat templates.
Please mind, the tooling is provided for reference only and is not a part of the product itself. Mirantis does not guarantee its interoperability with any MOSK version.
Management cluster storage¶
The management cluster requires minimum two storage devices per node. Each device is used for different type of storage:
One storage device for boot partitions and root file system. SSD is recommended. A RAID device is not supported.
One storage device per server is reserved for local persistent volumes. These volumes are served by the Local Storage Static Provisioner, that is
local-volume-provisioner, and used by many services of Mirantis Container Cloud.
You can configure host storage devices using BareMetalHostProfile
resources. For details, see Create a custom bare metal host profile.
System requirements for the seed node¶
The seed node is only necessary to deploy the management cluster. When the bootstrap is complete, the bootstrap node can be discarded and added back to the MOSK cluster as a node of any type.
The minimum reference system requirements for a baremetal-based bootstrap seed node are as follow:
Basic Ubuntu 18.04 server with the following configuration:
Kernel of version 4.15.0-76.86 or later
8 GB of RAM
4 CPU
10 GB of free disk space for the bootstrap cluster cache
No DHCP or TFTP servers on any NIC networks
Routable access IPMI network for the hardware servers.
Internet access for downloading of all required artifacts
If you use a firewall or proxy, make sure that the bootstrap and management clusters have access to the following IP ranges and domain names:
IP ranges:
Microsoft Azure (only IP addresses for
MicrosoftContainerRegistry)Amazon AWS (only IP addresses for
"service": "CLOUDFRONT")
Domain names:
mirror.mirantis.com and repos.mirantis.com for packages
binary.mirantis.com for binaries and Helm charts
mirantis.azurecr.io and *.blob.core.windows.net for Docker images
mcc-metrics-prod-ns.servicebus.windows.net:9093 for Telemetry (port 443 if proxy is enabled)
mirantis.my.salesforce.com and login.salesforce.com for Salesforce alerts
Note
Access to Salesforce is required from any Container Cloud cluster type.
If any additional Alertmanager notification receiver is enabled, for example, Slack, its endpoint must also be accessible from the cluster.
Components collocation¶
MOSK uses Kubernetes labels to place components onto hosts. For the default locations of components, see MOSK cluster hardware requirements. Additionally, MOSK supports component collocation. This is mostly useful for OpenStack compute and Ceph nodes. For component collocation, consider the following recommendations:
When calculating hardware requirements for nodes, consider the requirements for all collocated components.
When performing maintenance on a node with collocated components, execute the maintenance plan for all of them.
When combining other services with the OpenStack compute host, verify that
reserved_host_*has increased accordingly to the needs of collocated components by using node-specific overrides for thecomputeservice.
Infrastructure requirements¶
This section lists the infrastructure requirements for the Mirantis OpenStack for Kubernetes (MOSK) reference architecture.
Service |
Description |
|---|---|
MetalLB |
MetalLB exposes external IP addresses of cluster services to access applications in a Kubernetes cluster. |
DNS |
The Kubernetes Ingress NGINX controller is used to expose OpenStack services outside of a Kubernetes deployment. Access to the Ingress services is allowed only by its FQDN. Therefore, DNS is a mandatory infrastructure service for an OpenStack on Kubernetes deployment. |
See also
Automatic upgrade of a host operating system¶
To keep operating system on a bare metal host up to date with the latest security updates, the operating system requires periodic software packages upgrade that may or may not require the host reboot.
Mirantis Container Cloud uses life cycle management tools to update the operating system packages on the bare metal hosts.
In a management cluster, software package upgrade and host restart are applied automatically when a new Container Cloud version with available kernel or software packages upgrade is released.
In a managed cluster, package upgrade and host restart are applied as part of usual cluster update, when applicable. To start planning the maintenance window and proceed with the managed cluster update, see Cluster update.
Operating system upgrade and host restart are applied to cluster nodes one by one. If Ceph is installed in the cluster, the Container Cloud orchestration securely pauses the Ceph OSDs on the node before restart. This allows avoiding degradation of the storage service.
Cloud services¶
Each section below is dedicated to a particular service provided by MOSK. They contain configuration details and usage samples of supported capabilities provided through the custom resources.
Core cloud services:
Compute service¶
Mirantis OpenStack for Kubernetes (MOSK) provides instances management capability through the Compute service (OpenStack Nova). The Compute service interacts with other OpenStack components of an OpenStack environment to provide life-cycle management of the virtual machine instances.
Resource oversubscription¶
The Compute service (OpenStack Nova) enables you to spawn instances that can collectively consume more resources than what is physically available on a compute node through resource oversubscription, also known as overcommit or allocation ratio.
Resources available for oversubscription on a compute node include the number of CPUs, amount of RAM, and amount of available disk space. When making a scheduling decision, the scheduler of the Compute service takes into account the actual amount of resources multiplied by the allocation ratio. Thereby, the service allocates resources based on the assumption that not all instances will be using their full allocation of resources at the same time.
Oversubscription enables you to increase the density of workloads and compute resource utilization and, thus, achieve better Return on Investment (ROI) on compute hardware. In addition, oversubscription can also help avoid the need to create too many fine-grained flavors, which is commonly known as flavor explosion.
Available since MOSK 23.1
There are two ways to control the oversubscription values for compute nodes:
The legacy approach entails utilizing the
{cpu,disk,ram}_allocation_ratioconfiguration options offered by the Compute service. A drawback of this method is that restarting the Compute service is mandatory to apply the new configuration. This introduces the risk of possible interruptions of cloud user operations, for example, instance build failures.The modern and recommended approach, adopted in MOSK 23.1, involves using the
initial_{cpu,disk,ram}_allocation_ratioconfiguration options, which are employed exclusively during the initial provisioning of a compute node. This may occur during the initial deployment of the cluster or when new compute nodes are added subsequently. Any further alterations can be performed dynamically using the OpenStack Placement service API without necessitating the restart of the service.
There is no definitive method for selecting optimal oversubscription values. As a cloud operator, you should continuously monitor your workloads, ideally have a comprehensive understanding of their nature, and experimentally determine the maximum values that do not impact performance. This approach ensures maximum workload density and cloud resource utilization.
To configure the initial compute resource oversubscription in
MOSK, specify the spec:features:nova:allocation_ratios
parameter in the OpenStackDeployment custom resource as explained in the
table below.
Parameter |
|
|---|---|
Configuration |
Configure initial oversubscription of CPU, disk space, and RAM resources on compute nodes. By default, the following values are applied:
Note In MOSK 22.5 and earlier, the effective
default value of RAM allocation ratio is Warning Mirantis strongly advises against oversubscribing RAM, by any amount. See Preventing resource overconsumption for details. Changing the resource oversubscription configuration through the
|
Usage |
Configuration example: kind: OpenStackDeployment
spec:
features:
nova:
allocation_ratios:
cpu: 8
disk: 1.6
ram: 1.0
Configuration example of setting different oversubscription values for specific nodes: spec:
nodes:
compute-type::hi-perf:
features:
nova:
allocation_ratios:
cpu: 2.0
disk: 1.0
In the example configuration above, the compute nodes labeled with
|
When using oversubscription, it is important to conduct thorough cloud management and monitoring to avoid system overloading and performance degradation. If many or all instances on a compute node start using all allocated resources at once and, thereby, overconsume physical resources, failure scenarios depend on the resource being exhausted.
Affected resource |
Symptoms |
|---|---|
CPU |
Workloads are getting slower as they actively compete for physical CPU usage. A useful indicator is the steal time as reported inside the workload, which is a percentage of time the operating system in the workload is waiting for actual physical CPU core availability to run instructions. To verify the steal time in the Linux-based workload, use the top command: top -bn1 | head | grep st$ | awk -F ',' '{print $NF}'
Generally, steal times of >10 for 20-30 minutes are considered alarming. |
RAM |
Operating system on the compute node starts to aggressively use physical swap space, which significantly slows the workloads down. Sometimes, when the swap is also exhausted, the operating system of a compute node can outright OOM kill most offending processes, which can cause major disruptions to workloads or a compute node itself. Warning While it may seem like a good idea to make the most of available resources, oversubscribing RAM can lead to various issues and is generally not recommended due to potential performance degradation, reduced stability, and security risks for the workloads. Mirantis strongly advises against oversubscribing RAM, by any amount. |
Disk space |
Depends on the physical layout of storage. Virtual root and ephemeral storage devices that are hosted on a compute node itself are put in the read-only mode negatively affecting workloads. Additionally, the file system used by the operating system on a compute node may become read-only too blocking the compute node operability. See also |
There are workload types that are not suitable for running in an oversubscribed environment, especially those with high performance, latency-sensitive, or real-time requirements. Such workloads are better suited for compute nodes with dedicated CPUs, ensuring that only processes of a single instance run on each CPU core.
Virtual CPU¶
MOSK provides the capability to configure virtual CPU types
for OpenStack instances through the OpenStackDeployment custom resource.
This feature enables cloud user to tailor performance and resource allocation
within their OpenStack environment to meet specific workload demands
effectively.
Parameter |
|
|---|---|
Usage |
Configures the type of virtual CPU that Nova will use when creating instances. The list of supported CPU models include |
The host-model CPU model (default) mimics the host CPU and provides for
decent performance, good security, and moderate compatibility with live
migrations.
With this mode, libvirt finds an available predefined CPU model that best matches the host CPU, and then explicitly adds the missing CPU feature flags to closely match the host CPU features. To mitigate known security flaws, libvirt automatically adds critical CPU flags, supported by installed libvirt, QEMU, kernel, and CPU microcode versions.
This is a safe choice if your OpenStack compute node CPUs are of the same generation. If your OpenStack compute node CPUs are sufficiently different, for example, span multiple CPU generations, Mirantis strongly recommends setting explicit CPU models supported by all of your OpenStack compute node CPUs or organizing your OpenStack compute nodes into host aggregates and availability zones that have largely identical CPUs.
Note
The host-model model does not guarantee two-way live migrations
between nodes.
When migrating instances, the libvirt domain XML is first copied as is to the destination OpenStack compute node. Once the instance is hard rebooted or shut down and started again, the domain XML will be re-generated. If versions of libvirt, kernel, CPU microcode, or BIOS firmware differ from what they were on the source compute node the instance was started before, libvirt may pick up additional CPU feature flags, making it impossible to live-migrate back to the original compute node.
The host-passthrough CPU model provides maximum performance, especially
when nested virtualization is required or if live migration support is not
a concern for workloads. Live migration requires exactly the same CPU
on all OpenStack compute nodes, including the CPU microcode and kernel
versions. Therefore, for live migrations support, organize your compute
nodes into host aggregates and availability zones. For workload migration
between non-identical OpenStack compute nodes, contact Mirantis support.
For example, to set the host-passthrough CPU model for all OpenStack
compute nodes:
spec:
features:
nova:
vcpu_type: host-passthrough
MOSK enables you to specify a comma-separated list of exact QEMU CPU models to create and emulate. Specify the common and less advanced CPU models first. All explicit CPU models provided must be compatible with the OpenStack compute node CPUs.
To specify an exact CPU model, review the available CPU models and their
features. List and inspect the /usr/share/libvirt/cpu_map/*.xml files in
the libvirt containers of pods of the libvirt DeamonSet or multiple
DaemonSets if you are using node-specific settings.
To review the available CPU models
Identify the available libvirt DaemonSets:
kubectl -n openstack get ds -l application=libvirt --show-labels
Example of system response:
NAME DESIRED CURRENT READY UP-TO-DATE AVAILABLE NODE SELECTOR AGE LABELS libvirt-libvirt-default 2 2 2 2 2 openstack-compute-node=enabled 34d app.kubernetes.io/managed-by=Helm,application=libvirt,component=libvirt,release_group=openstack-libvirt
Identify the pods of libvirt DaemonSets:
kubectl -n openstack get po -l application=libvirt,release_group=openstack-libvirt
Example of system response:
NAME READY STATUS RESTARTS AGE libvirt-libvirt-default-5zs8m 2/2 Running 0 8d libvirt-libvirt-default-vt8wd 2/2 Running 0 3d14h
List and review the available CPU model definition files. For example:
kubectl -n openstack exec -ti libvirt-libvirt-default-5zs8m -c libvirt -- ls /usr/share/libvirt/cpu_map/*.xml
List and review the content of all CPU model definition files. For example:
kubectl -n openstack exec -ti libvirt-libvirt-default-5zs8m -c libvirt -- bash -c 'for f in `ls /usr/share/libvirt/cpu_map/*.xml`; do echo $f; cat $f; done'
For example, for nodes that are labeled with processor=amd-epyc, set
a custom EPYC CPU model:
spec:
nodes:
processor::amd-epyc:
features:
nova:
vcpu_type: EPYC
Instance migration¶
OpenStack supports the following types of instance migrations:
- Cold migration (also referred to simply as migration)
The process involves shutting down the instance, copying its definition and disk, if necessary, to another host, and then starting the instance again on the new host.
This method disrupts the workload running inside the instance but allows for more reliability and works for most types of instances and consumed resources.
- Live migration
The process involves copying the instance definition, memory, and disk, if necessary, to another host while the instance continues running, without shutting it down. The instance then momentarily switches to run on the new host.
While generally less disruptive to workloads, this method is less reliable and imposes more restrictions on the instance and target host properties to succeed.
As a cloud operator, you can configure live migration through the
OpenStackDeployment custom resource. The following table provides
the details on available configuration.
Parameter |
Usage |
|---|---|
|
Specifies the name of the NIC device on the actual host that will be used by Nova for the live migration of instances. Mirantis recommends setting up your Kubernetes hosts in such a way that networking is configured identically on all of them, and names of the interfaces serving the same purpose or plugged into the same network are consistent across all physical nodes. Also, set the option to
|
|
Available since MOSK 23.2.
If set to spec:
features:
nova:
libvirt:
tls:
enabled: true
See also Encryption of live migration data. |
Available since MOSK 24.3
MOSK provides the following distinct sets of policies that govern access to cold and live migrations:
os_compute_api:os-migrate-server:migrateandos_compute_api:os-migrate-server:migrate_livedefine the ability to initiate migrations without specifying the target host. In this case, the OpenStack Compute scheduler selects the best suited target host automatically.os_compute_api:os-migrate-server:migrate:hostandos_compute_api:os-migrate-server:migrate_live:hostdefine the ability to initiate migration together with specifying the target host. Depending on the API microversion used to start the migration, the host is either validated by the scheduler (recommended) or forced regardless of other considerations. The latter option is not recommended as it may lead to inconsistencies in the internal state of the Compute service.
Since MOSK 24.3, the default policies for migrations without
the target host specification is set to rule: project_member_or_admin.
This means that migration is available to both cloud administrators and
project users with the member role.
The migration to a specific host requires administrative privileges.
If the default policy does not suit your deployment, you can require
administrative access for all instance migrations by setting these policy
values to rule:context_is_admin, or any other value appropriate for your
use case.
If you use the default policies and want to revert to the old defaults, ensure
that the following snippet is present in your OpenStackDeployment custom
resource:
kind: OpenStackDeployment
spec:
features:
policies:
nova:
os_compute_api:os-migrate-server:migrate: rule:context_is_admin
os_compute_api:os-migrate-server:migrate_live: rule:context_is_admin
Image storage backend¶
Parameter |
|
|---|---|
Usage |
Defines the type of storage for Nova to use on the compute hosts for the images that back up the instances. The list of supported options include:
|
Remote console access to virtual machines¶
MOSK provides a number of different methods to interact
with OpenStack virtual machines including VNC (default) and SPICE remote
consoles. This section outlines how you can configure these different
console services through the OpenStackDeployment custom resource.
The noVNC client provides remote control or remote desktop access to guest virtual machines through the Virtual Network Computing (VNC) system. The MOSK Compute service users can access their instances using the noVNC clients through the noVNC proxy server.
The VNC remote console is enabled by default in MOSK.
To disable VNC remote console through the OpenStackDeployment custom
resource, set spec:features:nova:console:novnc to false:
spec:
features:
nova:
console:
novnc:
enabled: false
Available since MOSK 23.1
MOSK uses TLS to secure public-facing VNC access on networks between a noVNC client and noVNC proxy server.
The features:nova:console:novnc:tls:enabled ensures that the data
transferred between the instance and the noVNC proxy server is encrypted.
Both servers use the VeNCrypt authentication scheme for the data
encryption.
To enable the encrypted data transfer for noVNC, use the following
structure in the OpenStackDeployment custom resource:
kind: OpenStackDeployment
spec:
features:
nova:
console:
novnc:
tls:
enabled: true
Available since MOSK 24.1 TechPreview
The VNC protocol has its limitations, such as the lack of support for multiple monitors, bi-directional audio, reliable cut-and-paste, video streaming, and others. The SPICE protocol aims to overcome these limitations and deliver a robust remote desktop support.
The SPICE remote console is disabled by default in MOSK.
To enable SPICE remote console through the OpenStackDeployment custom
resource, set spec:features:nova:console:spice:enabled to true:
spec:
features:
nova:
console:
spice:
enabled: true
GPU virtualization¶
Available since MOSK 24.1 TechPreview
MOSK provides GPU virtualization capabilities to its users through the NVIDIA vGPU and Multi-Instance GPU (MIG) technologies.
GPU virtualization is a capability offered by modern datacenter-grade GPUs, enabling the partitioning of a single physical GPU into smaller virtual devices, that can then be attached to individual virtual machines.
In contrast to the Peripheral Component Interconnect (PCI) passthrough feature, leveraging the GPU virtualization enables concurrent utilization of the same physical GPU device by multiple virtual machines. This enhances hardware utilization and fosters a more elastic consumption of expensive hardware resources.
When using GPU virtualization, the physical device and its drivers manage computing resource partitioning and isolation.
The use case for GPU virtualization aligns with any application necessitating or benefiting from accelerated parallel floating-point calculations, such as graphic-intensive desktop workloads, for example, 3D modeling and rendering, as well as computationally intensive tasks, for example, artifial intelligence, specifically, machine learning training and classification.
At its core, GPU virtualization operates on base of the single-root input/output virtualization framework (SR-IOV), which is already widely used by datacenter-grade network adapters and mediated devices Linux kernel framework.
Typically, using GPU virtualization requires the installation of specific physical GPU drivers on the host system. For detailed instructions on obtaining and installing the required drivers, refer to official documentation from the vendor of your GPU.
For the latest family of NVIDIA GPUs under NVIDIA AI Enterprise, start with NVIDIA AI Enterprise documentation.
You can automate the configuration of drivers by adding a custom post-install
script to the BareMetalHostProfile object of your
MOSK cluster. See Configure GPU virtualization for details.
Certain NVIDIA GPUs, for example, Ampere GPU architecture and later, support GPU virtualization in two modes: time sliced (vGPU) or Multi-Instance GPU (MIG). Older architectures support only the time-sliced mode.
The distinction between these modes lies in resource isolation, dedicated performance levels, and partitioning flexibility.
Typically, there is no fixed rule dictating which mode should be used, as it depends on the intended workloads for the virtual GPUs and the level of experience and assurances the cloud operator aims to offer users. Below, there is a brief overview of the differences between these two modes.
In time-sliced vGPU mode, each virtual GPU is allocated dedicated slices of the physical GPU memory while sharing the physical GPU engines. Only one vGPU operates at a time, with full access to all physical GPU engines. The resource scheduler within the physical GPU regulates the timing of each vGPU execution, ensuring fair allocation of resources.
Therefore, this setup may encounter issues with noisy neighbors, where the performance of one vGPU is affected by resource contention from others. However, when not all available vGPU slots are occupied, the active ones can fully utilize the power of its physical GPU.
Advantages:
Potential ability to fully utilize the compute power of physical GPU, even if not all possible vGPUs have yet been created on that physical GPU.
Easier configuration.
Disadvantages:
Only a single vGPU type (size of the vGPU) can be created on any given physical GPU. The cloud operator must decide beforehand what type of vGPU each physical GPU will be providing.
Less strict resource isolation. Noisy neighbors and unpredictable level of performance for every single guest vGPU.
In Multi-Instance GPUs (MIG) mode, each virtual GPU is allocated dedicated physical GPU engines, exclusively utilized by that specific virtual GPU. Virtual GPUs run in parallel, each on its own engines according to their type.
Advantages:
Ability to partition a single physical GPU into various types of virtual GPUs. This approach provides cloud operators with enhanced flexibility in determining the available vGPU types for cloud users. However, the cloud operator has to decide beforehand what types of virtual GPU each physical GPU will be providing and partition each GPU accordingly.
Better resource isolation and guaranteed resource access with predictable performance levels for every virtual GPU.
Disadvantages:
Under-utilization of physical GPU when not all possible virtual GPU slots are occupied.
Comparatively complicated configuration, especially in heterogeneous hardware environments.
Note
Some of these restrictions may be lifted in future releases of MOSK.
Cloud users will face the following limitations when working with GPU virtualization in MOSK:
Inability to create several instances with virtual GPUs in one request if there is no physical GPU available that can fit all of them at once. For NVIDIA MIG, this effectively means that you cannot create several instances with virtual GPUs in one request.
Inability to create an instance with several virtual GPUs.
Inability to attach virtual GPU to or detach virtual GPU from a running instance.
Inability to live-migrate instances with virtual GPU attached.
Cloud operator will face the following limitations when configuring GPU virtualization in MOSK:
Partition of physical GPUs to virtual GPUs is static and not on-demand. You need to decide beforehand what types of virtual GPUs each physical GPU will get partinioned into. Changing of the partitioning requires removing all instances using virtual GPUs from the compute node before initiating the repartitioning process.
Repartitioning may require additional manual steps to eliminate orphan resource providers in the placement service, and thus, avoid resource over-reporting and instance scheduling problems.
Configuration of multiple virtual GPU types per node may be very verbose since configuration depends on particular PCI addresses of physical GPUs on each node.
See also
Learn more
Graceful instance shutdown¶
Available since MOSK 24.3
Management of compute node reboots is an important Day 2 operation. Before shutting down a host, guest instances must either be migrated to other compute nodes or gracefully powered off. This ensures the integrity of disk filesystems and prevents damage to running applications.
MOSK provides the capability to automatically power off the instances during the compute node shutdown or reboot through the ACPI power event.
Graceful instance shutdown is managed using the systemd inhibit
tool. When the nova-compute service starts, it creates locks. For example:
systemd-inhibit --list
Example system response:
WHO UID USER PID COMM WHAT WHY MODE
Nova Shutdown Handler 0 root 28927 python3 shutdown Handle events on shutdown notification delay
The process runs in the nova-compute-inhibit-lock container within
the nova-compute pod. It intercepts systemd power event and starts
graceful guest shutdown. When all guest instances are powered off,
the inhibit lock is released.
To initiate a proper shutdown, use the following commands: systemctl shutdown and systemctl reboot.
Networking service¶
Mirantis OpenStack for Kubernetes (MOSK) Networking service (OpenStack Neutron) provides cloud applications with Connectivity-as-a-Service enabling instances to communicate with each other and the outside world.
The API provided by the service abstracts all the nuances of implementing a virtual network infrastructure on top of your own physical network infrastructure. The service allows cloud users to create advanced virtual network topologies that may include load balancing, virtual private networking, traffic filtering, and other services.
MOSK Networking service supports Open vSwitch and Tungsten Fabric SDN technologies as backends.
Backends¶
MOSK offers various networking backends. Selecting the appropriate backend option for the Networking service is essential for building a robust and efficient cloud networking infrastructure. Whether you choose Open vSwitch (OVS), Open Virtual Network (OVN), or Tungsten Fabric, understanding their features, capabilities, and suitability for your specific use case is crucial for achieving optimal performance and scalability in your OpenStack environment.
Refer to Networking backend configuration for the configuration details.
Capability |
Tungsten Fabric |
Open vSwitch (OVS) |
Open Virtual Network (OVN) |
|---|---|---|---|
Logical routers |
|||
Static routes |
|||
SNAT |
|||
Floating IPs |
|||
External IPs on VMs |
|||
Per-tenant floating networks and SNAT pools |
|||
IPv6 |
|||
Bare Metal as a Service (Ironic) |
|||
DNS as a Service |
Designate and Tungsten Fabric vDNS |
Designate |
Designate |
Firewalling |
Security groups and application policies |
OVS firewall |
OVS firewall |
Load balancing |
Tungsten Fabric built in HAProxy, OpenStack Octavia/Amphora |
OpenStack Octavia/Amphora |
OpenStack Octavia/Amphora, Octavia/OVN native load balancer |
BGP VPNs |
TechPreview |
||
VPN as a Service (IPsec) |
TechPreview |
TechPreview |
|
Data plane acceleration |
SR-IOV/DPDPK |
SR-IOV/DPDK |
SR-IOV/DPDK |
QoS |
|||
Network equipment management |
Netconf/OVSDB |
Neutron ML2 plugins/networking-generic-switch |
Neutron ML2 plugins/networking-generic-switch |
East-West traffic encryption |
Open vSwitch is a production-quality, multilayer virtual switch licensed under the open source Apache 2.0 license. It is designed to enable massive network automation through programmatic extension, while supporting standard management interfaces and protocols.
Open vSwitch is suitable for general-purpose networking requirements in OpenStack deployments. It provides flexibility and scalability for various network topologies.
Key characteristics of Open vSwitch:
Depends on RabbitMQ and RPC communication
Uses keepalived to set up HA routers
Uses namespace and Veth routing to provide its capabilities
Locates metadata in router or DHCP namespaces
Centralizes the DHCP service, which is running in a separate namespace
Available since MOSK 25.1 as GA (Caracal) Available since MOSK 24.2 as TechPreview (Antelope)
Open Virtual Network is a solution for Open vSwitch that provides native virtual networking support for Open vSwitch environments. It provides enhanced scalability and performance compared to traditional Open vSwitch deployments.
Key characteristics of Open Virtual Network:
Uses the OVSDB protocol for commmunication
Is distributed by design
Handles all traffic with OpenFlow
Runs metadata on all nodes
Provides DHCP through local Open vSwitch instances
Caution
There are numerous limitations related to VLAN/Flat tenant networks in Open Virtual Network with distributed floating IPs for bare metal SR-IOV and Octavia VIP ports. For more information about Open Virtual Network limitations, see relevant upstream documentation.
OpenStack official documentation
Tungsten Fabric is an open-source SDN based on Juniper Contrail. Its design allows for simplified creation and management of virtual networks in cloud environments. Tungsten Fabric supports advanced networking scenarious, such as BGP integration and scalability.
Key characteristics of Tungsten Fabric:
Uses well scalable protocols to set up tunnels, such as BGP/MPLS
Provides out-of-the-box BGPaaS/Service chaining capabilities
General configuration¶
MOSK offers the Networking service as a part of its
core setup. You can configure the service through the
spec:features:neutron section of the OpenStackDeployment custom
resource.
Parameter |
|
|---|---|
Usage |
Defines the networking backend. The list of supported options includes:
Refer to Backends to learn more about the networking backends supported by MOSK. |
Parameter |
|
|---|---|
Usage |
Defines the name of the NIC device on the actual host that will be used for Neutron. Mirantis recommends setting up your Kubernetes hosts in such a way that networking is configured identically on all of them, and names of the interfaces serving the same purpose or plugged into the same network are consistent across all physical nodes. |
Parameter |
|
|---|---|
Usage |
Defines the list of IPs of DNS servers that are accessible from virtual networks. Used as default DNS servers for VMs. |
Parameter |
|
|---|---|
Usage |
Contains the data structure that defines external (provider) networks on top of which the Neutron networking will be created. |
Parameter |
|
|---|---|
Usage |
If enabled, must contain the data structure defining the floating IP network that will be created for Neutron to provide external access to your Nova instances. |
BGP dynamic routing¶
Available since MOSK 23.2 TechPreview
The BGP dynamic routing extension to the Networking service (OpenStack Neutron) is particularly useful for the MOSK clouds where private networks managed by cloud users need to be transparently integrated into the networking of the data center.
For example, the BGP dynamic routing is a common requirement for IPv6-enabled environments, where clients need to seamlessly access cloud workloads using dedicated IP addresses with no address translation involved in between the cloud and the external network.
BGP dynamic routing changes the way self-service (private) network prefixes are communicated to BGP-compatible physical network devices, such as routers, present in the data center. It eliminates the traditional reliance on static routes or ICMP-based advertising by enabling the direct passing of private network prefix information to router devices.
Note
To effectively use the BGP dynamic routing feature, Mirantis recommends acquiring good understanding of OpenStack address scopes and how they work.
The components of the OpenStack BGP dynamic routing are:
- Service plugin
An extension to the Networking service (OpenStack Neutron) that implements the logic for BGP-related entities orhestration and provides the cloud user-facing API. A cloud administrator creates and configures a BGP speaker using the CLI or API and manually schedules it to one or more hosts running the agent.
- Agent
Manages BGP peering sessions. In MOSK, the BGP agent runs on nodes labeled with
openstack-gateway=enabled.
Prefix advertisement depends on the binding of external networks to a BGP speaker and the address scope of external and internal IP address ranges or subnets.
BGP dynamic routing advertises prefixes for self-service networks and host routes for floating IP addresses.
To successfully advertise a self-service network, you need to fulfill the following conditions:
External and self-service networks reside in the same address scope.
The router contains an interface on the self-service subnet and a gateway on the external network.
The BGP speaker associates with the external network that provides a gateway on the router.
The BGP speaker has the
advertise_tenant_networksattribute set toTrue.
To successfully advertise a floating IP address, you need to fulfill the following conditions:
The router with the floating IP address binding contains a gateway on an external network with the BGP speaker association.
The BGP speaker has the
advertise_floating_ip_host_routesattribute set totrue.
The diagram below is an example of the BGP dynamic routing in the non-DVR mode with self-service networks and the following advertisements:
B>* 192.168.0.0/25 [200/0]through10.11.12.1B>* 192.168.0.128/25 [200/0]through10.11.12.2B>* 10.11.12.234/32 [200/0]through10.11.12.1
For both floating IP and IPv4 fixed IP addresses, the BGP speaker advertises the gateway of the floating IP agent on the corresponding compute node as the next-hop IP address. When using IPv6 fixed IP addresses, the BGP speaker advertises the DVR SNAT node as the next-hop IP address.
The diagram below is an example of the BGP dynamic routing in the DVR mode with self-service networks and the following advertisements:
B>* 192.168.0.0/25 [200/0]through10.11.12.1B>* 192.168.0.128/25 [200/0]through10.11.12.2B>* 10.11.12.234/32 [200/0]through10.11.12.12
DVR incompatibility with ARP announcements and VRRP¶
Due to the known issue #1774459 in the upstream implementation, Mirantis does not recommend using Distributed Virtual Routing (DVR) routers in the same networks as load balancers or other applications that utilize the Virtual Router Redundancy Protocol (VRRP) such as Keepalived. The issue prevents the DVR functionality from working correctly with network protocols that rely on the Address Resolution Protocol (ARP) announcements such as VRRP.
The issue occurs when updating permanent ARP entries for
allowed_address_pair IP addresses in DVR routers because DVR performs
the ARP table update through the control plane and does not allow any
ARP entry to leave the node to prevent the router IP/MAC from
contaminating the network.
This results in various network failover mechanisms not functioning in virtual networks that have a distributed virtual router plugged in. For instance, the default backend for MOSK Load Balancing service, represented by OpenStack Octavia with the OpenStack Amphora backend when deployed in the HA mode in a DVR-connected network, is not able to redirect the traffic from a failed active service instance to a standby one without interruption.
Block Storage service¶
Mirantis OpenStack for Kubernetes (MOSK) provides volume management capability through the Block Storage service (OpenStack Cinder).
Backup configuration¶
MOSK provides support for the following backends for the Block Storage service (OpenStack Cinder):
Backend |
Support status |
|---|---|
Ceph |
Full support, default |
NFS |
|
S3 |
|
In MOSK, Cinder backup is enabled and uses the Ceph back
end for Cinder by default. The backup configuration is stored
in the spec:features:cinder:backup structure in the
OpenStackDeployment custom resource. If necessary, you can disable
the backup feature in Cinder as follows:
kind: OpenStackDeployment
spec:
features:
cinder:
backup:
enabled: false
Using this structure, you can also configure another backup driver supported by MOSK for Cinder as described below. At any given time, only one backend can be enabled.
Available since MOSK 23.2 TechPreview
MOSK supports NFS Unix authentication exclusively.
To use an NFS driver with MOSK, ensure you have
a preconfigured NFS server with an NFS share accessible to a Unix
Cinder user. This user must be the owner of the exported NFS folder,
and the folder must have the permission value set to 775.
All Cinder services run with the same user by default. To obtain the Unix user ID:
kubectl -n openstack get pod -l application=cinder,component=api -o jsonpath='{.items[0].spec.securityContext.runAsUser}'
Note
The NFS server must be accessible through the network from all OpenStack control plane nodes of the cluster.
To enable the NFS storage for Cinder backup, configure the following
structure in the OpenStackDeployment object:
spec:
features:
cinder:
backup:
drivers:
<BACKEND_NAME>:
type: nfs
enabled: true
backup_share: <URL_TO_NFS_SHARE>
You can specify the backup_share parameter in following formats:
hostname:path, ipv4addr:path, or [ipv6addr]:path.
For example: 1.2.3.4:/cinder_backup.
Available since MOSK 23.2 TechPreview
To use an S3 driver with MOSK, ensure you have a preconfigured S3 storage with a user account created for access.
Note
The S3 storage must be accessible through the network from all OpenStack control plane nodes of the cluster.
To enable the S3 storage for Cinder backup:
Create a dedicated secret in Kuberbetes to securely store the credentials required for accessing the S3 storage:
--- apiVersion: v1 kind: Secret metadata: labels: openstack.lcm.mirantis.com/osdpl_secret: "true" name: cinder-backup-s3-hidden namespace: openstack type: Opaque data: access_key: <ACCESS_KEY_FOR_S3_ACCOUNT> secret_key: <ACCESS_KEY_FOR_S3_ACCOUNT>
Configure the following structure in the
OpenStackDeploymentobject:spec: features: cinder: backup: drivers: <BACKEND_NAME>: type: s3 enabled: true endpoint_url: <URL_TO_S3_STORAGE> store_bucket: <S3_BUCKET_NAME> store_access_key: value_from: secret_key_ref: key: access_key name: cinder-backup-s3-hidden store_secret_key: value_from: secret_key_ref: key: secret_key name: cinder-backup-s3-hidden
Volume encryption¶
TechPreview
The Block Storage service (OpenStack Cinder) supports volume encryption using a key stored in the Key Manager service (OpenStack Barbican). Such configuration uses Linux Unified Key Setup (LUKS) to create an encrypted volume type and attach it to the Compute service (OpenStack Nova) instances. Nova retrieves the asymmetric key from Barbican and stores it on the OpenStack compute node as a libvirt key to encrypt the volume locally or on the backend and only after that transfers it to Cinder.
Note
To create an encrypted volume under a non-admin user, the
creatorrole must be assigned to the user.When planning your cloud, consider that encryption may impact CPU.
Volume configuration¶
The MOSK Block Storage service (OpenStack Cinder) uses Ceph
as the default backend for Cinder Volume. Also, MOSK enables
its clients to define their own volume backends using the
OpenStackDeployment custom resource. This section provides all the details
required to properly configure a custom Cinder Volume backend as a StatefulSet
or a DaemonSet.
MOSK stores the configuration for the default Ceph backend
in the spec:features:cinder:volume structure in the OpenStackDeployment
custom resource.
To disable the Ceph backend for Cinder Volume, modify the
spec:features:cinder:volume structure as follows:
spec:
features:
cinder:
volume:
enabled: false
services:
block-storage:
cinder:
values:
conf:
DEFAULT:
default_volume_type: <NEW-DEFAULT-VOLUME-TYPE-NAME>
When disabling the Ceph backend for Cinder Volume, you must explicitly specify
the new default_volume_type parameter. Refer to the sections below to learn
how you can configure it.
Before you start deploying your custom Cinder Volume backend, decide on key backend parameters and understand how they affect other services:
Note
Make sure to navigate to the documentation for the specific OpenStack version used to deploy your environment when referring to the official OpenStack documentation.
In addition, you may need to build your own Cinder image as described in Customize OpenStack container images.
Next, review the following key considerations:
Configuration option |
Details |
|---|---|
StatefulSet or DaemonSet |
If the Cinder volume backend you prefer must run on all nodes with a specific label and scale automatically as nodes are added or removed, use a DaemonSet. This type of backend typically requires that its data remains on the same node where its pod is running. A common example of such a backend is the LVM backend. Otherwise, Mirantis recommends using a StatefulSet, which offers more flexibility than a DaemonSet. |
Support for Active/Active High Availability |
If the driver does not support Active/Active High Availability, ensure
that only a single copy of the backend runs and that the When deploying the backend using a StatefulSet, set
|
Support for Multi-Attach |
If the driver supports Multi-Attach, it allows multiple connections to the same volume. This capability is important for certain services, such as Glance. If the driver does not support Multi-Attach, the backend cannot be used for services that require this functionality. |
Support for iSCSI and access to the |
Some drivers require access to the |
Privileged access for the container |
For security reasons, Mirantis recommends running the Cinder Volume
backend container with the minimum required privileges. However,
if the drivers require privileged access, you can enable it for the
StatefulSet by setting the parameter
|
Access to the host network namespace |
If the driver requires access to the host network namespace, or if you
need to ensure that the Cinder Volume backend’s IP address remains
unchanged after pod recreation or restart, set hostNetwork
to
|
Access to the host IPC namespace |
If the driver requires access to the host’s IPC namespace, set hostIPC
to
|
Access to host PID namespace |
If the driver requires access to the host’s PID namespace, set hostPID
to
|
Available since MOSK 24.3 TechPreview
MOSK enables its clients to define volume backends as a StatefulSet.
To configure a custom StatefulSet backend for the MOSK
Block Storage service (OpenStack Cinder), use the
spec:features:cinder:volume:backends structure in the
OpenStackDeployment custom resource:
spec:
features:
cinder:
volume:
backends:
<UNIQUE_BACKEND_NAME>:
enabled: true
type: statefulset
create_volume_type: true
values:
conf:
images:
labels:
pod:
The enabled and create_volume_type parameters are optional. With
create_volume_type set to true (default), the new backend
will be added to the Cinder bootstrap job. Once this job is completed,
the volume type for the custom backend will be created in OpenStack.
The supported value for type is statefulset.
The list of keys you can override in the values.yaml file of the Cinder
chart includes conf, images, labels, and pod.
When you define the custom backend for the Block Storage service, MOSK deploys individual pods for it. These pods have separate Secrets for configuration files and ConfigMaps for scripts.
Example of configuration of a custom StatefulSet backend for Cinder:
The configuration example deploys a StatefulSet for the Cinder
volume backend that uses the NFS driver, running a single replica on node
labeled kubernetes.io/hostname:service-node. Privilege escalation for
the Cinder volume pod is driver-specific.
spec:
features:
cinder:
volume:
enabled: false
backends:
nfs-volume:
type: statefulset
values:
conf:
cinder:
DEFAULT:
cluster: ""
enabled_backends: volumes-nfs
volumes-extra-nfs:
nas_host: 1.2.3.4
nas_share_path: /cinder_volume
nas_secure_file_operations: false
nfs_mount_point_base: /tmp/mountpoints
nfs_snapshot_support: true
volume_backend_name: volumes-nfs
volume_driver: cinder.volume.drivers.nfs.NfsDriver
pod:
replicas:
volume: 1
security_context:
cinder_volume:
container:
cinder_volume:
privileged: true
labels:
volume:
node_selector_key: kubernetes.io/hostname
node_selector_value: service-node
services:
block-storage:
cinder:
values:
conf:
DEFAULT:
default_volume_type: volumes-nfs
TechPreview
MOSK enables its clients to define volume backends as a DaemonSet, LVM in particular.
To configure a custom DaemonSet backend for the MOSK
Block Storage service (OpenStack Cinder), use the spec:nodes structure
in the OpenStackDeployment custom resource:
spec:
nodes:
<node label>:
features:
cinder:
volume:
backends:
<backend name>:
lvm:
<CINDER-LVM-DRIVER-PARAMETERS>
Example of configuration of a custom DaemonSet backend for Cinder:
The configuration example deploys a DaemonSet for the Cinder volume backend
that uses the LVM driver and runs on nodes with the
openstack-compute-node=enabled label:
Caution
For data storage, this backend uses the LVM cinder-vol
group that must be present on nodes before the new backend is applied.
For the procedure on how to deploy an LVM backend, refer to
Enable LVM block storage.
spec:
features:
cinder:
volume:
enabled: false
nodes:
openstack-compute-node::enabled:
features:
cinder:
volume:
backends:
volumes-lvm:
lvm:
volume_group: "cinder-vol"
services:
block-storage:
cinder:
values:
conf:
DEFAULT:
default_volume_type: volumes-lvm
MOSK provides the cinder-service-cleaner CronJob
by default. This CronJob periodically checks whether all Cinder services in
OpenStack are up to date and removes any stale ones.
This CronJob is tested only with backends supported by MOSK.
If cinder-service-cleaner does not work properly with your custom Cinder
volume backend, you can disable it at the OpenStackDeployment service
level in the OpenStackDeployment custom resource:
spec:
services:
block-storage:
cinder:
values:
manifests:
cron_service_cleaner: false
Note
Make sure to navigate to the documentation for the specific OpenStack version used to deploy your environment when referring to the official OpenStack documentation.
Identity service¶
Mirantis OpenStack for Kubernetes (MOSK) provides authentication, service discovery, and distributed multi-tenant authorization through the OpenStack Identity service, aka Keystone.
Federation¶
MOSK integrates with Mirantis Container Cloud Identity and Access Management (IAM) subsystem to allow centralized management of users and their permissions across multiple clouds.
The core component of Container Cloud IAM is Keycloak, the open-source identity and access management software. Its primary function is to perform secure authentication of cloud users against its built-in or various external identity databases, such as LDAP directories, OpenID Connect or SAML compatible identity providers.
By default, every MOSK cluster is integrated with the
Keycloak running in the Container Cloud management cluster. The integration
automatically provisions the necessary configuration on the
MOSK and Container Cloud IAM sides, such as the os
client object in Keycloak. However, for the federated users to get proper
permissions after logging in, the cloud operator needs to define the role
mapping rules specific to each MOSK environment.
See also
MOSK enables you to connect to the Keycloak identity
provider through the following structure in the OpenStackDeployment
custom resource:
spec:
features:
keystone:
keycloak:
enabled: true
url: https://keycloak.it.just.works
oidc:
OIDCSSLValidateServer: false
OIDCOAuthSSLValidateServer: false
OIDCScope: "openid email profile groups"
Available since MOSK 24.3 TechPreview
MOSK enables you to connect external identity
provider to Keystone directly through the following structure in the
OpenStackDeployment custom resource:
spec:
features:
keystone:
federations:
openid:
enabled: true
oidc_auth_type: oauth2
providers:
keycloak:
issuer: https://keycloak.it.just.works/auth/realms/iam
mapping:
- local:
- user:
email: '{1}'
name: '{0}'
- domain:
name: Default
groups: '{2}'
remote:
- type: OIDC-iam_username
- type: OIDC-email
- type: OIDC-iam_roles
metadata:
client:
client_id: os
conf:
response_type: id_token
scope: openid email profile
ssl_validate_server: false
provider:
value_from:
from_url:
url: https://keycloak.it.just.works/auth/realms/iam/.well-known/openid-configuration
okta:
description: OKTA provider
enabled: true
issuer: https://dev-68495932.okta.com/oauth2/default
mapping:
- local:
- user:
email: '{1}'
name: '{0}'
- domain:
name: Default
groups: m:os@admin
remote:
- type: OIDC-name
- type: OIDC-email
metadata:
client:
client_id: 0oaixfwyqcAkCbC335d7
client_secret: aKOtnqHwu37ricQJfOD9ShECqj7DY7SVHgh8nm1NwlAhGbQjGqREHencsGagyfmQ
conf: {}
provider:
value_from:
from_url:
url: https://dev-68495932.okta.com/oauth2/default/.well-known/openid-configuration
oauth2:
OAuth2TokenVerify: jwks_uri https://dev-68495932.okta.com/oauth2/default/v1/keys
token_endpoint: https://dev-68495932.okta.com/oauth2/default/v1/token
The oidc_auth_type parameter specifies the Apache module to use:
oauth20 or oauth2. The oauth20 functionality is deprecated
and superseded by a new oauth2 module. You can configure two and more
identity providers only with the oauth2 module.
Regions¶
A region in MOSK represents a complete OpenStack cluster that has a dedicated control plane and set of API endpoints. It is not uncommon for operators of large clouds to offer their users several OpenStack regions, which differ by their geographical location or purpose. In order to easily navigate in a multi-region environment, cloud users need a way to distinguish clusters by their names.
The region_name parameter of an OpenStackDeployment custom resource
specifies the name of the region that will be configured in all the OpenStack
services comprising the MOSK cluster upon the initial
deployment.
Important
Once the cluster is up and running, the cloud operator cannot set or change the name of the region. Therefore, Mirantis recommends selecting a meaningful name for the new region before the deployment starts. For example, the region name can be based on the name of the data center the cluster is located in.
Usage sample:
apiVersion: lcm.mirantis.com/v1alpha1
kind: OpenStackDeployment
metadata:
name: openstack-cluster
namespace: openstack
spec:
region_name: <your-region-name>
Application credentials¶
Application credentials is a mechanism in the MOSK Identity service that enables application automation tools, such as shell scripts, Terraform modules, Python programs, and others, to securely perform various actions in the cloud API in order to deploy and manage application components.
Application credentials is a modern alternative to the legacy approach where every application owner had to request several technical user accounts to ensure their tools could authenticate in the cloud.
For the details on how to create and authenticate with application credentials, refer to Manage application credentials.
By default, cloud users logging in to the cloud through the Mirantis Container Cloud IAM or any external identity provider cannot use the application credentials mechanism.
An application credential is heavily tied to the account of the cloud user owning it. An application automation tool that is a consumer of the credential acts on behalf of the human user who created the credential. Each action that the application automation tool performs gets authorized against the permissions, including roles and groups, the user currently has.
The source of truth about a federated user permissions is the identity provider. This information gets temporary transferred to the cloud’s Identity service inside a token once the user authenticates. By default, if such a user creates an application credential and passes it to the automation tool, there is no data to validate the tool’s action on the user’s behalf.
However, a cloud operator can configure the authorization_ttl parameter
for an identity provider object to enable caching of its users authorization
data. The parameter defines for how long in minutes the information about
user permissions is preserved in the database after the user successfully
logs in to the cloud.
Warning
Authorization data caching has security implications. In case a federated user account is revoked or his permissions change in the identity provider, the cloud Identity service will still allow performing actions on the user behalf until the cached data expires or the user re-authenticates in the cloud.
To set authorization_ttl to, for example, 60 minutes for the keycloak
identity provider in Keystone:
Log in to the
keystone-clientPod:kubectl -n openstack exec $(kubectl -n openstack get po -l application=keystone,component=client -oname) -ti -c keystone-client -- bash
Inside the Pod, run the following command:
openstack identity provider set keycloak --authorization-ttl 60
Domain-specific configuration¶
Parameter |
|
|---|---|
Usage |
Defines the domain-specific configuration and is useful for integration
with LDAP. An example of OsDpl with LDAP integration, which will create
a separate spec:
features:
keystone:
domain_specific_configuration:
enabled: true
domains:
domain.with.ldap:
enabled: true
config:
assignment:
driver: keystone.assignment.backends.sql.Assignment
identity:
driver: ldap
ldap:
chase_referrals: false
group_desc_attribute: description
group_id_attribute: cn
group_member_attribute: member
group_name_attribute: ou
group_objectclass: groupOfNames
page_size: 0
password: XXXXXXXXX
query_scope: sub
suffix: dc=mydomain,dc=com
url: ldap://ldap01.mydomain.com,ldap://ldap02.mydomain.com
user: uid=openstack,ou=people,o=mydomain,dc=com
user_enabled_attribute: enabled
user_enabled_default: false
user_enabled_invert: true
user_enabled_mask: 0
user_id_attribute: uid
user_mail_attribute: mail
user_name_attribute: uid
user_objectclass: inetOrgPerson
|
Image service¶
Mirantis OpenStack for Kubernetes (MOSK) provides the image management capability through the OpenStack Image service, aka Glance.
The Image service enables you to discover, register, and retrieve virtual machine images. Using the Glance API, you can query virtual machine image metadata and retrieve actual images.
MOSK deployment profiles include the Image service in the
core set of services. You can configure the Image service through the
spec:features definition in the OpenStackDeployment custom resource.
Image signature verification¶
TechPreview
MOSK can automatically verify the cryptographic signatures
associated with images to ensure the integrity of their data. A signed image
has a few additional properties set in its metadata that include
img_signature, img_signature_hash_method, img_signature_key_type,
and img_signature_certificate_uuid. You can find more information about
these properties and their values in the upstream OpenStack documentation.
MOSK performs image signature verification during the following operations:
A cloud user or a service creates an image in the store and starts to upload its data. If the signature metadata properties are set on the image, its content gets verified against the signature. The Image service accepts non-signed image uploads.
A cloud user spawns a new instance from an image. The Compute service ensures that the data it downloads from the image storage matches the image signature. If the signature is missing or does not match the data, the operation fails. Limitations apply, see Known limitations.
A cloud user boots an instance from a volume, or creates a new volume from an image. If the image is signed, the Block Storage service compares the downloaded image data against the signature. If there is a mismatch, the operation fails. The service will accept a non-signed image as a source for a volume. Limitations apply, see Known limitations.
spec:
features:
glance:
signature:
enabled: true
Every MOSK cloud is pre-provisioned with a baseline set of images containing most popular operating systems, such as Ubuntu, Fedora, CirrOS.
In addition, a few services in MOSK rely on the creation of service instances to provide their functions, namely the Load Balancer service and the Bare Metal service, and require corresponding images to exist in the image store.
When image signature verification is enabled during the cloud deployment, all these images get automatically signed with a pre-generated self-signed certificate. Enabling the feature in an already existing cloud requires manual signing of all of the images stored in it. Consult the OpenStack documentation for an example of the image signing procedure.
The image signature verification is supported for LVM and local backends for ephemeral storage.
The functionality is not compatible with Ceph-backed ephemeral storage combined with RAW formatted images. The Ceph copy-on-write mechanism enables the user to create instance virtual disks without downloading the image to a compute node, the data is handled completely on the side of a Ceph cluster. This enables you to spin up instances almost momentarily but makes it impossible to verify the image data before creating an instance from it.
The Image service does not enforce the presence of a signature in the metadata when the user creates a new image. The service will accept the non-signed image uploads.
The Image service does not verify the correctness of an image signature upon update of the image metadata.
MOSK does not validate if the certificate used to sign an image is trusted, it only ensures the correctness of the signature itself. Cloud users are allowed to use self-signed certificates.
The Compute service does not verify image signature for Ceph backend when the RAW image format is used as described in Supported storage backends.
The Compute service does not verify image signature if the image is already cached on the target compute node.
The Instance HA service may experience issues when auto-evacuating instances created from signed images if it does have access to the corresponding secrets in the Key manager service.
The Block Storage service does not perform image signature verification when a Ceph backend is used and the images are in the RAW format.
The Block Storage service does not enforce the presence of a signature on the images.
Object Storage service¶
Ceph Object Gateway provides Object Storage (Swift) API for end users in MOSK deployments. For the API compatibility, refer to Ceph Documentation: Ceph Object Gateway Swift API.
Object storage enablement¶
Parameter |
|
|---|---|
Usage |
Enables the object storage and provides a RADOS Gateway Swift API that is compatible with the OpenStack Swift API. To enable the service, add spec:
features:
services:
- object-storage
To create the RADOS Gateway pool in Ceph, see :ref: Operations Guide: Enable Ceph RGW Object Storage <enable-rgw>. |
Object storage server-side encryption¶
TechPreview
Ceph Object Gateway also provides Amazon S3 compatible API. For details, see Ceph Documentation: Ceph Object Gateway S3 API. Using integration with the OpenStack Key Manager service (Barbican), the objects uploaded through S3 API can be encrypted by Ceph Object Gateway according to the AWS Documentation: Protecting data using server-side encryption with customer-provided encryption keys (SSE-C) specification.
Instead of Swift, such configuration uses an S3 client to upload server-side encrypted objects. Using server-side encryption, the data is sent over a secure HTTPS connection in an unencrypted form and the Ceph Object Gateway stores that data in the Ceph cluster in an encrypted form.
Dashboard¶
MOSK Dashboard (OpenStack Horizon) provides a web-based interface for users to access the functions of the cloud services.
Custom theme¶
Parameter |
|
|---|---|
Usage |
Defines the list of custom OpenStack Dashboard themes. Content of the archive file with a theme depends on the level of customization and can include static files, Django templates, and other artifacts. For the details, refer to OpenStack official documentation: Customizing Horizon Themes. spec:
features:
horizon:
themes:
- name: theme_name
description: The brand new theme
url: https://<path to .tgz file with the contents of custom theme>
sha256summ: <SHA256 checksum of the archive above>
|
Message of the Day (MOTD)¶
Available since MOSK 25.1
MOSK enables a cloud operator to configure Message of the Day (MOTD) for the MOSK Dashboard (OpenStack Horizon). These short messages inform users about current infrastructure issues, upcoming maintenance, and other events, helping them plan their work with minimal service disruption.
Cloud operators can configure messages to appear before or after users log in to Horizon, or both. Messages can also be visually distinguished based on severity and support minimal HTML formatting, including links.
To define the MOTD, populate the following structure in the
OpenStackDeployment custom resource:
spec:
features:
horizon:
motd:
<NAME>:
level: <LEVEL>
message: <MESSAGE>
afterLogin: <true|false>
beforeLogin: <true|false>
Parameters:
<NAME>: A unique symbolic name to distinguish messageslevel: The severity level of the message. Supported values:success,info,warning, anderrorbeforeLogin: Boolean. Iftrue, the message appears on the login page for unauthorized users. Default:false.afterLogin: Boolean. Iftrue, the message appears after users log in. Default:true.
Configuration example:
spec:
features:
horizon:
motd:
errorBefore:
level: error
message: "We are experiencing <b>issues</b> with the authentication provider<br>Check the status at the <a href='https://foo.bar'>status page</a>"
afterLogin: false
beforeLogin: true
warnAfter:
level: warning
message: "Planned maintenance tomorrow"
The above configuration results in the following two messages displayed for all cloud users:
Login page:
After logging in:
Auxiliary cloud services:
Bare Metal service¶
The Bare Metal service (Ironic) is an extra OpenStack service that can be
deployed by the OpenStack Controller (Rockoon). This section provides the
baremetal-specific configuration options of the OpenStackDeployment
resource.
Enabling the Bare Metal service¶
The Bare Metal service is not included into the core set of services and needs
to be explicitly enabled in the OpenStackDeployment custom resource.
To install bare metal services, add the baremetal keyword to the
spec:features:services list:
spec:
features:
services:
- baremetal
Note
All bare metal services are scheduled to the nodes with the
openstack-control-plane: enabled label.
Ironic agent deployment images¶
To provision a user image onto a bare metal server, Ironic boots a node with
a ramdisk image. Depending on the node’s deploy interface and hardware, the
ramdisk may require different drivers (agents). MOSK
provides tinyIPA-based ramdisk images and uses the direct deploy interface
with the ipmitool power interface.
Example of agent_images configuration:
spec:
features:
ironic:
agent_images:
base_url: https://binary.mirantis.com/openstack/bin/ironic/tinyipa
initramfs: tinyipa-stable-ussuri-20200617101427.gz
kernel: tinyipa-stable-ussuri-20200617101427.vmlinuz
Since the bare metal nodes hardware may require additional drivers, you may need to build a deploy ramdisk for particular hardware. For more information, see Ironic Python Agent Builder. Be sure to create a ramdisk image with the version of Ironic Python Agent appropriate for your OpenStack release.
Bare metal networking¶
Ironic supports the flat and multitenancy networking modes.
The flat networking mode assumes that all bare metal nodes are
pre-connected to a single network that cannot be changed during the
virtual machine provisioning. This network with bridged interfaces
for Ironic should be spread across all nodes including compute nodes
to allow plug-in regular virtual machines to connect to Ironic network.
In its turn, the interface defined as provisioning_interface should
be spread across gateway nodes. The cloud operator can perform
all these underlying configuration through the L2 templates.
Example of the OsDpl resource illustrating the configuration for the flat
network mode:
spec:
features:
services:
- baremetal
neutron:
external_networks:
- bridge: ironic-pxe
interface: <baremetal-interface>
network_types:
- flat
physnet: ironic
vlan_ranges: null
ironic:
# The name of neutron network used for provisioning/cleaning.
baremetal_network_name: ironic-provisioning
networks:
# Neutron baremetal network definition.
baremetal:
physnet: ironic
name: ironic-provisioning
network_type: flat
external: true
shared: true
subnets:
- name: baremetal-subnet
range: 10.13.0.0/24
pool_start: 10.13.0.100
pool_end: 10.13.0.254
gateway: 10.13.0.11
# The name of interface where provision services like tftp and ironic-conductor
# are bound.
provisioning_interface: br-baremetal
The multitenancy network mode uses the neutron Ironic network
interface to share physical connection information with Neutron. This
information is handled by Neutron ML2 drivers when plugging a Neutron port
to a specific network. MOSK supports the
networking-generic-switch Neutron ML2 driver out of the box.
Example of the OsDpl resource illustrating the configuration for the
multitenancy network mode:
spec:
features:
services:
- baremetal
neutron:
tunnel_interface: ens3
external_networks:
- physnet: physnet1
interface: <physnet1-interface>
bridge: br-ex
network_types:
- flat
vlan_ranges: null
mtu: null
- physnet: ironic
interface: <physnet-ironic-interface>
bridge: ironic-pxe
network_types:
- vlan
vlan_ranges: 1000:1099
ironic:
# The name of interface where provision services like tftp and ironic-conductor
# are bound.
provisioning_interface: <baremetal-interface>
baremetal_network_name: ironic-provisioning
networks:
baremetal:
physnet: ironic
name: ironic-provisioning
network_type: vlan
segmentation_id: 1000
external: true
shared: false
subnets:
- name: baremetal-subnet
range: 10.13.0.0/24
pool_start: 10.13.0.100
pool_end: 10.13.0.254
gateway: 10.13.0.11
DNS service¶
Mirantis OpenStack for Kubernetes (MOSK) provides DNS records managing capability through the DNS service (OpenStack Designate).
LoadBalancer type for PowerDNS¶
The supported backend for Designate is PowerDNS. If required, you can specify whether to use an external IP address or UDP, TCP, or TCP + UDP kind of Kubernetes for the PowerDNS service.
To configure LoadBalancer for PowerDNS, use the spec:features:designate
definition in the OpenStackDeployment custom resource.
The list of supported options includes:
external_ip- Optional. An IP address for the LoadBalancer service. If not defined, LoadBalancer allocates the IP address.protocol- A protocol for the Designate backend in Kubernetes. Can only beudp,tcp, ortcp+udp.type- The type of the backend for Designate. Can only bepowerdns.
For example:
spec:
features:
designate:
backend:
external_ip: 10.172.1.101
protocol: udp
type: powerdns
DNS service known limitations¶
Due to an issue in the dnspython library, Asynchronous Transfer Full Range
(AXFR) requests do not work and cause inability to set up a secondary DNS zone.
The issue affects OpenStack Victoria and is fixed in the Yoga release.
Key Manager service¶
MOSK Key Manager service (OpenStack Barbican) provides secure storage, provisioning, and management of cloud application secret data, such as Symmetric Keys, Asymmetric Keys, Certificates, and raw binary data.
Configuring the Vault backend¶
Parameter |
|
|---|---|
Usage |
Specifies the object containing the Vault parameters to connect to Barbican. The list of supported options includes:
If the Vault backend is used, configure it properly using the following parameters: spec:
features:
barbican:
backends:
vault:
enabled: true
approle_role_id: <APPROLE_ROLE_ID>
approle_secret_id: <APPROLE_SECRET_ID>
vault_url: <VAULT_SERVER_URL>
use_ssl: false
Mirantis recommeds hiding the Note Since MOSK does not currently support the
Vault SSL encryption, set the |
Instance High Availability service¶
TechPreview
Instance High Availability service (OpenStack Masakari) enables cloud users to ensure that their instances get automatically evacuated from a failed hypervisor.
The service consists of the following components:
API recieves requests from users and events from monitors, and sends them to engine
Engine executes recovery workflow
Monitors detect failures and notifies API. MOSK uses monitors of the following types:
Instance monitor performs liveness of instance processes
Introspective instance monitor enhances instance high availability within OpenStack environments by monitoring and identifying system-level failures through the QEMU Guest Agent
Host monitor performs liveness of a compute host, runs as part of the Node controller from the OpenStack Controller (Rockoon)
Note
The Processes monitor is not present in MOSK as far as HA for the compute processes is handled by Kubernetes.
This section describes how to enable various components of the Instance High Availability service for your MOSK deployment:
Enabling the Instance HA service¶
The Instance HA service is not included into the core set of services and needs
to be explicitly enabled in the OpenStackDeployment custom resource.
Parameter |
|
|---|---|
Usage |
Enables Masakari, the OpenStack service that ensures high availability
of instances running on a host. To enable the service, add
spec:
features:
services:
- instance-ha
|
Enabling introspective instance monitor¶
Available since MOSK 25.1 TechPreview
The introspective instance monitor in the Instance High Availability service enhances the reliability of the cloud environment by monitoring virtual machines for failure events, including operating system crashes, kernel panics, and unresponsive states. Upon detecting such events in real time, the monitor initiates automated recovery actions, such as rebooting the affected instance. This allows for reduced downtime and maintains high availability of an OpenStack environment.
As a cloud operator, you can enable and configure the instance introspection
through the spec:features:masakari:monitors:introspective definition in the
OpenStackDeployment custom resource. The list of supported options include:
enabled(boolean)Enables or disables the introspection monitor. Default:
false.
guest_monitoring_interval(integer)Defines the time interval (in seconds) for monitoring the status of the guest virtual machine. Default:
10.
guest_monitoring_timeout(integer)Sets the timeout (in seconds) for detecting a non-responsive guest VM before marking it as failed. Default:
2.
guest_monitoring_failure_threshold(integer)Defines the number of consecutive failures required before a notification is sent or recovery action is initiated. Default:
3.
Example configuration:
spec:
features:
masakari:
monitors:
introspective:
enabled: true
guest_monitoring_interval: 10
guest_monitoring_timeout: 2
guest_monitoring_failure_threshold: 3
The introspective instance monitor relies on the QEMU Guest Agent being installed within the guest virtual machine. This agent enables communication between the host and guest operating systems, ensuring precise monitoring of the virtual machine health. Without the QEMU Guest Agent, the introspection monitor cannot accurately assess the state of the virtual machine, which may prevent the initiation of necessary recovery actions. To start monitoring, refer to Configure the introspective instance monitor.
Dynamic Resource Balancer service¶
Available since MOSK 24.2 TechPreview
In a cloud environment where resources are shared across all workloads, those resources often become a point of contention.
For example, it is not uncommon for an oversubscribed compute node to experience the noisy neighbor problem, when one of the instances may start consuming a lot more resources than usually, negatively affecting performance of other instances running on the same node.
In such cases, an intervention is required from the cloud operators to manually re-distribute workloads in the cluster to achieve more equal utilization of resources.
The Dynamic Resource Balancer (DRB) service continiously measures resource usage on hypervisors and redistributes workloads to achieve some optimum target, thereby eliminating the need for manual interventions from cloud operators.
Architecture overview¶
The DRB service is implemented as a Kubernetes operator, controlled by
the custom resource of kind: DRBConfig. Unless at least one resource
of such kind is present, the service does not perform any operations. Cloud
operators who want to enable the DRB service for their MOSK
clouds, need to create the resource with proper configuration.
The DRB controller consists of the following сomponents interacting with each other:
collectorCollects the statistics of resource consumption in the cluster
schedulerBased on the data from the collector, makes decisions whether cloud resources need to be relocated to achieve the optimum
actuatorExecutes the resource relocation decisions made by
scheduler
Out of the box, these service components implement a very simple logic, which,
however, can be individually enhanced according to the needs of a specific
cloud environment by utilizing their pluggable architecture. The plugins
need to be written in Python programming language and injected as modules
into the DRB service by building a custom drb-controller container image.
Default plugins as well as custom plugins are configured through the
corresponding sections of DRBConfig custom resources.
Also, it is possible to limit the scope of DRB decisions and actions to only a subset of hosts. This way, you can model the node grouping schema that is configured in OpenStack, for example, compute node aggregates and availability zones, to avoid DRB service attempting resource placement changes that cannot be fulfilled by MOSK Compute service (OpenStack Nova).
Example configuration¶
apiVersion: lcm.mirantis.com/v1alpha1
kind: DRBConfig
metadata:
name: drb-test
namespace: openstack
spec:
actuator:
max_parallel_migrations: 10
migration_polling_interval: 5
migration_timeout: 180
name: os-live-migration
collector:
name: stacklight
hosts: []
migrateAny: false
reconcileInterval: 300
scheduler:
load_threshold: 80
min_improvement: 0
name: vm-optimize
The spec section of configuration consists of the following main parts:
collectorSpecifies and configures the collector plugin to collect the metrics on which decisions are based. At a minimum, the
nameof the plugin must be provided.
schedulerSpecifies and configures the scheduler plugin that will make decisions based on the collected metrics. At a minimum, the
nameof the plugin must be provided.
actuatorSpecifies and configures the actuator plugin that will move resources around. At a minimum, the
nameof the plugin must be provided.
reconcileIntervalDefines time in seconds between reconciliation cycles. Should be large enough for the metrics to settle after resources are moved around.
For the default
stacklightcollector plugin, this value must equal at least300.
hostsSpecifies the list of cluster hosts to which this given instance of
DRBConfigapplies. This means that only metrics from these hosts will be used for making decisions, only resources belonging to these hosts will be considered for re-distribution, and only these hosts will be considered as possible targets for re-distribution.You can create multiple
DRBConfigresources that watch over non-overlapping sets of hosts.Default of this setting is an empty list that implies all hosts.
migrateAnyA boolean flag that the scheduler plugin can consider when making decisions, allowing cloud operators and users to opt certain workloads in or out of redistribution.
For the default
vm-optimizescheduler plugin:migrateAny: true(default) - any instance can be migrated, except for instances tagged withlcm.mirantis.com:no-drb, explicitly opting out of the DRB functionalitymigrateAny: false- only instances tagged withlcm.mirantis.com:drbare migrated by the DRB service, explicitly opting in to the DRB functionality
Included default plugins¶
Collects node_load5, machine_cpu_cores, and
libvirt_domain_info_cpu_time_seconds:rate5m metrics
from the StackLight service running in the MOSK cluster.
Does not have options available.
Requires the reconcileInterval set to at least 300 (5 minutes),
as both the collected node and instance CPU usage metrics are effectively
averaged over a 5-minute sliding window.
Attempts to minimize the standard deviation of node load. The node load is normalized per CPU core, so heterogeneous compute hosts can be compared.
Available options:
load_thresholdThe value in percent of the compute host load after which the host will be considered overloaded and attempts will be made to migrate instances from it. Defaults to
80.
min_improvementMinimal improvement of the optimization metric in percent. While making decisions, the scheduler attempts to predict the resulting load distribution to determine if moving resources is beneficial. If the total improvement after all necessary decisions is calculated to be less than
min_improvement, no decisions will be executed.Defaults to
0, any potential improvement is acted upon. Setting this to a higher value should allow avoiding instance migrations that provide negligible improvements.
Warning
The current version of this plugin takes into account only basic resource classes when making scheduling decisions. These include only RAM, disk, and vCPU count from the instance flavor. It does not take into account any other information including specific image or aggregate metadata, custom resource classes, PCI devices, NUMA, hugepages, and so on. Moving around instances that consume such resources will more likely fail as the current implementation of the scheduler plugin cannot reliably predict if such instances fit onto the selected target host.
Live migrates instances to specific hosts. Assumes any migration is possible.
Refer to the hosts and migrateAny options above to learn how to control
which instances are migrated to which locations.
Available options:
max_parallel_migrationsDefines the number of instances to migrate in parallel.
Defaults to
10.This value applies to all decisions being processed, so it may involve instances from different hosts. Meanwhile, the
nova-computeservice may have its own limits on how many live migrations a given host can handle in parallel.
migration_polling_intervalDefines the interval in seconds for checking the instance status while the latter is being migrated
Defaults to
5.
migration_timeoutDefines the interval in seconds after which an unfinished migration is considered failed.
Defaults to
180.
Only logs the decisions that were scheduled for execution. Useful for debugging and dry-runs.
Note
The list of the services and their supported features included in this section is not full and is being constantly amended based on the complexity of the architecture and use of a particular service.
OpenStack¶
OpenStack cluster¶
OpenStack and auxiliary services are running as containers in the kind: Pod
Kubernetes resources. All long-running services are governed by one of
the ReplicationController-enabled Kubernetes resources, which include
either kind: Deployment, kind: StatefulSet, or kind: DaemonSet.
The placement of the services is mostly governed by the Kubernetes node labels. The labels affecting the OpenStack services include:
openstack-control-plane=enabled- the node hosting most of the OpenStack control plane services.openstack-compute-node=enabled- the node serving as a hypervisor for Nova. The virtual machines with tenants workloads are created there.openvswitch=enabled- the node hosting Neutron L2 agents and OpenvSwitch pods that manage L2 connection of the OpenStack networks.openstack-gateway=enabled- the node hosting Neutron L3, Metadata and DHCP agents, Octavia Health Manager, Worker and Housekeeping components.
Note
OpenStack is an infrastructure management platform. Mirantis OpenStack for Kubernetes (MOSK) uses Kubernetes mostly for orchestration and dependency isolation. As a result, multiple OpenStack services are running as privileged containers with host PIDs and Host Networking enabled. You must ensure that at least the user with the credentials used by Helm/Tiller (administrator) is capable of creating such Pods.
Infrastructure services¶
Service |
Description |
|---|---|
Storage |
While the underlying Kubernetes cluster is configured to use Ceph CSI for providing persistent storage for container workloads, for some types of workloads such networked storage is suboptimal due to latency. This is why the separate |
Database |
A single WSREP (Galera) cluster of MariaDB is deployed as the SQL
database to be used by all OpenStack services. It uses the storage class
provided by Local Volume Provisioner to store the actual database files.
The service is deployed as |
Messaging |
RabbitMQ is used as a messaging bus between the components of the OpenStack services. A separate instance of RabbitMQ is deployed for each OpenStack service that needs a messaging bus for intercommunication between its components. An additional, separate RabbitMQ instance is deployed to serve as a notification messages bus for OpenStack services to post their own and listen to notifications from other services. StackLight also uses this message bus to collect notifications for monitoring purposes. Each RabbitMQ instance is a single node and is deployed as
|
Caching |
A single multi-instance of the Memcached service is deployed to be used by all OpenStack services that need caching, which are mostly HTTP API services. |
Coordination |
A separate instance of etcd is deployed to be used by Cinder, which require Distributed Lock Management for coordination between its components. |
Ingress |
Is deployed as |
Image pre-caching |
A special This is especially useful for containers used in |
OpenStack services¶
Service |
Description |
|---|---|
Identity (Keystone) |
Uses MySQL backend by default.
|
Image (Glance) |
Supported backend is RBD (Ceph is required). |
Volume (Cinder) |
Supported backend is RBD (Ceph is required). |
Network (Neutron) |
Supported backends are Open vSwitch, Open Virtual Network, and Tungsten Fabric. |
Placement |
|
Compute (Nova) |
Supported hypervisor is Qemu/KVM through libvirt library. |
Dashboard (Horizon) |
|
DNS (Designate) |
Supported backend is PowerDNS. |
Load Balancer (Octavia) |
|
Ceph Object Gateway (SWIFT) |
Provides the object storage and a Ceph Object Gateway Swift API that is
compatible with the OpenStack Swift API. You can manually enable the
service in the |
Instance HA (Masakari) |
An OpenStack service that ensures high availability of instances running
on a host. You can manually enable Masakari in the
|
Orchestration (Heat) |
|
Key Manager (Barbican) |
The supported backends include:
|
Tempest |
Runs tests against a deployed OpenStack cloud. You can manually enable
Tempest in the |
Shared Filesystems (OpenStack Manila) |
Provides Shared Filesystems as a service that enables you to create and manage shared filesystems in a multi-project cloud environments. For details, refer to Shared Filesystems service. |
Shared Filesystems (OpenStack Manila) |
Provides Shared Filesystems as a service that enables you to create and manage shared filesystems in a multi-project cloud environments. For details, refer to Shared Filesystems service. |
OpenStack database architecture¶
A complete setup of a MariaDB Galera cluster for OpenStack is illustrated in the following image:
MariaDB server pods are running a Galera multi-master cluster. Clients
requests are forwarded by the Kubernetes mariadb service to the
mariadb-server pod that has the primary label. Other pods from
the mariadb-server StatefulSet have the backup label. Labels are
managed by the mariadb-controller pod.
The MariaDB Controller periodically checks the readiness of the
mariadb-server pods and sets the primary label to it if the following
requirements are met:
The
primarylabel has not already been set on the pod.The pod is in the ready state.
The pod is not being terminated.
The pod name has the lowest integer suffix among other ready pods in the StatefulSet. For example, between
mariadb-server-1andmariadb-server-2, the pod with themariadb-server-1name is preferred.
Otherwise, the MariaDB Controller sets the backup label. This means that
all SQL requests are passed only to one node while other two nodes are in
the backup state and replicate the state from the primary node.
The MariaDB clients are connecting to the mariadb service.
OpenStack lifecycle management¶
The OpenStack Operator component is a combination of the following entities:
OpenStack Controller (Rockoon)¶
The OpenStack Controller (Rockoon) runs in a set of containers in a pod in Kubernetes. Rockoon is deployed as a Deployment with 1 replica only. The failover is provided by Kubernetes that automatically restarts the failed containers in a pod.
However, given the recommendation to use a separate Kubernetes cluster for each OpenStack deployment, the controller in envisioned mode for operation and deployment will only manage a single OpenStackDeployment resource, making the proper HA much less of an issue.
Rockoon is written in Python using Kopf, as a Python framework to build Kubernetes operators, and Pykube, as a Kubernetes API client.
Using Kubernetes API, the controller subscribes to changes to resources of
kind: OpenStackDeployment, and then reacts to these changes by creating,
updating, or deleting appropriate resources in Kubernetes.
The basic child resources managed by the controller are Helm releases.
They are rendered from templates taking into account
an appropriate values set from the main and features fields in the
OpenStackDeployment resource.
Then, the common fields are merged to resulting data structures. Lastly, the services fields are merged providing the final and precise override for any value in any Helm release to be deployed or upgraded.
The constructed values are then used by Rockoon during a Helm release
installation.
Container |
Description |
|---|---|
|
The core container that handles changes in the |
|
The container that watches the |
|
The container that watches all Kubernetes native
resources, such as |
|
The container that provides data exchange between different components such as Ceph. |
|
The container that handles the node events. |
OpenStackDeployment Admission Controller¶
The CustomResourceDefinition resource in Kubernetes uses the
OpenAPI Specification version 2 to specify the schema of the resource
defined. The Kubernetes API outright rejects the resources that do not
pass this schema validation.
The language of the schema, however, is not expressive enough to define a specific validation logic that may be needed for a given resource. For this purpose, Kubernetes enables the extension of its API with Dynamic Admission Control.
For the OpenStackDeployment (OsDpl) CR the ValidatingAdmissionWebhook
is a natural choice. It is deployed as part of OpenStack Controller (Rockoon)
by default and performs specific extended validations when an OsDpl CR is
created or updated.
The inexhaustive list of additional validations includes:
Deny the OpenStack version downgrade
Deny the OpenStack version skip-level upgrade
Deny the OpenStack master version deployment
Deny upgrade to the OpenStack master version
Deny upgrade if any part of an OsDpl CR specification changes along with the OpenStack version
Under specific circumstances, it may be viable to disable the Admission Controller, for example, when you attempt to deploy or upgrade to the master version of OpenStack.
Warning
Mirantis does not support MOSK deployments performed without the OpenStackDeployment Admission Controller enabled. Disabling of the OpenStackDeployment Admission Controller is only allowed in staging non-production environments.
To disable the Admission Controller, ensure that the following structures and
values are present in the rockoon HelmBundle resource:
apiVersion: lcm.mirantis.com/v1alpha1
kind: HelmBundle
metadata:
name: openstack-operator
namespace: osh-system
spec:
releases:
- name: openstack-operator
values:
admission:
enabled: false
At that point, all safeguards except for those expressed by the CR definition are disabled.
OpenStack Exporter¶
The OpenStack Exporter collects metrics from the OpenStack services and exposes them to Prometheus for integration with StackLight. The Exporter interacts with the REST APIs of various OpenStack services to gather data about the infrastructure state and performance for visualization, alerting, and analysis within the monitoring system.
To retrieve metrics from the OpenStack Exporter:
Locate the Exporter pod. The OpenStack Exporter runs in the
osh-systemnamespace:kubectl -n osh-system get pods | grep exporter
Query the metrics by executing the curl request inside the exporter container:
kubectl -n osh-system exec -t <EXPORTER-POD> -c exporter curl http://localhost:9102/
OpenStack configuration¶
MOSK provides the configurational capabilities through a number of custom resources. This section is intended to provide detailed overview of these custom resources and their possible configuration.
OpenStackDeployment custom resource¶
The detailed information about schema of an OpenStackDeployment
custom resource can be obtained by running:
kubectl get crd openstackdeployments.lcm.mirantis.com -o yaml
The definition of a particular OpenStack deployment can be obtained by running:
kubectl -n openstack get osdpl -o yaml
apiVersion: lcm.mirantis.com/v1alpha1
kind: OpenStackDeployment
metadata:
name: openstack-cluster
namespace: openstack
spec:
openstack_version: victoria
preset: compute
size: tiny
internal_domain_name: cluster.local
public_domain_name: it.just.works
features:
neutron:
tunnel_interface: ens3
external_networks:
- physnet: physnet1
interface: veth-phy
bridge: br-ex
network_types:
- flat
vlan_ranges: null
mtu: null
floating_network:
enabled: False
nova:
live_migration_interface: ens3
images:
backend: local
Available since MOSK 23.1
The OpenStackDeployment custom resource enables you to securely store
sensitive fields in Kubernetes secrets. To do that, verify that the
reference secret is present in the same namespace as the
OpenStackDeployment object and the
openstack.lcm.mirantis.com/osdpl_secret label is set to true.
The list of fields that can be hidden from OpenStackDeployment is limited
and defined by the OpenStackDeployment schema.
For example, to hide spec:features:ssl:public_endpoints:api_cert, use the
following structure:
spec:
features:
ssl:
public_endpoints:
api_cert:
value_from:
secret_key_ref:
key: api_cert
name: osh-dev-hidden
Element |
Sub-element |
Description |
|---|---|---|
|
n/a |
Specifies the version of the Kubernetes API that is used to create this object |
|
n/a |
Specifies the kind of the object |
|
|
Specifies the name of metadata. Should be set in compliance with the Kubernetes resource naming limitations |
|
Specifies the metadata namespace. While technically it is possible to
deploy OpenStack on top of Kubernetes in other than Warning Both OpenStack and Kubernetes platforms provide resources to applications. When OpenStack is running on top of Kubernetes, Kubernetes is completely unaware of OpenStack-native workloads, such as virtual machines, for example. For better results and stability, Mirantis recommends using a dedicated Kubernetes cluster for OpenStack, so that OpenStack and auxiliary services, Ceph, and StackLight are the only Kubernetes applications running in the cluster. |
|
|
|
Specifies the OpenStack release to deploy |
|
String that specifies the name of the
Every supported deployment profile incorporates an OpenStack preset. Refer to Deployment profiles for the list of possible values. |
|
|
String that specifies the size category for the OpenStack cluster. The size category defines the internal configuration of the cluster such as the number of replicas for service workers and timeouts, etc. The list of supported sizes include:
|
|
|
Specifies the public DNS name for OpenStack services. This is a base DNS name that must be accessible and resolvable by API clients of your OpenStack cloud. It will be present in the OpenStack endpoints as presented by the OpenStack Identity service catalog. The TLS certificates used by the OpenStack services (see below) must also be issued to this DNS name. |
|
|
Specifies the Kubernetes storage class name used for services to create
persistent volumes. For example, backups of MariaDB. If not specified,
the storage class marked as |
|
|
Contains the top-level collections of settings for the OpenStack deployment that potentially target several OpenStack services. The section where the customizations should take place. The |
region_name¶TechPreview
The name of the region used for deployment, defaults to RegionOne.
features:policies¶Defines the list of custom policies for OpenStack services.
Configuration structure:
spec:
features:
policies:
nova:
custom_policy: custom_value
The list of services available for configuration includes: Cinder, Nova, Designate, Keystone, Glance, Neutron, Heat, Octavia, Barbican, Placement, Ironic, aodh, Gnocchi, and Masakari.
Learn more about OpenStack API access policies in MOSK in OpenStack API access policies.
Caution
Mirantis is not responsible for cloud operability in case of default policies modifications but provides API to pass the required configuration to the core OpenStack services.
features:policies:strict_admin¶TechPreview
Enables a tested set of policies that limits the global admin role to
only the user with admin role in the admin project or user with the
service role. The latter should be used only for service users utilizied
for communication between OpenStack services.
Configuration structure:
spec:
features:
policies:
strict_admin:
enabled: true
services:
identity:
keystone:
values:
conf:
keystone:
resource:
admin_project_name: admin
admin_project_domain_name: Default
Note
The spec.services part of the above section will become
redundant in one of the following releases.
artifacts¶A low-level section that defines the base URI prefixes for images and binary artifacts.
common¶A low-level section that defines values that will be passed to all
OpenStack (spec:common:openstack) or auxiliary
(spec:common:infra) services Helm charts.
Configuration structure:
spec:
artifacts:
common:
openstack:
values:
infra:
values:
services¶A section of the lowest level, enables the definition of specific values to pass to specific Helm charts on a one-by-one basis:
Warning
Mirantis does not recommend changing the default settings for
spec:artifacts, spec:common, and spec:services elements.
Customizations can compromise the OpenStack deployment update and upgrade
processes.
However, you may need to edit the spec:services section to limit
hardware resources in case of a hyperconverged architecture as described in
Limit HW resources for hyperconverged OpenStack compute nodes.
Parameter |
|
|---|---|
Usage |
Specifies the standard logging levels for OpenStack services that
include the following, at increasing severity: Configuration example: spec:
features:
logging:
nova:
level: DEBUG
|
Depending on the use case, you may need to configure the same application components differently on different hosts. MOSK enables you to easily perform the required configuration through node-specific overrides at the OpenStack Controller side.
The limitation of using the node-specific overrides is that they override only the configuration settings while other components, such as startup scripts and others, should be reconfigured as well.
Caution
The overrides have been implemented in a similar way to the OpenStack node and node label specific DaemonSet configurations. Though, the OpenStack Controller node-specific settings conflict with the upstream OpenStack node and node label specific DaemonSet configurations. Therefore, we do not recommend configuring node and node label overrides.
The list of allowed node labels is located in the Cluster object status
providerStatus.releaseRef.current.allowedNodeLabels field.
If the value field is not defined in allowedNodeLabels, a label can
have any value.
Before or after a machine deployment, add the required label from the allowed
node labels list with the corresponding value to
spec.providerSpec.value.nodeLabels in machine.yaml. For example:
nodeLabels:
- key: <NODE-LABEL>
value: <NODE-LABEL-VALUE>
The addition of a node label that is not available in the list of allowed node labels is restricted.
The node-specific settings are activated through the spec:nodes
section of the OsDpl CR. The spec:nodes section contains the following
subsections:
features- implements overrides for a limited subset of fields and is constructed similarly tospec::featuresservices- similarly tospec::services, enables you to override settings in general for the components running as DaemonSets.
Example configuration:
spec:
nodes:
<NODE-LABEL>::<NODE-LABEL-VALUE>:
features:
# Detailed information about features might be found at
# openstack_controller/admission/validators/nodes/schema.yaml
services:
<service>:
<chart>:
<chart_daemonset_name>:
values:
# Any value from specific helm chart
Parameter |
|
|---|---|
Usage |
Enables tests against a deployed OpenStack cloud: spec:
features:
services:
- tempest
|
See also
OpenStackDeploymentStatus custom resource¶
The resource of kind OpenStackDeploymentStatus is a custom resource that
describes the status of an OpenStack deployment. To obtain detailed information
about the schema of an OpenStackDeploymentStatus custom resource:
kubectl get crd openstackdeploymentstatus.lcm.mirantis.com -o yaml
To obtain the status definition for a particular OpenStack deployment:
kubectl -n openstack get osdplst
Example of system response:
NAME OPENSTACK VERSION CONTROLLER VERSION STATE LCM PROGRESS HEALTH MOSK RELEASE
osh-dev antelope 0.16.1.dev104 APPLIED 20/20 21/22 MOSK 24.1.3
Where:
OPENSTACK VERSIONdisplays the actual OpenStack version of the deploymentCONTROLLER VERSIONindicates the version of the OpenStack Controller (Rockoon) responsible for the deploymentSTATEreflects the current status of life-cycle management. The list of possible values includes:APPLYINGindicates that some Kubernetes objects for applications are in the process of being appliedAPPLIEDindicates that all Kubernetes objects for applications have been applied to the latest state
LCM PROGRESSreflects the current progress ofSTATEin the format X/Y, where X denotes the number of applications with Kubernetes objects applied and in the actual state, and Y represents the total number of applications managed by the OpenStack Controller (Rockoon)HEALTHprovides an overview of the current health status of the OpenStack deployment in the format X/Y, where X represents the number of applications withnotReadypods, and Y is the total number of applications managed by the OpenStack Controller (Rockoon)MOSK RELEASEdisplays the current product release of the OpenStack deployment
NAME OPENSTACK VERSION CONTROLLER VERSION STATE MOSK RELEASE
osh-dev antelope 0.16.1.dev104 APPLIED MOSK 24.1
Where:
OPENSTACK VERSIONdisplays the actual OpenStack version of the deploymentCONTROLLER VERSIONindicates the version of the OpenStack Controller (Rockoon) responsible for the deploymentSTATEreflects the current status of life-cycle management. The list of possible values includes:APPLYINGindicates that some Kubernetes objects for applications are in the process of being appliedAPPLIEDindicates that all Kubernetes objects for applications have been applied to the latest state
MOSK RELEASEdisplays the current product release of the OpenStack deployment
Example of an OpenStackDeploymentStatus custom resource
configuration
1 kind: OpenStackDeploymentStatus
2 metadata:
3 name: osh-dev
4 namespace: openstack
5 spec: {}
6 status:
7 handle:
8 lastStatus: update
9 health:
10 barbican:
11 api:
12 generation: 2
13 status: Ready
14 cinder:
15 api:
16 generation: 2
17 status: Ready
18 backup:
19 generation: 1
20 status: Ready
21 scheduler:
22 generation: 1
23 status: Ready
24 volume:
25 generation: 1
26 status: Ready
27 osdpl:
28 cause: update
29 changes: '((''add'', (''status'',), None, {''watched'': {''ceph'': {''secret'':
30 {''hash'': ''0fc01c5e2593bc6569562b451b28e300517ec670809f72016ff29b8cbaf3e729''}}}}),)'
31 controller_version: 0.5.3.dev12
32 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
33 openstack_version: ussuri
34 state: APPLIED
35 timestamp: "2021-09-08 17:01:45.633143"
36 services:
37 baremetal:
38 controller_version: 0.5.3.dev12
39 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
40 openstack_version: ussuri
41 state: APPLIED
42 timestamp: "2021-09-08 17:00:54.081353"
43 block-storage:
44 controller_version: 0.5.3.dev12
45 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
46 openstack_version: ussuri
47 state: APPLIED
48 timestamp: "2021-09-08 17:00:57.306669"
49 compute:
50 controller_version: 0.5.3.dev12
51 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
52 openstack_version: ussuri
53 state: APPLIED
54 timestamp: "2021-09-08 17:01:18.853068"
55 coordination:
56 controller_version: 0.5.3.dev12
57 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
58 openstack_version: ussuri
59 state: APPLIED
60 timestamp: "2021-09-08 17:01:00.593719"
61 dashboard:
62 controller_version: 0.5.3.dev12
63 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
64 openstack_version: ussuri
65 state: APPLIED
66 timestamp: "2021-09-08 17:00:57.652145"
67 database:
68 controller_version: 0.5.3.dev12
69 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
70 openstack_version: ussuri
71 state: APPLIED
72 timestamp: "2021-09-08 17:01:00.233777"
73 dns:
74 controller_version: 0.5.3.dev12
75 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
76 openstack_version: ussuri
77 state: APPLIED
78 timestamp: "2021-09-08 17:00:56.540886"
79 identity:
80 controller_version: 0.5.3.dev12
81 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
82 openstack_version: ussuri
83 state: APPLIED
84 timestamp: "2021-09-08 17:01:00.961175"
85 image:
86 controller_version: 0.5.3.dev12
87 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
88 openstack_version: ussuri
89 state: APPLIED
90 timestamp: "2021-09-08 17:00:58.976976"
91 ingress:
92 controller_version: 0.5.3.dev12
93 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
94 openstack_version: ussuri
95 state: APPLIED
96 timestamp: "2021-09-08 17:01:01.440757"
97 key-manager:
98 controller_version: 0.5.3.dev12
99 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
100 openstack_version: ussuri
101 state: APPLIED
102 timestamp: "2021-09-08 17:00:51.822997"
103 load-balancer:
104 controller_version: 0.5.3.dev12
105 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
106 openstack_version: ussuri
107 state: APPLIED
108 timestamp: "2021-09-08 17:01:02.462824"
109 memcached:
110 controller_version: 0.5.3.dev12
111 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
112 openstack_version: ussuri
113 state: APPLIED
114 timestamp: "2021-09-08 17:01:03.165045"
115 messaging:
116 controller_version: 0.5.3.dev12
117 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
118 openstack_version: ussuri
119 state: APPLIED
120 timestamp: "2021-09-08 17:00:58.637506"
121 networking:
122 controller_version: 0.5.3.dev12
123 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
124 openstack_version: ussuri
125 state: APPLIED
126 timestamp: "2021-09-08 17:01:35.553483"
127 object-storage:
128 controller_version: 0.5.3.dev12
129 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
130 openstack_version: ussuri
131 state: APPLIED
132 timestamp: "2021-09-08 17:01:01.828834"
133 orchestration:
134 controller_version: 0.5.3.dev12
135 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
136 openstack_version: ussuri
137 state: APPLIED
138 timestamp: "2021-09-08 17:01:02.846671"
139 placement:
140 controller_version: 0.5.3.dev12
141 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
142 openstack_version: ussuri
143 state: APPLIED
144 timestamp: "2021-09-08 17:00:58.039210"
145 redis:
146 controller_version: 0.5.3.dev12
147 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec
148 openstack_version: ussuri
149 state: APPLIED
150 timestamp: "2021-09-08 17:00:36.562673"
The health subsection provides a brief output on services health.
The osdpl subsection describes the overall status of the OpenStack
deployment.
Element |
Description |
|---|---|
|
The cause that triggered the LCM action: |
|
A string representation of changes in the |
|
The version of |
|
The SHA sum of the |
|
The current OpenStack version specified in the |
|
The current state of the LCM action. Possible values include:
|
|
The timestamp of the |
The services subsection provides detailed information of LCM performed with
a specific service. This is a dictionary where keys are service names, for
example, baremetal or compute and values are dictionaries with the
following items.
Element |
Description |
|---|---|
|
The version of the |
|
The SHA sum of the |
|
The OpenStack version specified in the |
|
The current state of the LCM action performed on a service. Possible values include:
|
|
The timestamp of the |
OpenStack Controller configuration¶
Available since MOSK 23.2
OpenStack Controller (Rockoon)
Since MOSK 25.1, the OpenStack Controller has been open-sourced under the name Rockoon and is maintained as an independent open-source project going forward.
As part of this transition, all openstack-controller pods are named
rockoon pods across the MOSK documentation and deployments. This change
does not affect functionality, but this is the reminder for the users to
utilize the new naming for pods and other related artifacts accordingly.
The OpenStack Controller (Rockoon) enables you to modify its configuration at
runtime without restarting. MOSK stores the controller
configuration in the rockoon-config ConfigMap in the osh-system
namespace of your cluster.
To retrieve the Rockoon configuration ConfigMap, run:
kubectl get configmaps rockoon-config -o yaml
Example of a Rockoon configuration ConfigMap
apiVersion: v1
data:
extra_conf.ini: |
[maintenance]
respect_nova_az = false
kind: ConfigMap
metadata:
annotations:
openstackdeployments.lcm.mirantis.com/skip_update: "true"
name: rockoon-config
namespace: osh-system
Section |
Parameter |
Default value |
Description |
|---|---|---|---|
|
|
|
The number of seconds to wait for all application components to become ready. |
|
|
The number of seconds before going to the sleep mode between attempts to verify if the application is ready. |
|
|
|
The amount of time to wait for the flapping node. |
|
|
|
|
The number of seconds to wait until the values set in the manifest are propagated to the dependent objects. |
|
|
The number of seconds between attempts to verify if the values were applied. |
|
|
|
The number of seconds to wait until the values are removed from the manifest and propagated to the child objects. |
|
|
|
The number of seconds between attempts to verify if the values were removed from the release. |
|
|
|
The number of seconds to wait until the Kubernetes object is removed. |
|
|
|
The number of seconds between attempts to verify if the Kubernetes object is removed. |
|
|
|
The number of seconds to pause for the Helm bundle changes. |
|
|
|
|
The number of instances to migrate concurrently. |
|
|
The maximum number of compute nodes allowed for a parallel update. |
|
|
|
The maximum number of gateway nodes allowed for a parallel update. |
|
|
|
Respect Nova availability zone (AZ). The |
|
|
|
The flag to skip the instance verification on a host before proceeding
with the node removal. The |
|
|
|
The flag to skip the volume verification on a host before proceeding
with the node removal. The |
Custom OpenStack images¶
Available since MOSK 25.1
The OpenStack Controller enables you to use customized images in your OpenStack
deployments. To start using such images, create a ConfigMap in the
openstack namespace with the following content, replacing
<OPENSTACKDEPLOYMENT-NAME> with the name of your OpenStackDeployment
custom resource:
apiVersion: v1
kind: ConfigMap
metadata:
labels:
openstack.lcm.mirantis.com/watch: "true"
name: <OPENSTACKDEPLOYMENT-NAME>-artifacts
namespace: openstack
data:
caracal: |
dep_check: <KUBERNETES-ENTRYPOINT-IMAGE-URL>
openvswitch_db_server: <OPENVSWITCH-IMAGE-URL>
openvswitch_vswitchd: <OPENVSWITCH-IMAGE-URL>
OpenStack database¶
MOSK relies on the MariaDB Galera cluster to provide its OpenStack components with a reliable storage of persistent data.
For successful long-term operations of a MOSK cloud, it is crucial to ensure the healthy state of the OpenStack database as well as the safety of the data stored in it. To help you with that, MOSK provides built-in automated procedures for OpenStack database maintenance, backup, and restoration. The hereby chapter describes the internal mechanisms and configuration details for the provided tools.
Overview of the OpenStack database backup and restoration¶
MOSK relies on the MariaDB Galera cluster to provide its OpenStack components with a reliable storage for persistent data. Mirantis recommends backing up your OpenStack databases daily to ensure the safety of your cloud data. Also, you should always create an instant backup before updating your cloud or performing any kind of potentially disruptive experiment.
MOSK has a built-in automated backup routine that can be triggered manually or by schedule. For detailed information about the process of MariaDB Galera cluster backup, refer to Workflows of the OpenStack database backup and restoration.
Backup and restoration can only be performed against the OpenStack database as a whole. Granular per-service or per-table procedures are not supported by MOSK.
Important
Because database restoration reverts to a specific snapshot, the resulting database state may not accurately reflect the current state of dynamic resources such as running VMs, volumes, and similar components. For example:
A VM is removed after a snapshot for restoration is created. Such VM will be present as an orphan entry in the database and the OpenStack API after restoration.
A VM is created after a snapshot for restoration is created. Such VM will disappear from the OpenStack API after database restoration but will still be present as a process on the compute host.
Manually analyze and resolve such inconsistencies.
Restoring a database may also impact applications that rely on it for heartbeats. For instance, Octavia Amphorae may become unresponsive after restoration, potentially requiring LoadBalancer failover to maintain service availability.
By default, periodic backups are turned off. Though, a cloud operator can
easily enable this capability by adding the following structure to the
OpenStackDeployment custom resource:
spec:
features:
database:
backup:
enabled: true
For the configuration details, refer to Periodic OpenStack database backups.
Along with the automated backup routine, MOSK provides the Mariabackup tool for the OpenStack database restoration. For the database restoration procedure, refer to Restore OpenStack databases from a backup. For more information about the restoration process, consult Workflows of the OpenStack database backup and restoration.
By default, MOSK backup routine stores the OpenStack database data into the Mirantis Ceph cluster, which is a part of the same cloud. This is sufficient for the vast majority of clouds. However, you may want to have the backup data stored off the cloud to comply with specific enterprise practices for infrastructure recovery and data safety.
To achieve that, MOSK enables you to point the backup routine to an external data volume. For details, refer to Remote storage for OpenStack database backups.
The size of a backup storage volume depends directly on the size of the
MOSK cluster, which can be determined through the
size parameter in the OpenStackDeployment CR.
The list of the recommended sizes for a minimal backup volume includes:
20 GB for the
tinycluster size40 GB for the
smallcluster size80 GB for the
mediumcluster size
If required, you can change the default size of a database backup volume. However, make sure that you configure the volume size before OpenStack deployment is complete. This is because there is no automatic way to resize the backup volume once the cloud is deployed. Also, only the local backup storage (Ceph) supports the configuration of the volume size.
To change the default size of the backup volume, use the following structure
in the OpenStackDeployment CR:
spec:
services:
database:
mariadb:
values:
volume:
phy_backup:
size: "200Gi"
To store the backup data to a local Mirantis Ceph, the MOSK underlying Kubernetes cluster needs to have a preconfigured storage class for Kubernetes persistent volumes with the Ceph cluster as a storage backend.
When restoring the OpenStack database from a local Ceph storage, the cron job restores the state on each MariaDB node sequentially. It is not possible to perform parallel restoration because Ceph Kubernetes volumes do not support concurrent mounting from multiple places.
Additionally, MOSK allows for increased backup safety through synchronization of local MariaDB backups with a remote S3 storage. For details, see Synchronization of local MariaDB backups with a remote S3 storage.
MOSK provides you with a capability to store the OpenStack database data outside of the cloud, on an external storage device that supports common data access protocols, such as third-party NAS appliances.
Refer to Remote storage for OpenStack database backups for the configuration details.
Available since MOSK 25.1 TechPreview
Security compliance may require storing backups of databases in an encrypted
format. MOSK enables encryption of database backups, both
local and remote, using the OpenSSL aes-256-cbc encryption.
To encrypt database backups, add the following configuration to the
OpenStackDeployment custom resource:
spec:
features:
database:
backup:
encryption:
enabled: true
Workflows of the OpenStack database backup and restoration¶
This section provides technical details about the internal implementation of automated backup and restoration routines built into MOSK. The below information would be helpful for troubleshooting of any issues related to the process or understanding the impact these procedures impose on a running cloud.
The OpenStack database backup workflow consists of the following phases.
The mariadb-phy-backup job is responsible for:
Performing basic sanity checks and choosing right node for backup
Verifying the wsrep status and changing the
wsrep_desyncparameter settingsChecking backup integrity (ensuring correct hash sums)
Managing the
mariadb-phy-backup-runnerpodIf enabled, synchronizing the local backup storage with the remote S3 storage
During the first backup phase, the following actions take place:
The
mariadb-phy-backuppod starts on the node where themariadb-serverreplica with the highest number in its name runs. For example, if the MariaDB server pods are namedmariadb-server-0,mariadb-server-1, andmariadb-server-2, themariadb-phy-backuppod starts on the same node asmariadb-server-2.The backup process verifies the hash sums of existing backup files based on ConfigMap information:
If the verification fails and synchronization with the remote S3 storage is enabled, the process checks the hash sums of remote backups as well. If the remote backups are valid, they are downloaded.
If the hash sums are incorrect for both local and remote backups, the backup job fails.
If no ConfigMap exists, these hash sum checks are skipped.
Sanity check: verification of the Kubernetes status and wsrep status of each MariaDB pod. If some pods have wrong statuses, the backup job fails unless the
--allow-unsafe-backupparameter is passed to the main script in the Kubernetes backup job.Note
Since MOSK 22.4, the
--allow-unsafe-backupfunctionality is removed from the product for security and backup procedure simplification purposes.Mirantis does not recommend setting the
--allow-unsafe-backupparameter unless it is absolutely required. To ensure the consistency of a backup, verify that the MariaDB Galera cluster is in a working state before you proceed with the backup.Desynchronize the replica from the Galera cluster. The script connects the target replica and sets the
wsrep_desyncvariable toON. Then, the replica stops receiving write-sets and receives the wsrep statusDonor/Desynced. The Kubernetes health check of thatmariadb-serverpod fails and the Kubernetes status of that pod becomesNot ready. If the pod has theprimarylabel, the MariaDB Controller sets thebackuplabel to it and the pod is removed from the endpoints list of the MariaDB service.Verify that there is enough space in the
/var/backupfolder to perform the backup. The amount of available space in the folder should exceed<DB-SIZE> * <MARIADB-BACKUP-REQUIRED-SPACE-RATIO>in KB.
The
mariadb-phy-backuppod performs the backup using the mariabackup tool.The script puts the backed up replica back to sync with the Galera cluster by setting
wsrep_desynctoOFFand waits for the replica to becomeReadyin Kubernetes.
The script calculates hash sums for backup files and stores them in a special ConfigMap.
If the number of existing backups exceeds the value of the
MARIADB_BACKUPS_TO_KEEPjob parameter, the script removes the oldest backups to maintain the allowed limit.If enabled, the script synchronizes the local backup storage with the remote S3 storage.
The OpenStack database restoration workflow consists of the following phases.
The mariadb-phy-restore job launches the mariadb-phy-restore pod.
This pod starts with the mariadb-server PVC with the highest number
in its name. This PVC is mounted to the /var/lib/mysql folder and the
backup PVC (or local filesystem if the hostapath backend is configured)
is mounted to /var/backup.
The mariadb-phy-restore pod contains the main restore script, which is
responsible for:
Scaling the
mariadb-serverStatefulSetVerifying the statuses of
mariadb-serverpodsManaging the
openstack-mariadb-phy-restore-runnerpodsChecking backup integrity (ensuring correct hash sums)
Caution
During the restoration, the database is not available for OpenStack services that means a complete outage of all OpenStack services.
During the first phase, the following actions take place:
The restoration process verifies the hash sums of existing backup files based on ConfigMap information:
If the verification fails and synchronization with the remote S3 storage is enabled, the process checks the hash sums of remote backups as well. If the remote backups are valid, they are downloaded.
If the hash sums are incorrect for both local and remote backups, the backup job fails.
Save the list of
mariadb-serverpersistent volume claims (PVC).Scale the
mariadbserver StatefulSet to0replicas. At this point, the database becomes unavailable for OpenStack services.
The
mariadb-phy-restorepod performs the following actions:Launches the
openstack-mariadb-phy-restore-runnerpod for eachmariadb-serverPVC. This pod cleans all MySQL data on each PVC.Collects logs from the
openstack-mariadb-phy-restore-runnerpod and then removes it.Unarchives the database backup files to a temporary directory within
/var/backup.Executes
mariabackup --prepareon the unarchived data.Restores the backup to
/var/lib/mysql.
The
mariadb-phy-restorepod scales themariadb-serverStatefulSet back to the configured number of replicas.The
mariadb-phy-restorepod waits until allmariadb-serverreplicas are ready.
OpenStack database auto-cleanup¶
By design, when deleting a cloud resource, for example, an instance, volume, or router, an OpenStack service does not immediately delete its data but marks it as removed so that it can later be picked up by the garbage collector.
Given that an OpenStack resource is often represented by more than one record in the database, deletion of all of them right away could affect the overall responsiveness of the cloud API. On the other hand, an OpenStack database being severely clogged with stale data is one of the most typical reasons for the cloud slowness.
To keep the OpenStack database small and performance fast, MOSK is pre-configured to automatically clean up the removed database records older than 30 days. By default, the clean up is performed for the following MOSK services every Monday according to the schedule:
Service |
Service identifier |
Clean up time |
|---|---|---|
Block Storage (OpenStack Cinder) |
|
12:01 a.m. |
Compute (OpenStack Nova) |
|
01:01 a.m. |
Image (OpenStack Glance) |
|
02:01 a.m. |
Instance HA (OpenStack Masakari) |
|
03:01 a.m. |
Key Manager (OpenStack Barbican) |
|
04:01 a.m. |
Orchestration (OpenStack Heat) |
|
05:01 a.m. |
If required, you can adjust the cleanup schedule for the OpenStack database by
adding the features:database:cleanup setting to the OpenStackDeployment
CR following the example below. The schedule parameter must contain a
valid cron expression. The age parameter specifies the number of days after
which a stale record gets cleaned up.
spec:
features:
database:
cleanup:
<os-service-identifier>:
enabled: true
schedule: "1 0 * * 1"
age: 30
batch: 1000
Periodic OpenStack database backups¶
MOSK uses the Mariabackup utility to back up the MariaDB Galera cluster data where the OpenStack data is stored. The Mariabackup gets launched on a periodic basis as a part of the Kubernetes CronJob included in any MOSK deployment and is suspended by default.
Note
If you are using the default backend to store the backup data, which is Ceph, you can increase the default size of a backup volume. However, make sure to configure the volume size before you deploy OpenStack.
For the default sizes and configuration details, refer to Size of a backup storage.
MOSK enables you to configure the periodic backup of the
OpenStack database through the OpenStackDeployment object. To enable the
backup, use the following structure:
spec:
features:
database:
backup:
enabled: true
TechPreview
To enhance cloud security, you can enable encryption of OpenStack
database backups using the OpenSSL aes-256-cbc encryption
through the OpenStackDeployment custom resource. Refer to
Backup encryption for configuration details.
By default, the backup job:
Runs backup on a daily basis at 01:00 AM
Creates incremental backups daily and full backups weekly
Keeps 10 latest full backups
Stores backups in the
mariadb-phy-backup-dataPVCHas the backup timeout of
3600secondsHas the
incrementalbackup type
To verify the configuration of the mariadb-phy-backup CronJob
object, run:
kubectl -n openstack get cronjob mariadb-phy-backup
Example of a mariadb-phy-backup CronJob object
apiVersion: batch/v1beta1
kind: CronJob
metadata:
annotations:
openstackhelm.openstack.org/release_uuid: ""
creationTimestamp: "2020-09-08T14:13:48Z"
managedFields:
<<<skipped>>>>
name: mariadb-phy-backup
namespace: openstack
resourceVersion: "726449"
selfLink: /apis/batch/v1beta1/namespaces/openstack/cronjobs/mariadb-phy-backup
uid: 88c9be21-a160-4de1-afcf-0853697dd1a1
spec:
concurrencyPolicy: Forbid
failedJobsHistoryLimit: 1
jobTemplate:
metadata:
creationTimestamp: null
labels:
application: mariadb-phy-backup
component: backup
release_group: openstack-mariadb
spec:
activeDeadlineSeconds: 4200
backoffLimit: 0
completions: 1
parallelism: 1
template:
metadata:
creationTimestamp: null
labels:
application: mariadb-phy-backup
component: backup
release_group: openstack-mariadb
spec:
containers:
- command:
- /tmp/mariadb_resque.py
- backup
- --backup-timeout
- "3600"
- --backup-type
- incremental
env:
- name: MARIADB_BACKUPS_TO_KEEP
value: "10"
- name: MARIADB_BACKUP_PVC_NAME
value: mariadb-phy-backup-data
- name: MARIADB_FULL_BACKUP_CYCLE
value: "604800"
- name: MARIADB_REPLICAS
value: "3"
- name: MARIADB_BACKUP_REQUIRED_SPACE_RATIO
value: "1.2"
- name: MARIADB_RESQUE_RUNNER_IMAGE
value: docker-dev-kaas-local.docker.mirantis.net/general/mariadb:10.4.14-bionic-20200812025059
- name: MARIADB_RESQUE_RUNNER_SERVICE_ACCOUNT
value: mariadb-phy-backup-runner
- name: MARIADB_RESQUE_RUNNER_POD_NAME_PREFIX
value: openstack-mariadb
- name: MARIADB_POD_NAMESPACE
valueFrom:
fieldRef:
apiVersion: v1
fieldPath: metadata.namespace
image: docker-dev-kaas-local.docker.mirantis.net/general/mariadb:10.4.14-bionic-20200812025059
imagePullPolicy: IfNotPresent
name: phy-backup
resources: {}
securityContext:
allowPrivilegeEscalation: false
readOnlyRootFilesystem: true
terminationMessagePath: /dev/termination-log
terminationMessagePolicy: File
volumeMounts:
- mountPath: /tmp
name: pod-tmp
- mountPath: /tmp/mariadb_resque.py
name: mariadb-bin
readOnly: true
subPath: mariadb_resque.py
- mountPath: /tmp/resque_runner.yaml.j2
name: mariadb-bin
readOnly: true
subPath: resque_runner.yaml.j2
- mountPath: /etc/mysql/admin_user.cnf
name: mariadb-secrets
readOnly: true
subPath: admin_user.cnf
dnsPolicy: ClusterFirst
initContainers:
- command:
- kubernetes-entrypoint
env:
- name: POD_NAME
valueFrom:
fieldRef:
apiVersion: v1
fieldPath: metadata.name
- name: NAMESPACE
valueFrom:
fieldRef:
apiVersion: v1
fieldPath: metadata.namespace
- name: INTERFACE_NAME
value: eth0
- name: PATH
value: /usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/
- name: DEPENDENCY_SERVICE
- name: DEPENDENCY_DAEMONSET
- name: DEPENDENCY_CONTAINER
- name: DEPENDENCY_POD_JSON
- name: DEPENDENCY_CUSTOM_RESOURCE
image: docker-dev-kaas-local.docker.mirantis.net/openstack/extra/kubernetes-entrypoint:v1.0.0-20200311160233
imagePullPolicy: IfNotPresent
name: init
resources: {}
securityContext:
allowPrivilegeEscalation: false
readOnlyRootFilesystem: true
runAsUser: 65534
terminationMessagePath: /dev/termination-log
terminationMessagePolicy: File
nodeSelector:
openstack-control-plane: enabled
restartPolicy: Never
schedulerName: default-scheduler
securityContext:
runAsUser: 999
serviceAccount: mariadb-phy-backup
serviceAccountName: mariadb-phy-backup
terminationGracePeriodSeconds: 30
volumes:
- emptyDir: {}
name: pod-tmp
- name: mariadb-secrets
secret:
defaultMode: 292
secretName: mariadb-secrets
- configMap:
defaultMode: 365
name: mariadb-bin
name: mariadb-bin
schedule: 0 1 * * *
successfulJobsHistoryLimit: 3
suspend: false
To override the default configuration, set the parameters and environment variables that are passed to the CronJob as described in the tables below.
Parameter |
Type |
Default |
Description |
|---|---|---|---|
|
String |
|
Type of a backup. The list of possible values include:
Usage example: spec:
features:
database:
backup:
backup_type: incremental
|
|
Integer |
|
Timeout in seconds for the system to wait for the backup operation to succeed. Usage example: spec:
services:
database:
mariadb:
values:
conf:
phy_backup:
backup_timeout: 30000
|
|
Boolean |
|
Not recommended, removed since MOSK 22.4. If set to
Usage example: spec:
services:
database:
mariadb:
values:
conf:
phy_backup:
allow_unsafe_backup: true
|
Variable |
Type |
Default |
Description |
|---|---|---|---|
|
Integer |
|
Number of full backups to keep. Usage example: spec:
features:
database:
backup:
backups_to_keep: 3
|
|
String |
|
Persistent volume claim used to store backups. Usage example: spec:
services:
database:
mariadb:
values:
conf:
phy_backup:
backup_pvc_name: mariadb-phy-backup-data
|
|
Integer |
|
Number of seconds that defines a period between 2 full backups.
During this period, incremental backups are performed. The parameter
is taken into account only if Usage example: spec:
features:
database:
backup:
full_backup_cycle: 70000
|
|
Floating |
|
Multiplier for the database size to predict the space required to create a backup, either full or incremental, and perform a restoration keeping the uncompressed backup files on the same file system as the compressed ones. To estimate the size of The Usage example: spec:
services:
database:
mariadb:
values:
conf:
phy_backup:
backup_required_space_ratio: 1.4
|
For example, to perform full backups monthly and incremental backups
daily at 02:30 AM and keep the backups for the last six months,
configure the database backup in your OpenStackDeployment object
as follows:
spec:
features:
database:
backup:
enabled: true
backups_to_keep: 6
schedule_time: '30 2 * * *'
full_backup_cycle: 2628000
Remote storage for OpenStack database backups¶
By default, MOSK stores the OpenStack database backups locally in the Mirantis Ceph cluster, which is a part of the same cloud.
Alternatively, MOSK provides you with a capability to create remote backups using an external storage. This section contains configuration details for a remote backend to be used for the OpenStack data backup.
In general, the built-in automated backup routine saves the data to the
mariadb-phy-backup-data PersistentVolumeClaim (PVC), which is provisioned
from StorageClass specified in the spec.persistent_volume_storage_class
parameter of the OpenstackDeployment custom resource (CR).
TechPreview
A preconfigured NFS server with NFS share that a Unix backup and restore user has access to. By default, it is the same user that runs MySQL server in a MariaDB image.
To get the Unix user ID, run:
kubectl -n openstack get cronjob mariadb-phy-backup -o jsonpath='{.spec.jobTemplate.spec.template.spec.securityContext.runAsUser}'
Note
Verify that the NFS server is accessible through the network from all of the OpenStack control plane nodes of the cluster.
The
nfs-commonpackage installed on all OpenStack control plane nodes.
Only NFS Unix authentication is supported.
Removal of the NFS persistent volume does not automatically remove the data.
No validation of mount options. If mount options are specified incorrectly in the
OpenStackDeploymentCR, the mount command fails upon the creation of a backup runner pod.
To enable the NFS backend, configure the following structure in the
OpenStackDeployment object:
spec:
features:
database:
backup:
enabled: true
backend: pv_nfs
pv_nfs:
server: <ip-address/dns-name-of-the-server>
path: <path-to-the-share-folder-on-the-server>
TechPreview
To enhance cloud security, you can enable encryption of OpenStack
database backups using the OpenSSL aes-256-cbc encryption
through the OpenStackDeployment custom resource. Refer to
Backup encryption for configuration details.
Optionally, MOSK enables you to set the required mount options for the NFS mount command. You can set as many options of mount as you need. For example:
spec:
services:
database:
mariadb:
values:
volume:
phy_backup:
nfs:
mountOptions:
- "nfsvers=4"
- "hard"
Synchronization of local MariaDB backups with a remote S3 storage¶
Available since MOSK 25.1 TechPreview
MOSK provides the capability to synchronize local MariaDB backups with a remote S3 storage. Distributing backups across multiple locations increases their safety. Optionally, backup archives stored in S3 can be encrypted on the server side.
To enable synchronization, you need to have a preconfigured S3 storage and a user account for access.
Only one remote S3 storage can be configured
Disabling the S3 synchronization does not automatically remove the data
Verify that the S3 storage is accessible through the network from all OpenStack control plane nodes.
Create the secret to store credentials for access to the S3 storage:
--- apiVersion: v1 kind: Secret metadata: labels: openstack.lcm.mirantis.com/osdpl_secret: "true" name: mariadb-backup-s3-hidden namespace: openstack type: Opaque data: access_key: <ACCESS-KEY-FOR-S3-ACCOUNT> secret_key: <SECRET-KEY-FOR-S3-ACCOUNT> sse_kms_key_id: <SECRET-KEY-FOR-SERVER-SIDE-ENCRYPTION>
Enable synchronization by adding the following structure to the
OpenStackDeploymentcustom resource. For example, to use Ceph RadosGW as the S3 storage provider and enable server-side encryption for stored archives:spec: features: database: backup: enabled: true sync_remote: enabled: true remotes: << remote name >>: conf: type: s3 provider: Ceph endpoint: <URL-TO-S3-STORAGE> path: <BUCKET-NAME-FOR-BACKUPS-ON-S3-STORAGE> server_side_encryption: aws:kms access_key_id: value_from: secret_key_ref: key: access_key name: mariadb-backup-s3-hidden secret_access_key: value_from: secret_key_ref: key: secret_key name: mariadb-backup-s3-hidden sse_kms_key_id: value_from: secret_key_ref: key: sse_kms_key_id name: mariadb-backup-s3-hidden
Alternatively, you can set the
providerparameter toAWSif you prefer using AWS as a provider for S3 storage and omit theserver_side_encryptionandsse_kms_key_idparameters if encryption is not required.
OpenStack message bus¶
The internal components of Mirantis OpenStack for Kubernetes (MOSK) coordinate their operations and exchange status information using the cluster’s message bus (RabbitMQ).
Exposable OpenStack notifications¶
Available since MOSK 22.5
MOSK enables you to configure OpenStack services to emit
notification messages to the MOSK cluster messaging bus
(RabbitMQ) every time an OpenStack resource, for example, an instance, image,
and so on, changes its state due to a cloud user action or through its
lifecycle. For example, MOSK Compute service (OpenStack
Nova) can publish the instance.create.end notification once a newly created
instance is up and running.
Note
In certain cases, RabbitMQ notifications may prove unreliable, such as when the RabbitMQ server undergoes a restart or when communication between the server and the client reading the notifications breaks down. To optimize reliability, Mirantis suggests using multiple channels to store notification events, encompassing:
StackLight notifications
Storing audit as part of the OpenStack logs
Sample of an instance.create.end notification
{
"event_type": "instance.create.end",
"payload": {
"nova_object.data": {
"action_initiator_project": "6f70656e737461636b20342065766572",
"action_initiator_user": "fake",
"architecture": "x86_64",
"auto_disk_config": "MANUAL",
"availability_zone": "nova",
"block_devices": [],
"created_at": "2012-10-29T13:42:11Z",
"deleted_at": null,
"display_description": "some-server",
"display_name": "some-server",
"fault": null,
"flavor": {
"nova_object.data": {
"description": null,
"disabled": false,
"ephemeral_gb": 0,
"extra_specs": {
"hw:watchdog_action": "disabled"
},
"flavorid": "a22d5517-147c-4147-a0d1-e698df5cd4e3",
"is_public": true,
"memory_mb": 512,
"name": "test_flavor",
"projects": null,
"root_gb": 1,
"rxtx_factor": 1.0,
"swap": 0,
"vcpu_weight": 0,
"vcpus": 1
},
"nova_object.name": "FlavorPayload",
"nova_object.namespace": "nova",
"nova_object.version": "1.4"
},
"host": "compute",
"host_name": "some-server",
"image_uuid": "155d900f-4e14-4e4c-a73d-069cbf4541e6",
"instance_name": "instance-00000001",
"ip_addresses": [
{
"nova_object.data": {
"address": "192.168.1.3",
"device_name": "tapce531f90-19",
"label": "private",
"mac": "fa:16:3e:4c:2c:30",
"meta": {},
"port_uuid": "ce531f90-199f-48c0-816c-13e38010b442",
"version": 4
},
"nova_object.name": "IpPayload",
"nova_object.namespace": "nova",
"nova_object.version": "1.0"
}
],
"kernel_id": "",
"key_name": "my-key",
"keypairs": [
{
"nova_object.data": {
"fingerprint": "1e:2c:9b:56:79:4b:45:77:f9:ca:7a:98:2c:b0:d5:3c",
"name": "my-key",
"public_key": "ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAAAgQDx8nkQv/zgGgB4rMYmIf+6A4l6Rr+o/6lHBQdW5aYd44bd8JttDCE/F/pNRr0lRE+PiqSPO8nDPHw0010JeMH9gYgnnFlyY3/OcJ02RhIPyyxYpv9FhY+2YiUkpwFOcLImyrxEsYXpD/0d3ac30bNH6Sw9JD9UZHYcpSxsIbECHw== Generated-by-Nova",
"type": "ssh",
"user_id": "fake"
},
"nova_object.name": "KeypairPayload",
"nova_object.namespace": "nova",
"nova_object.version": "1.0"
}
],
"launched_at": "2012-10-29T13:42:11Z",
"locked": false,
"locked_reason": null,
"metadata": {},
"node": "fake-mini",
"os_type": null,
"power_state": "running",
"progress": 0,
"ramdisk_id": "",
"request_id": "req-5b6c791d-5709-4f36-8fbe-c3e02869e35d",
"reservation_id": "r-npxv0e40",
"state": "active",
"tags": [
"tag"
],
"task_state": null,
"tenant_id": "6f70656e737461636b20342065766572",
"terminated_at": null,
"trusted_image_certificates": [
"cert-id-1",
"cert-id-2"
],
"updated_at": "2012-10-29T13:42:11Z",
"user_id": "fake",
"uuid": "178b0921-8f85-4257-88b6-2e743b5a975c"
},
"nova_object.name": "InstanceCreatePayload",
"nova_object.namespace": "nova",
"nova_object.version": "1.12"
},
"priority": "INFO",
"publisher_id": "nova-compute:compute"
}
OpenStack notification messages can be consumed and processed by various corporate systems to integrate MOSK clouds into the company infrastructure and business processes.
The list of the most common use cases includes:
Using notification history for retrospective security audit
Using the real-time aggregation of notification messages to gather statistics on cloud resource consumption for further capacity planning
Cloud billing considerations
Notifications alone should not be considered as a source of data for any kind of financial reporting. The delivery of the messages can not be guaranteed due to various technical reasons. For example, messages can be lost if an external consumer is not fetching them from the queue fast enough.
Mirantis strongly recommends that your cloud billing solutions rely on the combination of the following data sources:
Periodic polling of the OpenStack API as a reliable source of information about allocated resources
Subscription to notifications to receive timely updates about the resource status change
If you are looking for a ready-to-use billing solution for your cloud, contact Mirantis or one of our partners.
A cloud administrator can securely expose part of a MOSK cluster message bus to the outside world. This enables an external consumer to subscribe to the notification messages emitted by the cluster services.
Important
The latest OpenStack release available in MOSK supports notifications from the following services:
Block storage (OpenStack Cinder)
DNS (OpenStack Designate)
Image (OpenStack Glance)
Orchestration (OpenStack Heat)
Bare Metal (OpenStack Ironic)
Identity (OpenStack Keystone)
Shared Filesystems (OpenStack Manila)
Instance High Avalability (OpenStack Masakari)
Networking (OpenStack Neutron)
Compute (OpenStack Nova)
To enable the external notification endpoint, add the following structure
to the OpenStackDeployment custom resource. For example:
spec:
features:
messaging:
notifications:
external:
enabled: true
topics:
- external-consumer-A
- external-consumer-2
For each topic name specified in the topics field, MOSK
creates a topic exchange in its RabbitMQ cluster together with a set of queues
bound to this topic. All enabled MOSK services will
publish their notification messages to all configured topics so that
multiple consumers can receive the same messages in parallel.
A topic name must follow Kubernetes standard format for object names and IDs
that is only lowercase alphanumeric characters, -, or . The topic
name notifications is reserved for the internal use.
MOSK supports the connection to message bus (RabbitMQ) through an encrypted or non-encrypted endpoint. Once connected, it supports authentication through either a plain text user name and password or mutual TLS authentication using encrypted X.509 client certificates.
Each topic exchange is protected by automatically generated authentication
credentials and certificates for secure connection that are stored as a secret
in the openstack-external namespace of a MOSK underlying
Kubernetes cluster. A secret is identified by the name of the topic. The list
of attributes for the secret object includes:
hostsThe IP addresses which an external notification endpoint is available on
port_amqp,port_amqp-tlsThe TCP ports which external notification endpoint is available on
vhostThe name of the RabbitMQ virtual host which the topic queues are created on
username,passwordAuthentication data
ca_certThe client CA certificate
client_certThe client certificate
client_keyThe client private key
For the configuration example above, the following objects will be created:
kubectl -n openstack-external get secret
NAME TYPE DATA AGE
openstack-external-consumer-A-notifications Opaque 4 4m51s
openstack-external-consumer-2-notifications Opaque 4 4m51s
Tungsten Fabric¶
Tungsten Fabric provides basic L2/L3 networking to an OpenStack environment running on the MKE cluster and includes the IP address management, security groups, floating IP addresses, and routing policies functionality. Tungsten Fabric is based on overlay networking, where all virtual machines are connected to a virtual network with encapsulation (MPLSoGRE, MPLSoUDP, VXLAN). This enables you to separate the underlay Kubernetes management network. A workload requires an external gateway, such as a hardware EdgeRouter or a simple gateway to route the outgoing traffic.
The Tungsten Fabric vRouter uses different gateways for the control and data planes.
Tungsten Fabric cluster¶
All services of Tungsten Fabric are delivered as separate containers, which are deployed by the Tungsten Fabric Operator (TFO). Each container has an INI-based configuration file that is available on the host system. The configuration file is generated automatically upon the container start and is based on environment variables provided by the TFO through Kubernetes ConfigMaps.
The main Tungsten Fabric containers run with the host network as
DeploymentSet, without using the Kubernetes networking layer. The services
listen directly on the host network interface.
The following diagram describes the minimum production installation of Tungsten Fabric with a Mirantis OpenStack for Kubernetes (MOSK) deployment.
For the details about the Tungsten Fabric services included in MOSK deployments and the types of traffic and traffic flow directions, see the subsections below.
Tungsten Fabric cluster components¶
This section describes the Tungsten Fabric services and their distribution across the Mirantis OpenStack for Kubernetes (MOSK) deployment.
The Tungsten Fabric services run mostly as DaemonSets in separate containers for each service. The deployment and update processes are managed by the Tungsten Fabric Operator. However, Kubernetes manages the probe checks and restart of broken containers.
All configuration and control services run on the Tungsten Fabric Controller nodes.
Service name |
Service description |
|---|---|
|
Exposes a REST-based interface for the Tungsten Fabric API. |
|
Provisions the node for execution of configuration services. |
|
Communicates with the cluster gateways using BGP and with the vRouter agents using XMPP, as well as redistributes appropriate networking information. |
|
Provisions the node for execution of configuration services. |
|
Manages physical networking devices using |
|
Using the |
|
The customized Berkeley Internet Name Domain (BIND) daemon of
Tungsten Fabric that manages DNS zones for the |
|
Listens to configuration changes performed by a user and generates corresponding system configuration objects. In multi-node deployments, it works in the active-backup mode. |
|
Listens to configuration changes of |
|
Consists of the |
All analytics services run on Tungsten Fabric analytics nodes.
Service name |
Service description |
|---|---|
|
Evaluates and manages the alarms rules. |
|
Provides a REST API to interact with the Cassandra analytics database. |
|
Collects all Tungsten Fabric analytics process data and sends
this information to the Tungsten Fabric |
|
Provisions the init model if needed. Collects data of the |
|
Collects and analyzes data from all Tungsten Fabric services. |
|
Handles the queries to access data from the Cassandra database. |
|
Receives the authorization and configuration of the physical routers
from the |
|
Reads the SNMP information from the physical router user-visible entities (UVEs), creates a neighbor list, and writes the neighbor information to the physical router UVEs. The Tungsten Fabric web UI uses the neighbor list to display the physical topology. |
The Tungsten Fabric vRouter provides data forwarding to an OpenStack tenant instance and reports statistics to the Tungsten Fabric analytics service. The Tungsten Fabric vRouter is installed on all OpenStack compute nodes. Mirantis OpenStack for Kubernetes (MOSK) supports the kernel-based deployment of the Tungsten Fabric vRouter.
Service name |
Service description |
|---|---|
|
Connects to the Tungsten Fabric Controller container and the Tungsten Fabric DNS system using the Extensible Messaging and Presence Protocol (XMPP). The vRouter Agent acts as a local control plane. Each Tungsten Fabric vRouter Agent is connected to at least two Tungsten Fabric controllers in an active-active redundancy mode. The Tungsten Fabric vRouter Agent is responsible for all networking-related functions including routing instances, routes, and others. The Tungsten Fabric vRouter uses different gateways for the control and data planes. For example, the Linux system gateway is located on the management network, and the Tungsten Fabric gateway is located on the data plane network. |
|
Provisions the node for the vRouter agent execution. |
The following diagram illustrates the Tungsten Fabric kernel vRouter set up by the TF operator:
On the diagram above, the following types of networks interfaces are used:
eth0- for the management (PXE) network (eth1andeth2are the slave interfaces ofBond0)Bond0.x- for the MKE control plane networkBond0.y- for the MKE data plane network
Service name |
Service description |
|---|---|
|
|
|
The Kubernetes operator that enables the Cassandra clusters creation and management. |
|
Handles the messaging bus and generates alarms across the Tungsten Fabric analytics containers. |
|
The Kubernetes operator that enables Kafka clusters creation and management. |
|
Stores the physical router UVE storage and serves as a messaging bus for event notifications. |
|
The Kubernetes operator that enables Redis clusters creation and management. |
|
Holds the active-backup status for the |
|
The Kubernetes operator that enables ZooKeeper clusters creation and management. |
|
Exchanges messages between API servers and original request senders. |
|
The Kubernetes operator that enables RabbitMQ clusters creation and management. |
All Tungsten Fabric plugin services are installed on the OpenStack Controller (Rockoon) nodes.
Service name |
Service description |
|---|---|
|
The Neutron server that includes the Tungsten Fabric plugin. |
|
The Octavia API that includes the Tungsten Fabric Octavia driver. |
|
The Heat API that includes the Tungsten Fabric Heat resources and templates. |
Along with the Tungsten Fabric services, MOSK deploys and
updates special image precaching DaemonSets when the kind TFOperator
resource is created or image references in it get updated.
These DaemonSets precache container images on Kubernetes nodes minimizing
possible downtime when updating container images. Cloud operator can
disable image precaching through the TFOperator resource.
See also
Tungsten Fabric traffic flow¶
This section describes the types of traffic and traffic flow directions in a Mirantis OpenStack for Kubernetes (MOSK) cluster.
The following diagram illustrates all types of UI and API traffic in a Mirantis OpenStack for Kubernetes cluster, including the monitoring and OpenStack API traffic. The OpenStack Dashboard pod hosts Horizon and acts as a proxy for all other types of traffic. TLS termination is also performed for this type of traffic.
SDN or Tungsten Fabric traffic goes through the overlay Data network and processes east-west and north-south traffic for applications that run in a MOSK cluster. This network segment typically contains tenant networks as separate MPLS-over-GRE and MPLS-over-UDP tunnels. The traffic load depends on the workload.
The control traffic between the Tungsten Fabric controllers, edge routers, and vRouters uses the XMPP with TLS and iBGP protocols. Both protocols produce low traffic that does not affect MPLS over GRE and MPLS over UDP traffic. However, this traffic is critical and must be reliably delivered. Mirantis recommends configuring higher QoS for this type of traffic.
The following diagram displays both MPLS over GRE/MPLS over UDP and iBGP and XMPP traffic examples in a MOSK cluster:
Tungsten Fabric lifecycle management¶
Mirantis OpenStack for Kubernetes (MOSK) provides the Tungsten Fabric lifecycle management including pre-deployment custom configurations, updates, data backup and restoration, as well as handling partial failure scenarios, by means of the Tungsten Fabric operator.
This section is intended for the cloud operators who want to gain insight into the capabilities provided by the Tungsten Fabric operator along with the understanding of how its architecture allows for easy management while addressing the concerns of users of Tungsten Fabric-based MOSK clusters.
Tungsten Fabric Operator¶
The Tungsten Fabric Operator (TFO) is based on the Kubernetes operator SDK project. The Kubernetes operator SDK is a framework that uses the controller-runtime library to make writing operators easier by providing the following:
High-level APIs and abstractions to write the operational logic more intuitively.
Tools for scaffolding and code generation to bootstrap a new project fast.
Extensions to cover common operator use cases.
The TFO deploys the following sub-operators. Each sub-operator handles a separate part of a TF deployment:
Network |
Description |
|---|---|
TFControl |
Deploys the Tungsten Fabric control services, such as:
|
TFConfig |
Deploys the Tungsten Fabric configuration services, such as: |
TFAnalytics |
Deploys the Tungsten Fabric analytics services, such as: |
TFVrouter |
Deploys a vRouter on each compute node with the following services:
|
TFWebUI |
Deploys the following web UI services:
|
TFTool |
Deploys the following tools for debug purposes:
|
TFTest |
An operator to run Tempest tests. |
- 0(1,2,3,4,5,6,7)
Since MOSK 24.3,
Provisioneris a separate component for the vRouter, deployed as thetf-vrouter-provisionerDaemonSet. TheNodeManagerservice is no longer deployed in TF setups.
Besides the sub-operators that deploy TF services, TFO uses operators to deploy and maintain third-party services, such as different types of storage, cache, message system, and so on. The following table describes all third-party operators:
Network |
Description |
|---|---|
casandra-operator |
An upstream operator that automates the Cassandra HA storage operations for the configuration and analytics data. |
zookeeper-operator |
An upstream operator for deployment and automation of a ZooKeeper cluster. |
kafka-operator |
An operator for the Kafka cluster used by analytics services. |
redis-operator |
An upstream operator that automates the Redis cluster deployment and keeps it healthy. |
rabbitmq-operator |
An operator for the messaging system based on RabbitMQ. |
The following diagram illustrates a simplified TFO workflow:
TFOperator custom resource¶
The resource of kind TFOperator is a custom resource defined by a resource
of kind CustomResourceDefinition.
The CustomResourceDefinition resource in Kubernetes uses the OpenAPI
Specification version 2 to specify the schema of the defined resource.
The Kubernetes API outright rejects the resources that do not pass this schema
validation. Along with schema validation, TFOperator uses
ValidatingAdmissionWebhook for extended validations when a custom resource
is created or updated.
Important
Since 24.1, MOSK introduces the technical preview support for the API v2 for the Tungsten Fabric Operator. This version of the Tungsten Fabric Operator API aligns with the OpenStack Controller API and provides better interface for advanced configurations. Refer to Key differences between TFOperator API v1alpha1 and v2 for details.
For the list of configuration options available to a cloud operator, refer to Tungsten Fabric configuration. Also, check out the Tungsten Fabric Operator resources document of the MOSK version that your cluster has been deployed with.
Tungsten Fabric Operator uses ValidatingAdmissionWebhook to validate
environment variables set to Tungsten Fabric components upon the TFOperator
object creation or update. The following validations are performed:
Environment variables passed to the Tungsten Fabric components containers
Mapping between
tfVersionandtfImageTag, if definedSchedule for
dbBackupData capacity format
Feature variable values
Availability of the
dataStorageClassclass
If required, you can disable ValidatingAdmissionWebhook through the
TFOperator HelmBundle resource:
apiVersion: lcm.mirantis.com/v1alpha1
kind: HelmBundle
metadata:
name: tungstenfabric-operator
namespace: tf
spec:
releases:
- name: tungstenfabric-operator
values:
admission:
enabled: false
Warning
The features section of the TFOperator specification
allows for easy configuration of all Tungsten Fabric features. Mirantis
recommends updating the environment variables through envSettings
directly.
Environment variables |
Tungsten Fabric service and |
|---|---|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Environment variables |
Tungsten Fabric components and containers |
|---|---|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Key differences between TFOperator API v1alpha1 and v2¶
This section outlines the main differences between the v1alpha1 and v2 versions of the TFOperator API:
Introduction of the
featuressection:All non-default Tungsten Fabric and Tungsten Fabric Operator features can now be set in the
featuressection.Setting environment variables is no longer necessary but can still be done using the
envSettingfield in each Tungsten Fabric service section.
Relocation of
CustomSpecfrom the vRouter agent specification to the nodes section.Reorganization of the
controllerssection:The
controllerssection has been integrated into theservicessection.The
servicessection is now divided into groups:analytics,config,control,vRouter, andwebUI.Configuration of third-party services can be performed through the
analyticsorconfigsections.
Configuration of the logging levels can be performed using the
loggingfield, which is a separate field in each Tungsten Fabric services configuration.Movement of the
dataStorageClassandtfVersionfields to the upper level of the specification.Introduction of the
devOptionssection enabling the setup of experimental development-related options.
See also
Tungsten Fabric configuration¶
Mirantis OpenStack for Kubernetes (MOSK) allows you to easily adapt
your Tungsten Fabric deployment to the needs of your environment through the
TFOperator custom resource.
This section includes custom configuration details available to you.
Important
Since 24.1, MOSK introduces the technical preview support for the API v2 for the Tungsten Fabric Operator. This version of the Tungsten Fabric Operator API aligns with the OpenStack Controller API and provides better interface for advanced configurations. In MOSK 24.1, the API v2 is available only for the new product deployments with Tungsten Fabric.
Since 24.2, the API v2 becomes default for new product deployments and
includes the ability to convert existing v1alpha1 TFOperator to v2
during update.
During the update to the 24.3 series, the old Tungsten Fabric cluster
configuration API v1alpha1 is automatically converted and replaced
with the v2 version. Therefore, since MOSK 24.3,
start using the v2 TFOperator custom resource for any updates.
The v1alpha1 TFOperator custom resource remains in the cluster
but is no longer reconciled and will be automatically removed in
MOSK 25.1.
Cassandra configuration¶
This section describes the Cassandra configuration through the Tungsten Fabric Operator custom resource.
By default, Tungsten Fabric Operator sets up the following resource limits for Cassandra analytics and configuration StatefulSets:
Limits:
cpu: 8
memory: 32Gi
Requests:
cpu: 1
memory: 16Gi
This is a verified configuration suitable for most cases. However, if nodes
are under a heavy load, the KubeContainerCPUThrottlingHigh StackLight alert
may raise for Tungsten Fabric Pods of the tf-cassandra-analytics and
tf-cassandra-config StatefulSets. If such alerts appear constantly, you can
increase the limits through the TFOperator custom resource. For example:
spec:
services:
analytics:
enabled: true
cassandra:
resources:
limits:
cpu: "12"
memory: 32Gi
requests:
cpu: "2"
memory: 16Gi
config:
cassandra:
resources:
limits:
cpu: "12"
memory: 32Gi
requests:
cpu: "2"
memory: 16Gi
spec:
controllers:
cassandra:
deployments:
- name: tf-cassandra-config
resources:
limits:
cpu: "12"
memory: 32Gi
requests:
cpu: "2"
memory: 16Gi
- name: tf-cassandra-analytics
resources:
limits:
cpu: "12"
memory: 32Gi
requests:
cpu: "2"
memory: 16Gi
To specify custom configurations for Cassandra clusters, use the
configOptions settings in the TFOperator custom resource.
For example, you may need to increase the file cache size in case
of a heavy load on the nodes labeled with tfanalyticsdb=enabled
or tfconfigdb=enabled:
spec:
services:
analytics:
enabled: true
cassandra:
configOptions:
file_cache_size_in_mb: 1024
spec:
controllers:
cassandra:
deployments:
- name: tf-cassandra-analytics
configOptions:
file_cache_size_in_mb: 1024
Custom vRouter settings¶
TechPreview
Depending on the Tungsten Fabric Operator API version in use, proceed with one of the following options:
To specify custom settings for the Tungsten Fabric vRouter nodes, for
example, to change the name of the tunnel network interface or enable
debug level logging on some subset of nodes, use the nodes settings in
the TFOperator custom resource.
For example, to enable debug level logging on a specific node or multiple nodes:
spec:
nodes:
<CUSTOMSPEC-NAME>:
labels:
name: <NODE-LABEL>
value: <NODE-LABEL-VALUE>
nodeVRouter:
enabled: true
envSettings:
agent:
env:
- name: LOG_LEVEL
value: SYS_DEBUG
To specify custom settings for the Tungsten Fabric vRouter nodes, for
example, to change the name of the tunnel network interface or enable
debug level logging on some subset of nodes, use the customSpecs
settings in the TFOperator custom resource.
For example, to enable debug level logging on a specific node or multiple nodes:
spec:
controllers:
tf-vrouter:
agent:
customSpecs:
- name: <CUSTOMSPEC-NAME>
label:
name: <NODE-LABEL>
value: <NODE-LABEL-VALUE>
containers:
- name: agent
env:
- name: LOG_LEVEL
value: SYS_DEBUG
Caution
The customspecs:name value must follow the RFC 1123
international format. Verify that the name of a DaemonSet object
is a valid DNS subdomain name.
The customSpecs parameter inherits all settings for the tf-vrouter
containers that are set on the spec:controllers:agent level and
overrides or adds additional parameters. The example configuration above
overrides the logging level from SYS_INFO, which is the default logging
level, to SYS_DEBUG.
For clusters with a multi-rack architecture, you may need to redefine the
gateway IP for the Tungsten Fabric vRouter nodes using the
VROUTER_GATEWAY parameter. For details, see Multi-rack architecture.
Control plane traffic interface¶
By default, the TF control service uses the management interface for
the BGP and XMPP traffic. You can change the control service interface
using the controlInterface parameter in the TFOperator custom resource,
for example, to combine the BGP and XMPP traffic with the data (tenant)
traffic:
spec:
features:
control:
controlInterface: <tunnel-interface>
spec:
settings:
controlInterface: <tunnel-interface>
Traffic encapsulation¶
Tungsten Fabric implements cloud tenants’ virtual networks as Layer 3 overlays. Tenant traffic gets encapsulated into one of the supported protocols and is carried over the infrastructure network between 2 compute nodes or a compute node and an edge router device.
In addition, Tungsten Fabric is capable of exchanging encapsulated traffic with external systems in order to build advanced virtual networking topologies, for example, BGP VPN connectivity between 2 MOSK clouds or a MOSK cloud and a cloud tenant premises.
MOSK supports the following encapsulation protocols:
- MPLS over Generic Routing Encapsulation (GRE)
A traditional encapsulation method supported by several router vendors, including Cisco and Juniper. The feature is applicable when other encapsulation methods are not available. For example, an SDN gateway runs software that does not support MPLS over UDP.
- MPLS over User Datagram Protocol (UDP)
A variation of the MPLS over GRE mechanism. It is the default and the most frequently used option in MOSK. MPLS over UDP replaces headers in UDP packets. In this case, a UDP port stores a hash of the packet payload (entropy). It provides a significant benefit for equal-cost multi-path (ECMP) routing load balancing. MPLS over UDP and MPLS over GRE transfer Layer 3 traffic only.
- Virtual Extensible LAN (VXLAN) TechPrev
The combination of VXLAN and EVPN technologies is often used for creating advanced cloud networking topologies. For example, it can provide transparent Layer 2 interconnections between Virtual Network Functions running on top of the cloud and physical traffic generator appliances hosted somewhere else.
The ENCAP_PRIORIY parameter defines the priority in which the
encapsulation protocols are attempted to be used when setting the BGP VPN
connectivity between the cloud and external systems.
By default, the encapsulation order is set to MPLSoUDP,MPLSoGRE,VXLAN.
The cloud operator can change it depending their needs in the TFOperator
custom resource as it is illustrated in Configuring encapsulation.
The list of supported encapsulated methods along with their order is shared between BGP peers as part of the capabilities information exchange when establishing a BGP session. Both parties must support the same encapsulation methods to build a tunnel for the network traffic.
For example, if the cloud operator wants to set up a Layer 2 VPN between the cloud and their network infrastructure, they configure the cloud’s virtual networks with VXLAN identifiers (VNIs) and do the same on the other side, for example, on a network switch. Also, VXLAN must be set in the first position in encapsulation priority order. Otherwise, VXLAN tunnels will not get established between endpoints, even though both endpoints may support the VXLAN protocol.
However, setting VXLAN first in the encapsulation priority order will not enforce VXLAN encapsulation between compute nodes or between compute nodes and gateway routers that use Layer 3 VPNs for communication.
The TFOperator custom resource allows you to define encapsulation settings
for your Tungsten Fabric cluster.
Important
The TFOperator custom resource must be the only place to
configure the cluster encapsulation. Performing these configurations through
the Tungsten Fabric web UI, CLI, or API does not provide the configuration
persistency, and the settings defined this way may get reset to defaults
during the cluster services restart or update.
Note
Defining the default values for encapsulation parameters in
the TFOperator custom resource is unnecessary.
Depending on the Tungsten Fabric operator API version in use, proceed with one of the following options:
Parameter |
Default value |
Description |
|---|---|---|
|
|
Defines the encapsulation priority order. |
|
|
Defines the Virtual Network ID type. The list of possible values includes:
|
Example configuration:
features:
config:
vxlanVnIdMode: automatic
encapPriority: VXLAN,MPLSoUDP,MPLSoGRE
Parameter |
Default value |
Description |
|---|---|---|
|
|
Defines the encapsulation priority order. |
|
|
Defines the Virtual Network ID type. The list of possible values includes:
Typically, for a Layer 2 VPN use case, the |
Example configuration:
controllers:
tf-config:
provisioner:
containers:
- env:
- name: VXLAN_VN_ID_MODE
value: automatic
- name: ENCAP_PRIORITY
value: VXLAN,MPLSoUDP,MPLSoGRE
name: provisioner
Autonomous System Number (ASN)¶
In the routing fabric of a data centre, a MOSK cluster with Tungsten Fabric enabled can be represented either by a separate Autonomous System (AS) or as part of a bigger autonomous system. In either case, Tungsten Fabric needs to participate in the BGP peering, exchanging routes with external devices and within the cloud.
The Tungsten Fabric Controller acts as an internal (iBGP) route reflector for
the cloud AS by populating /32 routes pointing to VMs across all compute
nodes as well as the cloud’s edge gateway devices in case they belong to the
same AS. Apart from being an iBGP router reflector for the cloud AS, the
Tungsten Fabric Controller can act as a BGP peer for autonomous systems
external to the cloud, for example, for the AS configured across the
leaf-spine fabric of the data center.
The Autonomous System Number (ASN) setting contains the unique identifier of the autonomous system that the MOSK cluster with Tungsten Fabric belongs to. The ASN number does not affect the internal iBGP communication between vRouters running on the compute nodes. Such communication will work regardless of the ASN number settings. However, any network appliance that is not managed by the Tungsten Fabric control plane will have BGP configured manually. Therefore, the ASN settings should be configured accordingly on both sides. Otherwise, it would result in the inability to establish BPG sessions, regardless of whether the external device peers with Tungsten Fabric over iBGP or eBGP.
The TFOperator custom resource enables you to define ASN settings for
your Tungsten Fabric cluster.
Important
The TFOperator CR must be the only place to configure
the cluster ASN. Performing these configurations through the Tungsten
Fabric web UI, CLI, or API does not provide the configuration persistency,
and the settings defined this way may get reset to defaults during
the cluster services restart or update.
Note
Defining the default values for ASN parameters in the Tungsten Fabric Operator custom resource is unnecessary.
Depending on the Tungsten Fabric Operator API version in use, proceed with one of the following options:
Parameter |
Default value |
Description |
|---|---|---|
|
|
Defines ASN of the control node. |
|
|
Enables the 4-byte ASN format. |
Example configuration:
features:
config:
bgpAsn: 64515
enable4ByteAS: true
Parameter |
Default value |
Description |
|---|---|---|
|
|
Defines ASN of the control node. |
|
|
Enables the 4-byte ASN format. |
Example configuration:
controllers:
tf-config:
provisioner:
containers:
- env:
- name: BGP_ASN
value: "64515"
- name: ENABLE_4BYTE_AS
value: "true"
name: provisioner
tf-control:
provisioner:
containers:
- env:
- name: BGP_ASN
value: "64515"
name: provisioner
Access to external DNS¶
By default, the Tungsten Fabric tf-control-dns-external service is created
to expose the Tungsten Fabric control dns. You can disable creation of this
service through the enableDNSExternal parameter in the TFOperator
custom resource. For example:
spec:
features:
control:
enableDNSExternal: false
spec:
controllers:
tf-control:
dns:
enableDNSExternal: false
Gateway for vRouter data plane network¶
If an edge router is accessible from the data plane through a gateway, define
the vRouter gateway in the TFOperator custom resource. Otherwise,
the default system gateway is used.
Depending on the Tungsten Fabric Operator API version in use, proceed with one of the following configurations:
Define the vRouterGateway parameter in the features section of
the TFOperator custom resource:
spec:
features:
vRouter:
vRouterGateway: <data-plane-network-gateway>
You can also configure the parameter for Tungsten Fabric vRouter in the DPDK mode:
spec:
services:
vRouter:
agentDPDK:
enabled: true
envSettings:
agent:
env:
- name: VROUTER_GATEWAY
value: <data-plane-network-gateway>
Define the VROUTER_GATEWAY parameter in the TFOperator custom
resource:
spec:
controllers:
tf-vrouter:
agent:
containers:
- name: agent
env:
- name: VROUTER_GATEWAY
value: <data-plane-network-gateway>
You can also configure the parameter for Tungsten Fabric vRouter in the DPDK mode:
spec:
controllers:
tf-vrouter:
agent-dpdk:
enabled: true
containers:
- name: agent
env:
- name: VROUTER_GATEWAY
value: <data-plane-network-gateway>
Tungsten Fabric image precaching¶
By default, MOSK deploys image precaching DaemonSets
to minimize possible downtime when updating container images. You can disable
creation of these DaemonSets by setting the imagePreCaching parameter in
the TFOperator custom resource to false:
spec:
features:
imagePreCaching: false
spec:
settings:
imagePreCaching: false
Graceful restart and long-lived graceful restart¶
Available since MOSK 23.2 for Tungsten Fabric 21.4 only TechPreview
Graceful restart and long-lived graceful restart are vital mechanisms within BGP (Border Gateway Protocol) routing, designed to optimize the routing tables convergence in scenarios where a BGP router restarts or a networking failure is experienced, leading to interruptions of router peering.
During a graceful restart, a router can signal its BGP peers about its impending restart, requesting them to retain the routes it had previously advertised as active. This allows for seamless network operation and minimal disruption to data forwarding during the router downtime.
The long-lived aspect of the long-lived graceful restart extends the graceful restart effectiveness beyond the usual restart duration. This extension provides an additional layer of resilience and stability to BGP routing updates, bolstering the network ability to manage unforeseen disruptions.
Caution
Mirantis does not generally recommend using the graceful restart and long-lived graceful restart features with the Tungsten Fabric XMPP helper, unless the configuration is done by proficient operators with at-scale expertise in networking domain and exclusively to address specific corner cases.
Tungsten Fabric Operator allows for easy enablement and configuration
of the graceful restart and long-lived graceful restart features through
the TFOperator custom resource:
spec:
features:
control:
gracefulRestart:
enabled: <BOOLEAN>
bgpHelperEnabled: <BOOLEAN>
xmppHelperEnabled: <BOOLEAN>
restartTime: <TIME_IN_SECONDS>
llgrRestartTime: <TIME_IN_SECONDS>
endOfRibTimeout: <TIME_IN_SECONDS>
spec:
settings:
settings:
gracefulRestart:
enabled: <BOOLEAN>
bgpHelperEnabled: <BOOLEAN>
xmppHelperEnabled: <BOOLEAN>
restartTime: <TIME_IN_SECONDS>
llgrRestartTime: <TIME_IN_SECONDS>
endOfRibTimeout: <TIME_IN_SECONDS>
Parameter |
Default value |
Description |
|---|---|---|
|
|
Enables or disables graceful restart and long-lived graceful restart features. |
|
|
Specifies the time interval, when the Tungsten Fabric control services act as a graceful restart helper to the edge router or any other BGP peer by retaining the routes learned from this peer and advertising them to the rest of the network as applicable. Note BGP peer should support and be configured with graceful restart for all of the address families used. |
|
|
Specifies the time interval, when the datapath agent should retain the last route path from the Tungsten Fabric Controller when an XMPP-based connection is lost. |
|
|
Configures a non-zero restart time in seconds to advertise for graceful restart capability from peers. |
|
|
Specifies the amount of time in seconds the vRouter datapath should keep advertised routes from the Tungsten Fabric control services, when an XMPP connection between the control and vRouter agent services is lost. Note When graceful restart and long-lived graceful restart are both configured, the duration of the long-lived graceful restart timer is the sum of both timers. |
|
|
Specifies the amount of time in seconds a control node waits to remove stale routes from a vRouter agent Routing Information Base (RIB). |
Configuring the protocol for connecting to Cassandra clusters¶
To streamline and improve the efficiency of communication between clients and the database, Cassandra is transitioning away from the Thrift protocol in favor of the Query Language (CQL) protocol starting with MOSK 24.1. Since MOSK 24.2, Cassandra uses the CQL protocol by default.
CQL provides a more user-friendly and SQL-like interface for interacting with the database. With the move towards CQL, the Thrift-based client drivers are no longer actively supported encouraging the users to migrate to CQL-based client drivers to take advantage of new features and improvements in Cassandra.
If your cluster is running MOSK 24.1.x, you can enable the CQL protocol proceeding with one of the options below depending on the Tungsten Fabric Operator API version in use.
During update to MOSK 24.2, switching from Thrift to CQL
is performed automatically. While it is possible to switch back to Thrift,
Mirantis does not recommend it. If you choose to do so, specify thrift
instead of cql in the configuration examples below.
Define the cassandraDriver parameter in the devOptions section of
the TFOperator custom resource:
spec:
devOptions:
cassandraDriver: cql
Define the CONFIGDB_CASSANDRA_DRIVER variable for the tf-analytics,
tf-config, and tf-control controllers in the TFOperator custom
resource:
spec:
controllers:
tf-analytics:
alarm-gen:
containers:
- env:
- name: CONFIGDB_CASSANDRA_DRIVER
value: cql
name: alarm-gen
api:
containers:
- env:
- name: CONFIGDB_CASSANDRA_DRIVER
value: cql
name: api
collector:
containers:
- env:
- name: CONFIGDB_CASSANDRA_DRIVER
value: cql
name: collector
snmp:
containers:
- env:
- name: CONFIGDB_CASSANDRA_DRIVER
value: cql
name: snmp
topology:
containers:
- env:
- name: CONFIGDB_CASSANDRA_DRIVER
value: cql
name: topology
tf-config:
api:
containers:
- env:
- name: CONFIGDB_CASSANDRA_DRIVER
value: cql
name: api
devicemgr:
containers:
- env:
- name: CONFIGDB_CASSANDRA_DRIVER
value: cql
name: devicemgr
schema:
containers:
- env:
- name: CONFIGDB_CASSANDRA_DRIVER
value: cql
name: schema
svc-monitor:
containers:
- env:
- name: CONFIGDB_CASSANDRA_DRIVER
value: cql
name: svc-monitor
tf-control:
control:
containers:
- env:
- name: CONFIGDB_CASSANDRA_DRIVER
value: cql
name: control
dns:
containers:
- env:
- name: CONFIGDB_CASSANDRA_DRIVER
value: cql
name: dns
SR-IOV Spoof Check control for Tungsten Fabric¶
Available since MOSK 24.2 TechPreview
MOSK provides the capability to enable SR-IOV Spoof Check control with the Neutron Tungsten Fabric backend.
The capability can be useful for certain network configurations. For example, you might need to allow traffic from a virtual function interface even when its MAC address does not match the MAC address inside the virtual machine. In this scenario, known as MAC spoofing, disabling spoof check enables the traffic to pass through regardless of the MAC address mismatch.
Caution
Certain NICs and drivers may not handle the spoofchk setting.
For example, the Intel 82599ES NIC paired with the ixgbe driver disregards
the spoofchk setting when VLAN tagging is enabled. Therefore, ensure
compatibility with your hardware configuration regarding spoofchk
handling before proceeding.
To enable SR-IOV Spoof Check control for Tungsten Fabric, enable SR-IOV
interfaces handling by Nova os-vif plugin in the OpenStackDeployment
custom resource:
services:
compute:
nova:
values:
conf:
nova:
workarounds:
pass_hwveb_ports_to_os_vif_plugin: true
Now, you can enable and disable spoof checking for certain SR-IOV ports through the OpenStack CLI. To disable spoof checking on an SR-IOV port:
openstack port set --no-security-group --disable-port-security <SRIOV-PORT>
To enable spoof checking on an SR-IOV port:
openstack port set --enable-port-security <SRIOV-PORT>
Availability zones¶
The Tungsten Fabric Operator provides a capability to configure
the netns_availability_zone parameter of the Tungsten Fabric
svc-monitor service through the netnsAZ parameter. This configuration
enables MOSK users to specify an availability zone for
Tungsten Fabric instances, such as HAProxy (load balancer instances) or SNAT
routers.
Configuration:
spec:
features:
config:
netnsAZ: <NETNS_AVAILABILITY_ZONE>
Tungsten Fabric database¶
Tungsten Fabric (TF) uses Cassandra and ZooKeeper to store its data. Cassandra is a fault-tolerant and horizontally scalable database that provides persistent storage of configuration and analytics data. ZooKeeper is used by TF for allocation of unique object identifiers and transactions implementation.
To prevent data loss, Mirantis recommends that you simultaneously back up the ZooKeeper database dedicated to configuration services and the Cassandra database.
The backup of database must be consistent across all systems because the state of the Tungsten Fabric databases is associated with other system databases, such as OpenStack databases.
Periodic Tungsten Fabric database backups¶
MOSK enables you to perform the automatic TF
data backup in the JSON format using the tf-dbbackup-job cron job.
By default, it is disabled. To back up the TF databases, enable
tf-dbBackup in the TF Operator custom resource:
spec:
features:
dbBackup:
enabled: true
spec:
controllers:
tf-dbBackup:
enabled: true
By default, the tf-dbbackup-job job is scheduled for weekly execution,
allocating PVC of 5 Gi size for storing backups and keeping 5 previous
backups. To configure the backup parameters according to the needs of your
cluster, use the following structure:
spec:
features:
dbBackup:
enabled: true
dataCapacity: 30Gi
schedule: "0 0 13 * 5"
storedBackups: 10
spec:
controllers:
tf-dbBackup:
enabled: true
dataCapacity: 30Gi
schedule: "0 0 13 * 5"
storedBackups: 10
To temporarily disable tf-dbbackup-job, suspend the job:
spec:
features:
dbBackup:
enabled: true
suspend: true
spec:
controllers:
tf-dbBackup:
enabled: true
suspend: true
To delete the tf-dbbackup-job job, disable tf-dbBackup in
the TF Operator custom resource:
spec:
features:
dbBackup:
enabled: false
spec:
controllers:
tf-dbBackup:
enabled: false
Remote storage for Tungsten Fabric database backups¶
Available since MOSK 23.2 TechPreview
MOSK supports configuring a remote NFS storage for TF data backups through the TF Operator custom resource:
spec:
features:
dbBackup:
enabled: true
backupType: "pv_nfs"
nfsOptions:
path: <PATH_TO_SHARE_FOLDER_ON_SERVER>
server: <IP_ADDRESS/DNS_NAME_OF_SERVER>
spec:
controllers:
tf-dbBackup:
enabled: true
backupType: "pv_nfs"
nfsOptions:
path: <PATH_TO_SHARE_FOLDER_ON_SERVER>
server: <IP_ADDRESS/DNS_NAME_OF_SERVER>
If PVC backups were used previously, the old PVC will not be utilized. You can delete it with the following command:
kubectl -n tf delete pvc <TF_DB_BACKUP_PVC>
Tungsten Fabric services¶
The section explains specifics of the Tungsten Fabric services provided by Mirantis OpenStack for Kubernetes (MOSK). The list of the services and their supported features included in this section is not full and is being constantly amended based on the complexity of the architecture and use of a particular service.
Tungsten Fabric load balancing (HAProxy)¶
Note
Since 23.1, MOSK provides technology preview for Octavia Amphora load balancing. To start experimenting with the new load balancing solution, refer to Octavia Amphora load balancing.
MOSK ensures Octavia with Tungsten Fabric integration by OpenStack Octavia Driver with Tungsten Fabric HAProxy as a backend.
The Tungsten Fabric-based MOSK deployment supports creation, update, and deletion operations with the following standard load balancing API entities:
Load balancers
Note
For a load balancer creation operation, the driver supports only the
vip-subnet-idargument, thevip-network-idargument is not supported.Listeners
Pools
Health monitors
The Tungsten Fabric-based MOSK deployment does not support the following load balancing capabilities:
L7 load balancing capabilities, such as L7 policies, L7 rules, and others
Setting specific availability zones for load balancers and their resources
Using of the UDP protocol
Operations with Octavia quotas
Operations with Octavia flavors
Warning
The Tungsten Fabric-based MOSK deployment enables you to manage the load balancer resources by means of the OpenStack CLI or OpenStack Horizon. Do not perform any manipulations with the load balancer resources through the Tungsten Fabric web UI because in this case the changes will not be reflected on the OpenStack API side.
Octavia Amphora load balancing¶
Available since MOSK 23.1 TechPreview
Octavia Amphora (Amphora v2) load balancing provides a scalable and flexible solution for load balancing in cloud environments. MOSK deploys Amphora load balancer on each node of the OpenStack environment ensuring that load balancing services are easily accessible, highly scalable, and highly reliable.
Compared to the Octavia Tungsten Fabric driver for LBaaS v2 solution, Amphora offers several advanced features including:
Full compatibility with the Octavia API, which provides a standardized interface for load balancing in MOSK OpenStack environments. This makes it easier to manage and integrate with other OpenStack services.
Layer 7 policies and rules, which allow for more granular control over traffic routing and load balancing decisions. This enables users to optimize their application performance and improve the user experience.
Support for the UDP protocol, which is commonly used for real-time communications and other high-performance applications. This enables users to deploy a wider range of applications with the same load balancing infrastructure.
By default, MOSK uses the Octavia Tungsten Fabric load balancing. Once Octavia Amphora load balancing is enabled, the existing Octavia Tungsten Fabric driver load balancers will continue to function normally. However, you cannot migrate your load balancer workloads from the old LBaaS v2 solution to Amphora.
Note
As long as MOSK provides Octavia Amphora load balancing as a technology preview feature, Mirantis cannot guarantee the stability of this solution and does not provide a migration path from Tungsten Fabric load balancing (HAProxy), which is used by default.
To enable Octavia Amphora load balancing:
Assign
openstack-gateway: enabledlabels to the compute nodes in either order.Caution
Assigning the
openstack-gateway: enabledlabels on compute nodes is crucial for the effective operation of Octavia Amphora load balancing within an OpenStack environment. Double-check the labels assignment to guarantee proper configuration.To make Amphora the default provider, specify it in the
OpenStackDeploymentcustom resource:spec: features: octavia: default_provider: amphorav2
Verify that the OpenStack Controller (Rockoon) has scheduled new Octavia pods that include health manager, worker, and housekeeping pods.
kubectl get pods -n openstack -l 'application=octavia,component in (worker, health_manager, housekeeping)'
Example of output for an environment with two compute nodes:
NAME READY STATUS RESTARTS AGE octavia-health-manager-default-48znl 1/1 Running 0 4h32m octavia-health-manager-default-jk82v 1/1 Running 0 4h34m octavia-housekeeping-7bdf9cbd6c-24vc4 1/1 Running 0 4h34m octavia-housekeeping-7bdf9cbd6c-h9ccv 1/1 Running 0 4h34m octavia-housekeeping-7bdf9cbd6c-rptvv 1/1 Running 0 4h34m octavia-worker-665f84fc7-8kdqd 1/1 Running 0 4h34m octavia-worker-665f84fc7-j6jn9 1/1 Running 0 4h31m octavia-worker-665f84fc7-kqf9t 1/1 Running 0 4h33m
The workflow for creating new load balancers with Amphora is identical to the workflow for creating load balancers with Octavia Tungsten Fabric driver for LBaaS v2. You can do it either through the OpenStack Horizon UI or OpenStack CLI.
If you have not defined amphorav2 as default provider in the
OpenStackDeployment custom resource, you can specify it explicitly
when creating a load balancer using the provider argument:
openstack loadbalancer create --provider amphorav2
Tungsten Fabric known limitations¶
This section contains a summary of the Tungsten Fabric upstream features and use cases not supported in MOSK, features and use cases offered as Technology Preview in the current product release if any, and known limitations of Tungsten Fabric in integration with other product components.
Feature or use case |
Status |
Description |
|---|---|---|
Tungsten Fabric web UI |
Provided as is |
MOSK provides the TF web UI as is and does not include this service in the support Service Level Agreement |
Automatic generation of network port records in DNSaaS (Designate) |
Not supported |
As a workaround, you can use the Tungsten Fabric built-in DNS service that enables virtual machines to resolve each other names |
Secret management (Barbican) |
Not supported |
It is not possible to use the certificates stored in Barbican to terminate HTTPs on a load balancer in a Tungsten Fabric deployment |
Role Based Access Control (RBAC) for Neutron objects |
Not supported |
|
Advanced Tungsten Fabric features |
Provided as is |
MOSK provides the following advanced Tungsten Fabric features as is and does not include them in the support Service Level Agreement:
|
Technical Preview |
DPDK |
|
Tungsten Fabric and OpenStack Octavia Amphora integration |
Technical Preview |
Due to Tungsten Fabric Simple Virtual Gateway restriction, each virtual
network can have only one VGW interface. As a result,
MOSK should be limited to a single compute node
with the |
Tungsten Fabric integration with OpenStack¶
The levels of integration between OpenStack and Tungsten Fabric (TF) include controllers and services integration levels.
Controllers integration¶
The integration between the OpenStack and TF controllers is
implemented through the shared Kubernetes openstack-tf-shared namespace.
Both controllers have access to this namespace to read and write the Kubernetes
kind: Secret objects.
The OpenStack Controller (Rockoon) posts the data into the
openstack-tf-shared namespace required by the TF services. The TF
controller watches this namespace. Once an appropriate secret is created,
the TF controller obtains it into the internal data structures for further
processing.
The OpenStack Controller includes the following data for the TF Controller:
tunnel_intefaceName of the network interface for the TF data plane. This interface is used by TF for the encapsulated traffic for overlay networks.
- Keystone authorization information
Keystone Administrator credentials and an up-and-running IAM service are required for the TF Controller to initiate the deployment process.
- Nova metadata information
Required for the TF vRrouter agent service.
Also, the OpenStack Controller watches the openstack-tf-shared namespace
for the vrouter_port parameter that defines the vRouter port number and
passes it to the nova-compute pod.
Services integration¶
The list of the OpenStack services that are integrated with TF through their API include:
neutron-server- integration is provided by thecontrail-neutron-plugincomponent that is used by theneutron-serverservice for transformation of the API calls to the TF API compatible requests.nova-compute- integration is provided by thecontrail-nova-vif-driverandcontrail-vrouter-apipackages used by thenova-computeservice for interaction with the TF vRouter to the network ports.octavia-api- integration is provided by the Octavia TF Driver that enables you to use OpenStack CLI and Horizon for operations with load balancers. See Tungsten Fabric load balancing (HAProxy) for details.
Warning
TF is not integrated with the following OpenStack services:
DNS service (Designate)
Key management (Barbican)
Tungsten Fabric IPv6 support¶
Tungsten Fabric allows running IPv6-enabled OpenStack tenant networks on top of the IPv4 underlay. You can create an IPv6 virtual network through the Tungsten Fabric web UI or OpenStack CLI in the same way as an IPv4 virtual network. The IPv6 functionality is enabled out of the box and does not require major changes in the cloud configuration. This section lists the IPv6 capabilities supported by MOSK, as well as those available and unavailable in the upstream OpenContrail software.
The following IPv6 features are supported and verified in MOSK:
Virtual machines with IPv6 and IPv4 interfaces
Virtual machines with IPv6-only interfaces
DHCPv6 and neighbor discovery
Policy and security groups
IPv6 flow set up, tear down, and aging
Flow set up and tear down based on a TCP state machine
Fat flow
Allowed address pair configuration with IPv6 addresses
Equal Cost Multi-Path (ECMP)
Additionally, the following IPv6 features are available in upstream OpenContrail according to its official documentation:
Protocol-based flow aging
IPv6 service chaining
Connectivity with gateway (MX Series device)
Virtual Domain Name Services (vDNS), name-to-IPv6 address resolution
The following IPv6 features are not available in upstream OpenContrail:
Any IPv6 Network Address Translation (NAT)
Load Balancing as a Service (LBaaS)
IPv6 fragmentation
Floating IPv6 address
Link-local and metadata services
Diagnostics for IPv6
Contrail Device Manager
Virtual customer premises equipment (vCPE)
Networking¶
Depending on the size of an OpenStack environment and the components that you use, you may want to have a single or multiple network interfaces, as well as run different types of traffic on a single or multiple VLANs.
This section provides the recommendations for planning the network configuration and optimizing the cloud performance.
Networking overview¶
Mirantis OpenStack for Kubernetes (MOSK) cluster networking is complex and defined by the security requirements and performance considerations. It is based on the Kubernetes cluster networking provided by Mirantis Container Cloud and expanded to facilitate the demands of the OpenStack virtualization platform.
A Container Cloud Kubernetes cluster provides a platform for MOSK and is considered a part of its control plane. All networks that serve Kubernetes and related traffic are considered control plane networks. The Kubernetes cluster networking is typically focused on connecting pods of different nodes as well as exposing the Kubernetes API and services running in pods into an external network.
The OpenStack networking connects virtual machines to each other and the outside world. Most of the OpenStack-related networks are considered a part of the data plane in an OpenStack cluster. Ceph networks are considered data plane networks for the purpose of this reference architecture.
When planning your OpenStack environment, consider the types of traffic that your workloads generate and design your network accordingly. If you anticipate that certain types of traffic, such as storage replication, will likely consume a significant amount of network bandwidth, you may want to move that traffic to a dedicated network interface to avoid performance degradation.
The following diagram provides a simplified overview of the underlay networking in a MOSK environment:
Management cluster networking¶
This page summarizes the recommended networking architecture of a Mirantis Container Cloud management cluster for a Mirantis OpenStack for Kubernetes (MOSK) cluster.
We recommend deploying the management cluster with a dedicated interface for the provisioning (PXE) network. The separation of the provisioning network from the management network ensures additional security and resilience of the solution.
MOSK end users typically should have access to the Keycloak service in the management cluster for authentication to the Horizon web UI. Therefore, we recommend that you connect the management network of the management cluster to an external network through an IP router. The default route on the management cluster nodes must be configured with the default gateway in the management network.
If you deploy the multi-rack configuration, ensure that the provisioning network of the management cluster is connected to an IP router that connects it to the provisioning networks of all racks.
MOSK cluster networking¶
Mirantis OpenStack for Kubernetes (MOSK) clusters managed by Mirantis Container Cloud use the following networks to serve different types of traffic:
Network role |
Description |
|---|---|
Provisioning (PXE) network |
Facilitates the iPXE boot of all bare metal machines in a MOSK cluster and provisioning of the operating system to machines. This network is only used during provisioning of the host. It must not be configured on an operational MOSK node. |
Life-cycle management (LCM) network |
Connects LCM Agents running on the hosts to the Container Cloud LCM API.
The LCM API is provided by the management cluster.
The LCM network is also used for communication between The LCM subnet(s) provides IP addresses that are statically allocated by the IPAM service to bare metal hosts. This network must be connected to the Kubernetes API endpoint of the management cluster through an IP router. LCM Agents running on MOSK clusters will connect to the management cluster API through this router. LCM subnets may be different per MOSK cluster as long as this connection requirement is satisfied. You can use more than one LCM network segment in a MOSK cluster. In this case, separated L2 segments and interconnected L3 subnets are still used to serve LCM and API traffic. All IP subnets in the LCM networks must be connected to each other by IP routes. These routes must be configured on the hosts through L2 templates. All IP subnets in the LCM network must be connected to the Kubernetes API endpoints of the management cluster through an IP router. You can manually select the load balancer IP address for external
access to the cluster API and specify it in the Note When using the ARP announcement of the IP address for the cluster API load balancer, the following limitations apply:
When using the BGP announcement of the IP address for the cluster API load balancer, which is available as Technology Preview since MOSK 23.2.2, no segment stretching is required between Kubernetes master nodes. Also, in this scenario, the load balancer IP address is not required to match the LCM subnet CIDR address. |
Kubernetes workloads network |
Serves as an underlay network for traffic between pods in the MOSK cluster. Do not share this network between clusters. There might be more than one Kubernetes pods network segment in the cluster. In this case, they must be connected through an IP router. Kubernetes workloads network does not need an external access. The Kubernetes workloads subnet(s) provides IP addresses that are statically allocated by the IPAM service to all nodes and that are used by Calico for cross-node communication inside a cluster. By default, VXLAN overlay is used for Calico cross-node communication. |
Kubernetes external network |
Serves for an access to the OpenStack endpoints in a MOSK cluster. When using the ARP announcement of the external endpoints of load-balanced services, the network must contain a VLAN segment extended to all MOSK nodes connected to this network. When using the BGP announcement of the external endpoints of load-balanced services, which is available as Technology Preview since MOSK 23.2.2, there is no requirement of having a single VLAN segment extended to all MOSK nodes connected to this network. A typical MOSK cluster only has one external network. The external network must include at least two IP address ranges
defined by separate
|
Storage access network |
Serves for the storage access traffic from and to Ceph OSD services. A MOSK cluster may have more than one VLAN segment and IP subnet in the storage access network. All IP subnets of this network in a single cluster must be connected by an IP router. The storage access network does not require external access unless you want to directly expose Ceph to the clients outside of a MOSK cluster. Note A direct access to Ceph by the clients outside of a MOSK cluster is technically possible but not supported by Mirantis. Use at your own risk. The IP addresses from subnets in this network are statically allocated by the IPAM service to Ceph nodes. The Ceph OSD services bind to these addresses on their respective nodes. This is a public network in Ceph terms. 1 |
Storage replication network |
Serves for the storage replication traffic between Ceph OSD services. A MOSK cluster may have more than one VLAN segment and IP subnet in this network as long as the subnets are connected by an IP router. This network does not require external access. The IP addresses from subnets in this network are statically allocated by the IPAM service to Ceph nodes. The Ceph OSD services bind to these addresses on their respective nodes. This is a cluster network in Ceph terms. 1 |
Out-of-Band (OOB) network |
Connects Baseboard Management Controllers (BMCs) of the bare metal hosts. Must not be accessible from a MOSK cluster. |
- 1(1,2)
For more details about Ceph networks, see Ceph Network Configuration Reference.
The following diagram illustrates the networking schema of the Container Cloud deployment on bare metal with a MOSK cluster using ARP announcements:
Since 23.2.2, MOSK supports full L3 networking topology in the Technology Preview scope. The following diagram illustrates the networking schema of the Container Cloud deployment on bare metal with a MOSK cluster using BGP announcements:
Network types¶
This section describes network types for Layer 3 networks used for Kubernetes and Mirantis OpenStack for Kubernetes (MOSK) clusters along with requirements for each network type.
Note
Only IPv4 is currently supported by Container Cloud and IPAM for infrastructure networks. Both IPv4 and IPv6 are supported for OpenStack workloads.
The following diagram provides an overview of the underlay networks in a MOSK environment:
L3 networks for Kubernetes¶
A MOSK deployment typically requires the following types of networks:
- Provisioning network
Used for provisioning of bare metal servers.
- Management network
Used for management of the Container Cloud infrastructure and for communication between containers in Kubernetes.
- LCM/API network
Since 23.2.2, MOSK supports full L3 networking topology in the Technology Preview scope. This enables deployment of specific cluster segments in dedicated racks without the need for L2 layer extension between them. For configuration procedure, see Configure BGP announcement for cluster API LB address and Configure BGP announcement of external addresses of Kubernetes load-balanced services in Deployment Guide.
If BGP announcement is configured for the MOSK cluster API LB address, the API/LCM network is not required. Announcement of the cluster API LB address is done using the LCM network.
If you configure ARP announcement of the load-balancer IP address for the MOSK cluster API, the API/LCM network must be configured on the Kubernetes manager nodes of the cluster. This network contains the Kubernetes API endpoint with the VRRP virtual IP address.
- LCM network
Enables communication between the MKE cluster nodes. Multiple VLAN segments and IP subnets can be created for a multi-rack architecture. Each server must be connected to one of the LCM segments and have an IP from the corresponding subnet.
- External network
Used to expose the OpenStack, StackLight, and other services of the MOSK cluster.
- Kubernetes workloads network
Used for communication between containers in Kubernetes.
- Storage access network (Ceph)
Used for accessing the Ceph storage. In Ceph terms, this is a public network 0. We recommended that it is placed on a dedicated hardware interface.
- Storage replication network (Ceph)
Used for Ceph storage replication. In Ceph terms, this is a cluster network 0. To ensure low latency and fast access, place the network on a dedicated hardware interface.
- 0(1,2)
For details about Ceph networks, see Ceph Network Configuration Reference.
L3 networks for MOSK¶
The MOSK deployment additionally requires the following networks.
Service name |
Network |
Description |
VLAN name |
|---|---|---|---|
Networking |
Provider networks |
Typically, a routable network used to provide the external access to OpenStack instances (a floating network). Can be used by the OpenStack services such as Ironic, Manila, and others, to connect their management resources. |
|
Networking |
Overlay networks (virtual networks) |
The network used to provide denied, secure tenant networks with the help of the tunneling mechanism (VLAN/GRE/VXLAN). If the VXLAN and GRE encapsulation takes place, the IP address assignment is required on interfaces at the node level. |
|
Compute |
Live migration network |
The network used by the OpenStack compute service (Nova) to transfer data during live migration. Depending on the cloud needs, it can be placed on a dedicated physical network not to affect other networks during live migration. The IP address assignment is required on interfaces at the node level. |
|
The way of mapping of the logical networks described above to physical networks and interfaces on nodes depends on the cloud size and configuration. We recommend placing OpenStack networks on a dedicated physical interface (bond) that is not shared with storage and Kubernetes management network to minimize the influence on each other.
L3 networks requirements¶
The following tables describe networking requirements for a MOSK cluster, Container Cloud management and Ceph clusters.
Network type |
Provisioning |
Management |
|---|---|---|
Suggested interface name |
N/A |
|
Minimum number of VLANs |
1 |
1 |
Minimum number of IP subnets |
3 |
2 |
Minimum recommended IP subnet size |
|
|
External routing |
Not required |
Required, may use proxy server |
Multiple segments/stretch segment |
Stretch segment for management cluster due to MetalLB Layer 2 limitations 1 |
Stretch segment due to VRRP, MetalLB Layer 2 limitations |
Internal routing |
Routing to separate DHCP segments, if in use |
|
- 1
Multiple VLAN segments with IP subnets can be added to the cluster configuration for separate DHCP domains.
Since 23.2.2, MOSK supports full L3 networking topology in the Technology Preview scope. This enables deployment of specific cluster segments in dedicated racks without the need for L2 layer extension between them. For configuration procedure, see Configure BGP announcement for cluster API LB address and Configure BGP announcement of external addresses of Kubernetes load-balanced services in Deployment Guide.
If you configure BGP announcement of the load-balancer IP address for a MOSK cluster API and for load-balanced services of the cluster:
Network type |
Provisioning |
LCM |
External |
Kubernetes workloads |
|---|---|---|---|---|
Minimum number of VLANs |
1 (optional) |
1 |
1 |
1 |
Suggested interface name |
N/A |
|
|
|
Minimum number of IP subnets |
1 (optional) |
1 |
2 |
1 |
Minimum recommended IP subnet size |
16 IPs (DHCP range) |
|
|
1 IP per cluster node |
Stretch or multiple segments |
Multiple |
Multiple |
Multiple For details, see Configure node selectors for MetalLB speakers. |
Multiple |
External routing |
Not required |
Not required |
Required, default route |
Not required |
Internal routing |
Routing to the provisioning network of the management cluster |
|
Routing to the IP subnet of the Container Cloud Management API |
Routing to all IP subnets of Kubernetes workloads |
If you configure ARP announcement of the load-balancer IP address for a MOSK cluster API and for load-balanced services of the cluster:
Network type |
Provisioning |
LCM/API |
LCM |
External |
Kubernetes workloads |
|---|---|---|---|---|---|
Minimum number of VLANs |
1 (optional) |
1 |
1 (optional) |
1 |
1 |
Suggested interface name |
N/A |
|
|
|
|
Minimum number of IP subnets |
1 (optional) |
1 |
1 (optional) |
2 |
1 |
Minimum recommended IP subnet size |
16 IPs (DHCP range) |
|
1 IP per MOSK node (Kubernetes worker) |
|
1 IP per cluster node |
Stretch or multiple segments |
Multiple |
Stretch due to VRRP limitations |
Multiple |
Stretch connected to all MOSK controller nodes. For details, see Configure node selectors for MetalLB speakers. |
Multiple |
External routing |
Not required |
Not required |
Not required |
Required, default route |
Not required |
Internal routing |
Routing to the provisioning network of the management cluster |
|
|
Routing to the IP subnet of the Container Cloud Management API |
Routing to all IP subnets of Kubernetes workloads |
- 2(1,2)
The bridge interface with this name is mandatory if you need to separate Kubernetes workloads traffic. You can configure this bridge over the VLAN or directly over the bonded or single interface.
Network type |
Storage access |
Storage replication |
|---|---|---|
Minimum number of VLANs |
1 |
1 |
Suggested interface name |
|
|
Minimum number of IP subnets |
1 |
1 |
Minimum recommended IP subnet size |
1 IP per cluster node |
1 IP per cluster node |
Stretch or multiple segments |
Multiple |
Multiple |
External routing |
Not required |
Not required |
Internal routing |
Routing to all IP subnets of the Storage access network |
Routing to all IP subnets of the Storage replication network |
Note
When selecting externally routable subnets, ensure that the subnet ranges do not overlap with the internal subnets ranges. Otherwise, internal resources of users will not be available from the MOSK cluster.
- 3(1,2)
For details about Ceph networks, see Ceph Network Configuration Reference.
Multi-rack architecture¶
TechPreview
Mirantis OpenStack for Kubernetes (MOSK) enables you to deploy a cluster with a multi-rack architecture, where every data center cabinet (a rack) incorporates its own Layer 2 network infrastructure that does not extend beyond its top-of-rack switch. The architecture allows a MOSK cloud to integrate natively with the Layer 3-centric networking topologies such as Spine-Leaf that are commonly seen in modern data centers.
The architecture eliminates the need to stretch and manage VLANs across parts of a single data center, or to build VPN tunnels between the segments of a geographically distributed cloud.
The set of networks present in each rack depends on the backend used by the OpenStack networking service.
Bare metal provisioning network¶
In the Mirantis Container Cloud and MOSK multi-rack reference architecture, every rack has its own L2 segment (VLAN) to bootstrap and install servers.
Segmentation of the provisioning network requires additional configuration of the underlay networking infrastructure and certain Container Cloud API objects. You need to configure a DHCP Relay agent on the border of each VLAN in the provisioning network. The agent handles broadcast DHCP requests coming from the bare metal servers in the rack and forwards them as unicast packets across L3 fabric of the data center to a Container Cloud management cluster.
From the standpoint of Container Cloud API, you need to configure per-rack DHCP
ranges by adding Subnet resources in Container Cloud as described in
Configure multiple DHCP address ranges.
The DHCP server of Container Cloud automatically leases a temporary IP address from the DHCP range to the requester host depending on the address of the DHCP agent that relays the request.
Multi-rack MOSK cluster¶
To deploy a MOSK cluster with multi-rack reference architecture, you need to create a dedicated set of subnets and L2 templates for every rack in your cluster.
Every specific host type in the rack, which is defined by the role in the MOSK cluster and network-related hardware configuration, may require a specific L2 template.
Note
Since 23.2.2, MOSK supports full L3 networking topology in the Technology Preview scope. This enables deployment of specific cluster segments in dedicated racks without the need for L2 layer extension between them. For configuration procedure, see Configure BGP announcement for cluster API LB address and Configure BGP announcement of external addresses of Kubernetes load-balanced services in Deployment Guide.
For MOSK 23.1 and older versions, due to the Container Cloud limitations, you need to configure the following networks to have L2 segments (VLANs) stretch across racks to all hosts of certain types in a multi-rack environment:
- LCM/API network
Must be configured on the Kubernetes manager nodes of the MOSK cluster. Contains a Kubernetes API endpoint with a VRRP virtual IP address. Enables MKE cluster nodes to communicate with each other.
- External network
Exposes OpenStack, StackLight, and other services of the MOSK cluster to external clients.
For details, see Underlay networking: routing configuration.
When planning space allocation for IP addresses in your cluster, pick large IP ranges for each type of network. Then you will split these ranges into per-rack subnets.
For example, if you allocate a /20 address block for LCM network,
then you can create up to 16 Subnet objects with the /24 address block
each for up to 16 racks. This way you can simplify routing on your hosts using
the large /20 IP subnet as an aggregated route destination. For details,
see Underlay networking: routing configuration.
Multi-rack MOSK cluster with Tungsten Fabric¶
A typical medium and more sized MOSK cloud consists of three or more racks that can generally be divided into the following major categories:
Compute/Storage racks that contain the hypervisors and instances running on top of them. Additionally, they contain nodes that store cloud applications’ block, ephemeral, and object data as part of the Ceph cluster.
Control plane racks that incorporate all the components needed by the cloud operator to manage its life cycle. Also, they include the services through which the cloud users interact with the cloud to deploy their applications, such as cloud APIs and web UI.
A control plane rack may also contain additional compute and storage nodes.
The diagram below will help you to plan the networking layout of a multi-rack MOSK cloud with Tungsten Fabric.
Note
Since 23.2.2, MOSK supports full L3 networking topology in the Technology Preview scope. This enables deployment of specific cluster segments in dedicated racks without the need for L2 layer extension between them. For configuration procedure, see Configure BGP announcement for cluster API LB address and Configure BGP announcement of external addresses of Kubernetes load-balanced services in Deployment Guide.
For MOSK 23.1 and older versions, Kubernetes masters (3 nodes) either need to be placed into a single rack or, if distributed across multiple racks for better availability, require stretching of the L2 segment of the management network across these racks. This requirement is caused by the Mirantis Kubernetes Engine underlay for MOSK relying on the Layer 2 VRRP protocol to ensure high availability of the Kubernetes API endpoint.
The table below provides a mapping between the racks and the network types participating in a multi-rack MOSK cluster with the Tungsten Fabric backend.
Network |
VLAN name |
Control Plane rack |
Compute/Storage rack |
|---|---|---|---|
Common/PXE |
|
Yes |
Yes |
Management |
|
Yes |
Yes |
External (MetalLB) |
|
Yes |
No |
Kubernetes workloads |
|
Yes |
Yes |
Storage access (Ceph) |
|
Yes |
Yes |
Storage replication (Ceph) |
|
Yes |
Yes |
Overlay |
|
Yes |
Yes |
Live migration |
|
Yes |
Yes |
Physical networks layout¶
This section summarizes the requirements for the physical layout of underlay network and VLANs configuration for the multi-rack architecture of Mirantis OpenStack for Kubernetes (MOSK).
Physical networking of a Container Cloud management cluster¶
Due to limitations of virtual IP address for Kubernetes API and of MetalLB load balancing in Container Cloud, the management cluster nodes must share VLAN segments in the provisioning and management networks.
In the multi-rack architecture, the management cluster nodes may be placed to a single rack or spread across three racks. In either case, provisioning and management network VLANs must be stretched across ToR switches of the racks.
The following diagram illustrates physical and L2 connections of the Container Cloud management cluster.
Physical networking of a MOSK cluster¶
Since 23.2.2, MOSK supports full L3 networking topology in the Technology Preview scope. This enables deployment of specific cluster segments in dedicated racks without the need for L2 layer extension between them. For configuration procedure, see Configure BGP announcement for cluster API LB address and Configure BGP announcement of external addresses of Kubernetes load-balanced services in Deployment Guide.
If you configure BGP announcement for IP addresses of load-balanced services of a MOSK cluster, the external network can consist of multiple VLAN segments connected to all nodes of a MOSK cluster where MetalLB speaker components are configured to announce IP addresses for Kubernetes load-balanced services. Mirantis recommends that you use OpenStack controller nodes for this purpose.
If you configure ARP announcement for IP addresses of load-balanced services of a MOSK cluster, the external network must consist of a single VLAN stretched to the ToR switches of all the racks where MOSK nodes connected to the external network are located. Those are the nodes where MetalLB speaker components are configured to announce IP addresses for Kubernetes load-balanced services. Mirantis recommends that you use OpenStack controller nodes for this purpose.
Note
Since 23.2.2, MOSK supports full L3 networking topology in the Technology Preview scope. This enables deployment of specific cluster segments in dedicated racks without the need for L2 layer extension between them. For configuration procedure, see Configure BGP announcement for cluster API LB address and Configure BGP announcement of external addresses of Kubernetes load-balanced services in Deployment Guide.
If BGP announcement is configured for MOSK cluster API LB address, Kubernetes manager nodes have no requirement to share the single stretched VLAN segment in the API/LCM network. All VLANs may be configured per rack.
If ARP announcement is configured for MOSK cluster API LB address, Kubernetes manager nodes must share the VLAN segment in the API/LCM network. In the multi-rack architecture, Kubernetes manager nodes may be spread across three racks. The API/LCM network VLAN must be stretched to the ToR switches of the racks. All other VLANs may be configured per rack. This requirement is caused by the Mirantis Kubernetes Engine underlay for MOSK relying on the Layer 2 VRRP protocol to ensure high availability of the Kubernetes API endpoint.
The following diagram illustrates physical and L2 network connections of the Kubernetes manager nodes in a MOSK cluster.
Caution
Such configuration does not apply to a compact control plane MOSK installation. See Create a MOSK cluster.
The following diagram illustrates physical and L2 network connections of the control plane nodes in a MOSK cluster.
All VLANs for OpenStack compute nodes may be configured per rack. No VLAN should be stretched across multiple racks.
The following diagram illustrates physical and L2 network connections of the compute nodes in a MOSK cluster.
All VLANs for OpenStack storage nodes may be configured per rack. No VLAN should be stretched across multiple racks.
The following diagram illustrates physical and L2 network connections of the storage nodes in a MOSK cluster.
Underlay networking: routing configuration¶
This section describes requirements for the configuration of the underlay network for an MOSK cluster in a multi-rack reference configuration. The infrastructure operator must configure the underlay network according to these guidelines. Mirantis Container Cloud will not configure routing on the network devices.
Provisioning network¶
In the multi-rack reference architecture, every server rack has its own layer-2 segment (VLAN) for network bootstrap and installation of physical servers.
You need to configure top-of-rack (ToR) switches in each rack with the default gateway for the provisioning network VLAN. This gateway must also serve as a DHCP Relay Agent on the border of the VLAN. The agent handles broadcast DHCP requests coming from the bare metal servers in the rack and forwards them as unicast packets across the data center L3 fabric to the provisioning network of a Container Cloud management cluster.
Therefore, each ToR gateway must have an IP route to the IP subnet of the provisioning network of the management cluster. The provisioning network gateway, in turn, must have routes to all IP subnets of all racks.
The hosts of the management cluster must have routes to all IP subnets in the provisioning network through the gateway in the provisioning network of the management cluster.
All hosts in the management cluster must have IP addresses from the same IP subnet of the provisioning network. Even if the hosts of the management cluster are mounted to different racks, they must share a single provisioning VLAN segment.
Management network¶
All hosts of a management cluster must have IP addresses from the same subnet of the management network. Even if hosts of a management cluster are mounted to different racks, they must share a single management VLAN segment.
The gateway in this network is used as the default route on the nodes in a Container Cloud management cluster. This gateway must connect to external Internet networks directly or through a proxy server. If the Internet is accessible through a proxy server, you must configure Container Cloud bootstrap to use it as well. For details, see Deploy a management cluster.
This network connects a Container Cloud management cluster to Kubernetes API endpoints of MOSK clusters. It also connects LCM agents of MOSK nodes to the Kubernetes API endpoint of the management cluster.
The network gateway must have routes to all API/LCM subnets of all MOSK clusters.
LCM network¶
This network may include multiple VLANs, typically, one VLAN per rack. Each VLAN may have one or more IP subnets with gateways configured on ToR switches.
Each ToR gateway must provide routes to all other IP subnets in all other VLANs in the LCM network to enable communication between nodes in the cluster.
Note
Since 23.2.2, MOSK supports full L3 networking topology in the Technology Preview scope. This enables deployment of specific cluster segments in dedicated racks without the need for L2 layer extension between them. For configuration procedure, see Configure BGP announcement for cluster API LB address and Configure BGP announcement of external addresses of Kubernetes load-balanced services in Deployment Guide.
If you configure BGP announcement of the load-balancer IP address for a MOSK cluster API:
All nodes of a MOSK cluster must be connected to the LCM network. Each host connected to this network must have routes to all IP subnets in the LCM network and to the management subnet of the management cluster, through the ToR gateway for the rack of this host.
It is not required to configure a separate API/LCM network. Announcement of the IP address of the load balancer is done using the LCM network.
If you configure ARP announcement of the load-balancer IP address for a MOSK cluster API:
All nodes of a MOSK cluster excluding manager nodes must be connected to the LCM network. Each host connected to this network must have routes to all IP subnets in the LCM network, including the API/LCM network of this MOSK cluster and to the Management subnet of the management cluster, through the ToR gateway for the rack of this host.
It is required to configure a separate API/LCM network. All manager nodes of a MOSK cluster must be connected to the API/LCM network. IP address announcement for load balancing is done using the API/LCM network.
API/LCM network¶
Note
Since 23.2.2, MOSK supports full L3 networking topology in the Technology Preview scope. This enables deployment of specific cluster segments in dedicated racks without the need for L2 layer extension between them. For configuration procedure, see Configure BGP announcement for cluster API LB address and Configure BGP announcement of external addresses of Kubernetes load-balanced services in Deployment Guide.
If BGP announcement is configured for the MOSK cluster API LB address, the API/LCM network is not required. Announcement of the cluster API LB address is done using the LCM network.
If you configure ARP announcement of the load-balancer IP address for the MOSK cluster API, the API/LCM network must be configured on the Kubernetes manager nodes of the cluster. This network contains the Kubernetes API endpoint with the VRRP virtual IP address.
This network consists of a single VLAN shared between all MOSK manager nodes in a MOSK cluster, even if the nodes are spread across multiple racks. All manager nodes of a MOSK cluster must be connected to this network and have IP addresses from the same subnet in this network.
The gateway in the API/LCM network for a MOSK cluster must have a route to the Management subnet of the management cluster. This is required to ensure symmetric traffic flow between the management and MOSK clusters.
The gateway in this network must also have routes to all IP subnets in the LCM network of this MOSK cluster.
The load-balancer IP address for cluster API must be allocated from the same CIDR address that the API/LCM subnet uses.
External network¶
Note
Since 23.2.2, MOSK supports full L3 networking topology in the Technology Preview scope. This enables deployment of specific cluster segments in dedicated racks without the need for L2 layer extension between them. For configuration procedure, see Configure BGP announcement for cluster API LB address and Configure BGP announcement of external addresses of Kubernetes load-balanced services in Deployment Guide.
If you configure BGP announcement for IP addresses of load-balanced services of a MOSK cluster, the external network can consist of multiple VLAN segments connected to all nodes of a MOSK cluster where MetalLB speaker components are configured to announce IP addresses for Kubernetes load-balanced services. Mirantis recommends that you use OpenStack controller nodes for this purpose.
If you configure ARP announcement for IP addresses of load-balanced services of a MOSK cluster, the external network must consist of a single VLAN stretched to the ToR switches of all the racks where MOSK nodes connected to the external network are located. Those are the nodes where MetalLB speaker components are configured to announce IP addresses for Kubernetes load-balanced services. Mirantis recommends that you use OpenStack controller nodes for this purpose.
The IP gateway in this network is used as the default route on all nodes in the MOSK cluster, which are connected to this network. This allows external users to connect to the OpenStack endpoints exposed as Kubernetes load-balanced services.
Dedicated IP ranges from this network must be configured as address pools for the MetalLB service. MetalLB allocates addresses from these address pools to Kubernetes load-balanced services.
Ceph public network¶
This network may include multiple VLANs and IP subnets, typically, one VLAN and IP subnet per rack. All IP subnets in this network must be connected by IP routes on the ToR switches.
Typically, every node in a MOSK cluster is connected to this network and have routes to all IP subnets from this network through its rack IP gateway.
This network is not connected to the external networks.
Ceph cluster network¶
This network may include multiple VLANs and IP subnets, typically, one VLAN and IP subnet per rack. All IP subnets in this network must be connected by IP routes on the ToR switches.
Every Ceph OSD node in a MOSK cluster must be connected to this network and have routes to all IP subnets from this network through its rack IP gateway.
This network is not connected to the external networks.
Kubernetes workloads network¶
This network may include multiple VLANs and IP subnets, typically, one VLAN and IP subnet per rack. All IP subnets in this network must be connected by IP routes on the ToR switches.
All nodes in a MOSK cluster must be connected to this network and have routes to all IP subnets from this network through its rack IP gateway.
This network is not connected to the external networks.
Performance optimization¶
The following recommendations apply to all types of nodes in the Mirantis OpenStack for Kubernetes (MOSK) clusters.
Jumbo frames¶
To improve the goodput, we recommend that you enable jumbo frames where possible. The jumbo frames have to be enabled on the whole path of the packets traverse. If one of the network components cannot handle jumbo frames, the network path uses the smallest MTU.
Bonding¶
To provide fault tolerance of a single NIC, we recommend using the link aggregation, such as bonding. The link aggregation is useful for linear scaling of bandwidth, load balancing, and fault protection. Depending on the hardware equipment, different types of bonds might be supported. Use the multi-chassis link aggregation as it provides fault tolerance at the device level. For example, MLAG on Arista equipment or vPC on Cisco equipment.
The Linux kernel supports the following bonding modes:
active-backupbalance-xor802.3ad(LACP)balance-tlbbalance-alb
Since LACP is the IEEE standard 802.3ad supported by the majority of
network platforms, we recommend using this bonding mode.
Use the Link Aggregation Control Protocol (LACP) bonding mode
with MC-LAG domains configured on ToR switches. This corresponds to
the 802.3ad bond mode on hosts.
Additionally, follow these recommendations in regards to bond interfaces:
Use ports from different multi-port NICs when creating bonds. This makes network connections redundant if failure of a single NIC occurs.
Configure the ports that connect servers to the PXE network with PXE VLAN as native or untagged. On these ports, configure LACP fallback to ensure that the servers can reach DHCP server and boot over network.
Spanning tree portfast mode¶
Configure Spanning Tree Protocol (STP) settings on the network switch ports to ensure that the ports start forwarding packets as soon as the link comes up. It helps avoid iPXE timeout issues and ensures reliable boot over network.
Storage¶
A MOSK cluster uses Ceph as a distributed storage system for file, block, and object storage. This section provides an overview of a Ceph cluster deployed by Container Cloud.
Ceph overview¶
Mirantis Container Cloud deploys Ceph on MOSK using Helm charts with the following components:
- Rook Ceph Operator
A storage orchestrator that deploys Ceph on top of a Kubernetes cluster. Also known as
RookorRook Operator. Rook operations include:Deploying and managing a Ceph cluster based on provided Rook CRs such as
CephCluster,CephBlockPool,CephObjectStore, and so on.Orchestrating the state of the Ceph cluster and all its daemons.
KaaSCephClustercustom resource (CR)Represents the customization of a Kubernetes installation and allows you to define the required Ceph configuration through the Container Cloud web UI before deployment. For example, you can define the failure domain, Ceph pools, Ceph node roles, number of Ceph components such as Ceph OSDs, and so on. The
ceph-kcc-controllercontroller on the Container Cloud management cluster manages theKaaSCephClusterCR.- Ceph Controller
A Kubernetes controller that obtains the parameters from Container Cloud through a CR, creates CRs for Rook and updates its CR status based on the Ceph cluster deployment progress. It creates users, pools, and keys for OpenStack and Kubernetes and provides Ceph configurations and keys to access them. Also, Ceph Controller eventually obtains the data from the OpenStack Controller (Rockoon) for the Keystone integration and updates the Ceph Object Gateway services configurations to use Kubernetes for user authentication.
The Ceph Controller operations include:
Transforming user parameters from the Container Cloud Ceph CR into Rook CRs and deploying a Ceph cluster using Rook.
Providing integration of the Ceph cluster with Kubernetes.
Providing data for OpenStack to integrate with the deployed Ceph cluster.
- Ceph Status Controller
A Kubernetes controller that collects all valuable parameters from the current Ceph cluster, its daemons, and entities and exposes them into the
KaaSCephClusterstatus. Ceph Status Controller operations include:Collecting all statuses from a Ceph cluster and corresponding Rook CRs.
Collecting additional information on the health of Ceph daemons.
Provides information to the
statussection of theKaaSCephClusterCR.
- Ceph Request Controller
A Kubernetes controller that obtains the parameters from Container Cloud through a CR and manages Ceph OSD lifecycle management (LCM) operations. It allows for a safe Ceph OSD removal from the Ceph cluster. Ceph Request Controller operations include:
Providing an ability to perform Ceph OSD LCM operations.
Obtaining specific CRs to remove Ceph OSDs and executing them.
Pausing the regular Ceph Controller reconciliation until all requests are completed.
A typical Ceph cluster consists of the following components:
Ceph Monitors - three or, in rare cases, five Ceph Monitors.
Ceph Managers - one Ceph Manager in a regular cluster.
Ceph Object Gateway (
radosgw) - Mirantis recommends having three or moreradosgwinstances for HA.Ceph OSDs - the number of Ceph OSDs may vary according to the deployment needs.
Warning
A Ceph cluster with 3 Ceph nodes does not provide hardware fault tolerance and is not eligible for recovery operations, such as a disk or an entire Ceph node replacement.
A Ceph cluster uses the replication factor that equals 3. If the number of Ceph OSDs is less than 3, a Ceph cluster moves to the degraded state with the write operations restriction until the number of alive Ceph OSDs equals the replication factor again.
The placement of Ceph Monitors and Ceph Managers is defined in the
KaaSCephCluster CR.
The following diagram illustrates the way a Ceph cluster is deployed in Container Cloud:
The following diagram illustrates the processes within a deployed Ceph cluster:
See also
Ceph limitations¶
A Ceph cluster configuration in MOSK includes but is not limited to the following limitations:
Only one Ceph Controller per MOSK cluster and only one Ceph cluster per Ceph Controller are supported.
The replication size for any Ceph pool must be set to more than 1.
Only one CRUSH tree per cluster. The separation of devices per Ceph pool is supported through device classes with only one pool of each type for a device class.
All CRUSH rules must have the same
failure_domain.Only the following types of CRUSH buckets are supported:
topology.kubernetes.io/regiontopology.kubernetes.io/zonetopology.rook.io/datacentertopology.rook.io/roomtopology.rook.io/podtopology.rook.io/pdutopology.rook.io/rowtopology.rook.io/racktopology.rook.io/chassis
RBD mirroring is not supported.
Consuming an existing Ceph cluster is not supported.
Lifted since MOSK 23.1 CephFS is unsupported. Multiple CephFS are supported since MOSK 25.1.
Only IPv4 is supported.
If two or more Ceph OSDs are located on the same device, there must be no dedicated WAL or DB for this class.
Only a full collocation or dedicated WAL and DB configurations are supported.
The minimum size of any defined Ceph OSD device is 5 GB.
Ceph OSDs support only raw disks as data devices meaning that no
dmorlvmdevices are allowed.Lifted since MOSK 23.3 Ceph cluster does not support removable devices (with hotplug enabled) for deploying Ceph OSDs.
When adding a Ceph node with the Ceph Monitor role, if any issues occur with the Ceph Monitor,
rook-cephremoves it and adds a new Ceph Monitor instead, named using the next alphabetic character in order. Therefore, the Ceph Monitor names may not follow the alphabetical order. For example,a,b,d, instead ofa,b,c.Reducing the number of Ceph Monitors is not supported and causes the Ceph Monitor daemons removal from random nodes.
Removal of the
mgrrole in thenodessection of theKaaSCephClusterCR does not remove Ceph Managers. To remove a Ceph Manager from a node, remove it from thenodesspec and manually delete themgrpod in the Rook namespace.Lifted since MOSK 24.1 Ceph does not support allocation of Ceph RGW pods on nodes where the Federal Information Processing Standard (FIPS) mode is enabled.
Ceph integration with OpenStack¶
The integration between Ceph and OpenStack (Rockoon) Controllers is implemented
through the shared Kubernetes openstack-ceph-shared namespace. Both
controllers have access to this namespace to read and write the Kubernetes
kind: Secret objects.
As Ceph is required and only supported backend for several OpenStack
services, all necessary Ceph pools must be specified in the configuration
of the kind: MiraCeph custom resource as part of the deployment.
Once the Ceph cluster is deployed, the Ceph Controller posts the
information required by the OpenStack services to be properly configured
as a kind: Secret object into the openstack-ceph-shared namespace.
The OpenStack Controller watches this namespace. Once the corresponding
secret is created, the OpenStack Controller transforms this secret to the
data structures expected by the OpenStack-Helm charts. Even if an OpenStack
installation is triggered at the same time as a Ceph cluster deployment, the
OpenStack Controller halts the deployment of the OpenStack services that
depend on Ceph availability until the secret in the shared namespace is
created by the Ceph Controller.
For the configuration of Ceph Object Gateway as an OpenStack Object
Storage, the reverse process takes place. The OpenStack Controller waits
for the OpenStack-Helm to create a secret with OpenStack Identity
(Keystone) credentials that Ceph Object Gateway must use to validate the
OpenStack Identity tokens, and posts it back to the same
openstack-ceph-shared namespace in the format suitable for
consumption by the Ceph Controller. The Ceph Controller then reads this
secret and reconfigures Ceph Object Gateway accordingly.
Mirantis StackLight¶
StackLight is the logging, monitoring, and alerting solution that provides a single pane of glass for cloud maintenance and day-to-day operations as well as offers critical insights into cloud health including operational information about the components deployed with Mirantis OpenStack for Kubernetes (MOSK). StackLight is based on Prometheus, an open-source monitoring solution and a time series database, and OpenSearch, the logs and notifications storage.
Deployment architecture¶
Mirantis OpenStack for Kubernetes (MOSK) deploys the StackLight stack as a release of a Helm chart that contains the helm-controller and HelmBundle custom resources. The StackLight HelmBundle consists of a set of Helm charts describing the StackLight components. Apart from the OpenStack-specific components below, StackLight also includes the components described in Mirantis Container Cloud Reference Architecture: Deployment architecture. By default, StackLight logging stack is disabled.
During the StackLight configuration when deploying a MOSK cluster, you can define the HA or non-HA StackLight architecture type. Non-HA StackLight requires a backend storage provider, for example, a Ceph cluster. For details, see Mirantis Container Cloud Reference Architecture: StackLight database modes.
StackLight component |
Description |
|---|---|
Prometheus native exporters and endpoints |
Export the existing metrics as Prometheus metrics and include:
|
Telegraf OpenStack plugin |
Collects and processes the OpenStack metrics. |
Monitored components¶
StackLight measures, analyzes, and reports in a timely manner about failures that may occur in the following Mirantis OpenStack for Kubernetes (MOSK) components and their sub-components. Apart from the components below, StackLight also monitors the components listed in Mirantis Container Cloud Reference Architecture: Monitored components.
Libvirt
Memcached
MariaDB
NTP
OpenStack (Barbican, Cinder, Designate, Glance, Heat, Horizon, Ironic, Keystone, Neutron, Nova, Octavia)
OpenStack SSL certificates
Tungsten Fabric (Casandra, Kafka, Redis, ZooKeeper)
RabbitMQ
OpenSearch and Prometheus storage sizing¶
Caution
Calculations in this document are based on numbers from a real-scale test cluster with 34 nodes. The exact space required for metrics and logs must be calculated depending on the ongoing cluster operations. Some operations force the generation of additional metrics and logs. The values below are approximate. Use them only as recommendations.
During the deployment of a new cluster, you must specify the OpenSearch retention time and Persistent Volume Claim (PVC) size, Prometheus PVC, retention time, and retention size. When configuring an existing cluster, you can only set OpenSearch retention time, Prometheus retention time, and retention size.
The following table describes the recommendations for both OpenSearch
and Prometheus retention size and PVC size for a cluster with 34 nodes.
Retention time depends on the space allocated for the data. To calculate
the required retention time, use the
{retention time} = {retention size} / {amount of data per day} formula.
Service |
Required space per day |
Description |
|---|---|---|
OpenSearch |
|
When setting Persistent Volume Claim Size for OpenSearch during the cluster creation, take into account that it defines the PVC size for a single instance of the OpenSearch cluster. StackLight in HA mode has 3 OpenSearch instances. Therefore, for a total OpenSearch capacity, multiply the PVC size by 3. |
Prometheus |
|
Every Prometheus instance stores the entire database. Multiple replicas store multiple copies of the same data. Therefore, treat the Prometheus PVC size as the capacity of Prometheus in the cluster. Do not sum them up. Prometheus has built-in retention mechanisms based on the database size and time series duration stored in the database. Therefore, if you miscalculate the PVC size, retention size set to ~1 GB less than the PVC size will prevent disk overfilling. |
StackLight integration with OpenStack¶
StackLight integration with OpenStack includes automatic discovery of RabbitMQ
credentials for notifications and OpenStack credentials for OpenStack API
metrics. For details, see the
openstack.rabbitmq.credentialsConfig and
openstack.telegraf.credentialsConfig parameters description in
StackLight configuration parameters.
Workload monitoring¶
Lifecycle management operations of a MOSK cluster may impose impact on its workloads and, specifically, may cause network connectivity interruptions for instances running in OpenStack. To make sure that the downtime caused on the cloud applications still fits into Service Level Agreements (SLAs), MOSK provides the tooling to measure the network availability of instances.
Additionally, continuous monitoring of the network connectivity in the cluster is essential for early detection of infrastructure problems.
MOSK offers cloud operators to oversee the availability of workloads hosted in their OpenStack infrastructure on several levels:
Monitoring of floating IP addresses through the Cloudprober service
Monitoring of network ports availability through the Portprober service
Floating IP address availability monitoring (Cloudprober)¶
Available since MOSK 23.2 TechPreview
The floating IP address availability monitoring service (Cloudprober) is a special probing agent that starts on controller nodes and periodically pings selected floating IP addresses. As of today, the agent supports only Internet Control Message Protocol (ICMP) to determine the IP address availability.
To monitor the availability of floating IP addresses, your MOSK cluster and workloads need to meet the following requirements:
There must be the layer-3 connectivity between the clusters floating IP networks and nodes running the OpenStack control plane.
The guest operating system of the monitored OpenStack instances must allow the ICMP ingress and egress traffic.
OpenStack security groups used by the monitored instances must allow the ICMP ingress and egress traffic.
To enable the floating IP address availability monitoring service, use the
following OpenStackDeployment definition:
spec:
features:
services:
- cloudprober
For the detailed configuration procedure of the floating IP address availability monitoring service, refer to Configure monitoring of cloud workload availability.
Network port availability monitoring (Portprober)¶
Available since MOSK 24.2 TechPreview
The network port availability monitoring service (Portprober) is implemented
as an extension to OpenStack Neutron service which gets enabled automatically
together with the cloudprober service described above.
Also, you can enable Portprober explicitly, regardless of whether Cloudprober
is enabled or not. To do so, specify the following structure in the
OpenStackDeployment custom resource:
spec:
features:
neutron:
extensions:
portprober:
enabled: true
The Portprober service is supported only for the following cloud configurations:
OpenStack version is Antelope or newer
Neutron OVS backend for networking (Tungsten Fabric and OVN backends are not supported)
The Portprober agent automatically connects to all OpenStack virtual networks
and probes all the ports that are plugged in there and are in the bound
state, meaning they are associated with an instance or a network service.
The service makes no difference between private and external networks and also reports the availability of the ports that belong to virtual routers.
The service relies on the ARP protocol to determine port availability and does not require any security groups to be assigned to monitored instances, as opposed to the Floating IP address monitoring service (Cloudprober).
Among the known limitations of the network port availability monitoring service is the lack of support for IPv6. The service ignores the ports that do not have IPv4 addresses associated with them.
StackLight logging indices¶
Available since MCC 2.26.0 (17.1.0 and 16.1.0)
StackLight logging indices are managed by OpenSearch data streams, which are introduced in OpenSearch 2.6. It is a convenient way to manage insert-only pipelines such as log message collection. The solution consists of the following elements:
Data stream objects that can be referred to as alias:
Audit - dedicated for Container Cloud, MKE, and host audit logs, ensuring data integrity and security.
System - replaces Logstash for system logs, provides a streamlined approach to log management.
Write index - current index where ingestion can be performed without removing a data stream.
Read indices - indices created after the rollover mechanism is applied.
Rollover policy - creating new write index for data stream based on the size of shards
Example of an initial index list:
health status index uuid pri rep docs.count docs.deleted store.size pri.store.size
green open .ds-audit-000001 30q4HLGmR0KmpRR8Kvy5jw 1 1 2961719 0 496.3mb 248mb
green open .ds-system-000001 5_eFtMAFQa6aFB7nttHjkA 1 1 2476 0 6.1mb 3mb
Example of the index after the rollover is applied to the audit index:
health status index uuid pri rep docs.count docs.deleted store.size pri.store.size
green open .ds-audit-000001 30q4HLGmR0KmpRR8Kvy5jw 1 1 9819913 0 1.5gb 784.8mb
green open .ds-audit-000002 U1fbs0i9TJmOsAOoR7cERg 1 1 2961719 0 496.3mb 248mb
green open .ds-system-000001 5_eFtMAFQa6aFB7nttHjkA 1 1 2476 0 6.1mb 3mb
See also
Audit and system index templates¶
The following table contains a simplified template of the audit and system indices. The user can perform aggregation queries over keyword fields.
Field |
Type |
Description |
|---|---|---|
|
date |
Time when a log event was produced, if available in the parsed message. Otherwise time when the event was ingested. |
|
keyword |
Identifier of the Docker container that the application generating the event was running in. |
|
text |
Name of the Docker image defined as |
|
keyword |
Name of the Docker container that the application generating the event was running in. |
|
keyword |
Source of the event: |
|
keyword |
Name of the application that produced the message. |
|
keyword |
Name of the host that the message was collected from. |
|
keyword |
Path on the host to the source file for the message if the message was not produced by the application running in the container or system unit. |
|
keyword |
Severity level of the event taken from the parsed message content. |
|
text |
Unparsed content of the event message. |
|
flat_object |
Kubernetes metadata labels of the pod that runs the Docker container of the application. |
|
keyword |
Kubernetes namespace where the application pod was running. |
|
keyword |
Kubernetes pod name of the pod running the application Docker container. |
|
keyword |
Type of orchestrator: |
The following table contains a simplified template of extra fields for the system index that are not present in the audit template.
Field |
Type |
Description |
|---|---|---|
|
keyword |
IP address of the HTTP request destination. |
|
keyword |
Name of the OpenStack service that the HTTP request was sent to. |
|
long |
Request duration in nanoseconds. |
|
keyword |
Request ID generated by OpenStack. |
|
keyword |
HTTP request method. |
|
keyword |
Path of the HTTP URL request. |
|
long |
HTTP status code of the response. |
|
keyword |
IP address of the HTTP request source. |
System index mapping to the Logstash index¶
The following table lists mapping of the system index fields to the Logstash ones:
System |
Logstash Removed in MCC 2.26.0 (17.1.0 and 16.1.0) |
|---|---|
|
|
|
|
|
|
|
|
|
n/a |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
n/a |
|
|
|
|
|
|
|
|
|
|
|
n/a |
Blueprints¶
This section contains a collection of Mirantis OpenStack for Kubernetes (MOSK) architecture blueprints that include common cluster topology and configuration patterns that can be referred to when building a MOSK cloud. Every blueprint is validated by Mirantis and is known to work. You can use these blueprints alone or in combination, although the interoperability of all possible combinations can not be guaranteed.
The section provides information on the target use cases, pros and cons of every blueprint and outlines the extents of its applicability. However, do not hesitate to reach out to Mirantis if you have any questions or doubts on whether a specific blueprint can be applied when designing your cloud.
Remote compute nodes¶
Introduction to edge computing¶
Although a classic cloud approach allows resources to be distributed across multiple regions, it still needs powerful data centers to host control planes and compute clusters. Such regional centralization poses challenges when the number of data consumers grows. It becomes hard to access the resources hosted in the cloud even though the resources are located in the same geographic region. The solution would be to bring the data closer to the consumer. And this is exactly what edge computing provides.
Edge computing is a paradigm that brings computation and data storage closer to the sources of data or the consumer. It is designed to improve response time and save bandwidth.
A few examples of use cases for edge computing include:
Hosting a video stream processing application on premises of a large stadium during the Super Bowl match
Placing the inventory or augmented reality services directly in the industrial facilities, such as storage, powerplant, shipyard, and so on
A small computation node deployed in a far-distanced village supermarket to host an application for store automatization and accounting
These and many other use cases could be solved by deploying multiple edge clusters managed from a single central place. The idea of centralized management plays a significant role for the business efficiency of the edge cloud environment:
Cloud operators obtain a single management console for the cloud that simplifies the Day-1 provisioning of new edge sites and Day-2 operations across multiple geographically distributed points of presence
Cloud users get ability to transparently connect their edge applications with central databases or business logic components hosted in data centers or public clouds
Depending on the size, location, and target use case, the points of presence comprising an edge cloud environment can be divided into five major categories. Mirantis OpenStack powered by Mirantis Container Cloud offers reference architectures to address the centralized management in core and regional data centers as well as edge sites.
Overview of the remote compute nodes approach¶
Remote compute nodes is one of the approaches to the implementation of the edge computing concept offered by MOSK. The topology consists of a MOSK cluster residing in a data center, which is extended with multiple small groups of compute nodes deployed in geographically distanced remote sites. Remote compute nodes are integrated into the MOSK cluster just like the nodes in the central site with their configuration and life cycle managed through the same means.
Along with compute nodes, remote sites need to incorporate network gateway components that allow application users to consume edge services directly without looping the traffic through the central site.
Design considerations for a remote site¶
Deployment of an edge cluster managed from a single central place starts with a proper planning. This section provides recommendations on how to approach the deployment design.
Mirantis recommends organizing nodes in each remote site into separate Availability Zones in the MOSK Compute (OpenStack Nova), Networking (OpenStack Neutron), and Block Storage (OpenStack Cinder) services. This enables the cloud users to be aware of the failure domain represented by a remote site and distribute the parts of their applications accordingly.
Typically, high latency in between the central control plane and remote sites makes it not feasible to rely on Ceph as a storage for the instance root/ephemeral and block data.
Mirantis recommends that you configure the remote sites to use the following backends:
Local storage (LVM or QCOW2) as a storage backend for the MOSK Compute service. See Image storage backend for the configuration details.
LVM on iSCSI backend for the MOSK Block Storage service. See Enable LVM block storage for the enablement procedure.
To maintain the small size of a remote site, the compute nodes need to be hyper-converged and combine the compute and block storage functions.
There is no limitation on the number of the remote sites and their size.
However, when planning the cluster, ensure consistency between the total number
of nodes managed by a single control plane and the value of the size
parameter set in the OpenStackDeployment custom resource. For the list of
supported sizes, refer to Main elements.
Additionally, the sizing of the remote site needs to take into account the characteristics of the networking channel with the main site.
Typically, an edge site consists of 3-7 compute nodes installed in a single, usually rented, rack.
Mirantis recommends keeping the network latency between the main and remote sites as low as possible. For stable interoperability of cluster components, the latency needs to be around 30-70 milliseconds. Though, depending on the cluster configuration and dynamism of the workloads running in the remote site, the stability of the cluster can be preserved with the latency of up to 190 milliseconds.
The bandwidth of the communication channel between the main and remote sites needs to be sufficient to run the following traffic:
The control plane and management traffic, such as OpenStack messaging, database access, MOSK underlay Kubernetes cluster control plane, and so on. A single remote compute node in the idle state requires at minimum 1.5 Mbit/s of bandwidth to perform the non-data plane communications.
The data plane traffic, such as OpenStack image operations, instances VNC console traffic, and so on, that heavily depend on the profile of the workloads and other aspects of the cloud usage.
In general, Mirantis recommends having a minimum of 100 MBit/s bandwidth between the main and remote sites.
MOSK remote compute nodes architecture is designed to tolerate a temporary loss of connectivity between the main cluster and the remote sites. In case of a disconnection, the instances running on remote compute nodes will keep running normally preserving their ability to read and write ephemeral and block storage data presuming it is located in the same site, as well as connectivity to their neighbours and edge application users. However, the instances will not have access to any cloud services or applications located outside of their remote site.
Since the MOSK control plane communicates with remote compute nodes through the same network channel, cloud users will not be able to perform any manipulations, for example, instance creation, deletion, snapshotting, and so on, over their edge applications until the connectivity gets restored. MOSK services providing high availability to cloud applications, such as the Instance HA service and Network service, need to be connected to the remote compute nodes to perform a failover of application components running in the remote site.
Once the connectivity between the main and the remote site restores, all functions become available again. The period during which an edge application can sustain normal function after a connectivity loss is determined by multiple factors including the selected networking backend for the MOSK cluster. Mirantis recommends that a cloud operator performs a set of test manipulations over the cloud resources hosted in the remote site to ensure that it has been fully restored.
When configured in Tungsten Fabric-powered clouds, the Graceful restart and long-lived graceful restart feature significantly improves the MOSK ability to sustain the connectivity of workloads running at remote sites in situations when a site experiences a loss of connection to the central hosting location of the control plane.
Extensive testing has demonstrated that remote sites can effectively withstand a 72-hour control plane disconnection with zero impact on the running applications.
Given that a remote site communicates with its main MOSK cluster across a wide area network (WAN), it becomes important to protect sensitive data from being intercepted and viewed by a third party. Specifically, you should ensure the protection of the data belonging to the following cloud components:
- Mirantis Container Cloud life-cycle management plane
Bare metal servers provisioning and control, Kubernetes cluster deployment and management, Mirantis StackLight telemetry
- MOSK control plane
Communication between the components of OpenStack, Tungsten Fabric, and Mirantis Ceph
- MOSK data plane
Cloud application traffic
The most reliable way to protect the data is to configure the network equipment in the data center and the remote site to encapsulate all the bypassing remote-to-main communications into an encrypted VPN tunnel. Alternatively, Mirantis Container Cloud and MOSK can be configured to force encryption of specific types of network traffic, such as:
Kubernetes networking for MOSK underlying Kubernetes cluster that handles the vast majority of in-MOSK communications
OpenStack tenant networking that carries all the cloud application traffic
The ability to enforce traffic encryption depends on the specific version of the Mirantis Container Cloud and MOSK in use, as well as the selected SDN backend for OpenStack.
Remote compute nodes with Tungsten Fabric¶
TechPreview
In MOSK, the main cloud that controls remote computes can be the regional site that locates the regional cluster and the MOSK control plane. Additionally, it can contain a local storage and compute nodes.
The remote computes implementation in MOSK considers Tungsten Fabric as an SDN solution.
Remote computes bare metal servers are configured as Kubernetes workers hosting the deployments for:
Tungsten Fabric vRouter-gateway service
Nova-compute
Local storage (LVM with iSCSI block storage)
Large clusters¶
This section describes a validated MOSK cluster architecture that is capable of handling 10,000 instances under a single control plane.
Hardware characteristics¶
Role |
Nodes count |
Server specification |
|---|---|---|
Management cluster Kubernetes nodes |
3 |
|
MOSK cluster Kubernetes master nodes |
3 |
|
OpenStack controller nodes |
5 |
|
OpenStack compute and storage nodes |
Up to 500 total |
|
StackLight nodes |
3 |
|
Cluster architecture¶
Configuration |
Value |
|---|---|
Dedicated StackLight nodes |
Yes |
Dedicated Ceph storage nodes |
Yes |
Dedicated control plane Kubernetes nodes |
Yes |
Dedicated OpenStack gateway nodes |
No, collocated with OpenStack controller nodes |
OpenStack networking backend |
Open vSwitch, no Distributed Virtual Router |
Cluster size in the |
|
Cluster validation¶
The architecture validation is perfomed by means of simultanious creation of multiple OpenStack resources of various types and execution of functional tests against each resource. The amount of resources hosted in the cluster at the moment when a certain threshold of non-operational resources starts being observed, is described below as cluster capacity limit.
Note
A successfully created resource has the Active status in the API
and passes the functional tests, for example, its floating IP address is
accessible. The MOSK cluster is considered to be able to
handle the created resources if it successfully performs the LCM operations
including the OpenStack services restart, both on the control and data
plane.
Note
The key limiting factor for creating more OpenStack objects in this illustrative setup is hardware resources (vCPU and RAM) available on the compute nodes.
OpenStack resource |
Limit |
|---|---|
Instances |
11101 |
Network ports - instances |
37337 |
Network ports - service (avg. per gateway node) |
3517 |
Volumes |
2784 |
Routers |
2448 |
Networks |
3383 |
Orchestration stacks |
2419 |
Node role |
Load average |
vCPU |
RAM in GB |
|---|---|---|---|
OpenStack controller + gateway |
10 |
10 |
100 |
OpenStack compute |
30 |
25 |
160 |
Ceph storage |
2 |
2 |
15 |
StackLight |
10 |
8 |
102 |
Kubernetes master |
10 |
6 |
13 |
Cephless cloud¶
Available since MOSK 23.2 TechPreview
Persistent storage is a key component of any MOSK deployment. Out of the box, MOSK includes an open-source software-defined storage solution (Ceph), which hosts various kinds of cloud application data, such as root and ephemeral disks for virtual machines, virtual machine images, attachable virtual block storage, and object data. In addition, a Ceph cluster usually acts as a storage for the internal MOSK components, such as Kubernetes, OpenStack, StackLight, and so on.
Being distributed and redundant by design, Ceph requires a certain minimum amount of servers, also known as OSD or storage nodes, to work. A production-grade Ceph cluster typically consists of at least nine storage nodes, while a development and test environment may include four to six servers. For details, refer to MOSK cluster hardware requirements.
It is possible to reduce the overall footprint of a MOSK cluster by collocating the Ceph components with hypervisors on the same physical servers; this is also known as hyper-converged design. However, this architecture still may not satisfy the requirements of certain use cases for the cloud.
Standalone telco-edge MOSK clouds typically consist of three to seven servers hosted in a single rack, where every piece of CPU, memory and disk resources is strictly accounted and better be dedicated to the cloud workloads, rather than control plane. For such clouds, where the cluster footprint is more important than the resiliency of the application data storage, it makes sense either not to have a Ceph cluster at all or to replace it with some primitive non-redundant solution.
Enterprise virtualization infrastructure with third-party storage is not a rare strategy among large companies that rely on proprietary storage appliances, provided by NetApp, Dell, HPE, Pure Storage, and other major players in the data storage sector. These industry leaders offer a variety of storage solutions meticulously designed to suit various enterprise demands. Many companies, having already invested substantially in proprietary storage infrastructure, prefer integrating MOSK with their existing storage systems. This approach allows them to leverage this investment rather than incurring new costs and logistical complexities associated with migrating to Ceph.
Architecture¶
Kind of data |
MOSK component |
Data storage in Cephless architecture |
Configuration |
|---|---|---|---|
Root and ephemeral disks of instances |
Compute service (OpenStack Nova) |
|
|
Volumes |
Block Storage service (OpenStack Cinder) |
|
|
Volumes backups |
Block Storage service (OpenStack Cinder) |
Alternatively, you can disable the volume backup functionality. |
|
Tungsten Fabric database backups |
Tungsten Fabric (Cassandra, ZooKeeper) |
External NFS share TechPreview Alternatively, you can disable the Tungsten Fabric database backups functionality. |
|
OpenStack database backups |
OpenStack (MariaDB) |
Alternatively, you can disable the database backup functionality. |
|
Results of functional testing |
OpenStack Tempest |
Local file system of MOSK controller nodes. The |
|
Instance images and snapshots |
Image service (OpenStack Glance) |
You can configure the Block Storage service (OpenStack Cinder) to be used as a storage backend for images and snapshots. In this case, each image is represented as a volume. Important Representing volumes as images implies a hard requirement for the selected block storage backend to support multi-attach capability that is concurrent reads and writes to and from a single volume. |
|
Application object data |
Object storage service (Ceph RADOS Gateway) |
External S3, Swift, or any other third-party storage solutions compatible with object access protocols. Note An external object storage solution will not be integrated into the MOSK identity service (OpenStack Keystone), the cloud applications will need to take care of managing access to their object data themselves. If no Ceph is deployed as part of a cluster, the MOSK built-in Object Storage service API endpoints are disabled automatically. |
|
Logs, metrics, alerts |
Mirantis StackLight (Prometeus, Alertmanager, Patroni, OpenSearch) |
Local file system of MOSK controller nodes. StackLight must be deployed in the HA mode, when all its data gets stored on the local file system of the nodes running StackLight services. In this mode, StackLight components get configured to handle the data replication themselves. |
Limitations¶
The determination of whether a MOSK cloud will include Ceph or not should take place during its planning and design phase. Once the deployment is complete, reconfiguring the cloud to switch between Ceph and non-Ceph architectures becomes impossible.
Mirantis recommends avoiding substitution of Ceph-backed persistent volumes in the MOSK underlying Kubernetes cluster with local volumes (local volume provisioner) for production environments. MOSK does not support such configuration unless the components that rely on these volumes can replicate their data themselves, for example, StackLight. Volumes provided by the local volume provisioner are not redundant, as they are bound to just a single node and can only be mounted from the Kubernetes pods running on the same nodes.
Node maintenance API¶
This section describes internal implementation of the node maintenance API and how OpenStack and Tungsten Fabric controllers communicate with LCM and each other during a managed cluster update.
Node maintenance API objects¶
The node maintenance API consists of the following objects:
Cluster level:
ClusterWorkloadLockClusterMaintenanceRequest
Node level:
NodeWorkloadLockNodeMaintenanceRequest
WorkloadLock objects¶
The WorkloadLock objects are created by each Application Controller.
These objects prevent LCM from performing any changes on the cluster or node
level while the lock is in the active state. The inactive state of the lock
means that the Application Controller has finished its work and the LCM can
proceed with the node or cluster maintenance.
apiVersion: lcm.mirantis.com/v1alpha1
kind: ClusterWorkloadLock
metadata:
name: cluster-1-openstack
spec:
controllerName: openstack
status:
state: active # inactive;active;failed (default: active)
errorMessage: ""
release: "6.16.0+21.3"
apiVersion: lcm.mirantis.com/v1alpha1
kind: NodeWorkloadLock
metadata:
name: node-1-openstack
spec:
nodeName: node-1
controllerName: openstack
status:
state: active # inactive;active;failed (default: active)
errorMessage: ""
release: "6.16.0+21.3"
MaintenanceRequest objects¶
The MaintenanceRequest objects are created by LCM. These objects notify
Application Controllers about the upcoming maintenance of a cluster or
a specific node.
apiVersion: lcm.mirantis.com/v1alpha1
kind: ClusterMaintenanceRequest
metadata:
name: cluster-1
spec:
scope: drain # drain;os
apiVersion: lcm.mirantis.com/v1alpha1
kind: NodeMaintenanceRequest
metadata:
name: node-1
spec:
nodeName: node-1
scope: drain # drain;os
The scope parameter in the object specification defines the impact on
the managed cluster or node. The list of the available options include:
drainA regular managed cluster update. Each node in the cluster goes over a drain procedure. No node reboot takes place, a maximum impact includes restart of services on the node including Docker, which causes the restart of all containers present in the cluster.
osA node might be rebooted during the update. Triggers the workload evacuation by the OpenStack Controller (Rockoon).
When the MaintenanceRequest object is created, an Application Controller
executes a handler to prepare workloads for maintenance and put appropriate
WorkloadLock objects into the inactive state.
When maintenance is over, LCM removes MaintenanceRequest objects,
and the Application Controllers move their WorkloadLocks objects into
the active state.
OpenStack Controller maintenance API¶
When LCM creates the ClusterMaintenanceRequest object, the OpenStack
Controller (Rockoon) ensures that all OpenStack components are in the
Healthy state, which means that the pods are up and running, and the
readiness probes are passing.
ClusterMaintenanceRequest object creation flow¶
When LCM creates the NodeMaintenanceRequest, the OpenStack Controller:
Prepares components on the node for maintenance by removing
nova-computefrom scheduling.If the reboot of a node is possible, the instance migration workflow is triggered. The Operator can configure the instance migration flow through the Kubernetes node annotation and should define the required option before the managed cluster update. For configuration details, refer to Instance migration configuration for hosts.
Also, since MOSK 25.1, cloud users can mark their instances for LCM to handle them individually during host maintenance operations. This allows for greater flexibility during cluster updates, especially for workloads that are sensitive to live migration. For details, refer to Configure per-instance migration mode.
If the OpenStack Controller cannot migrate instances due to errors, it is suspended unless all instances are migrated manually or the
openstack.lcm.mirantis.com/instance_migration_modeannotation is set toskip.
NodeMaintenanceRequest object creation flow¶
When the node maintenance is over, LCM removes the NodeMaintenanceRequest
object and the OpenStack Controller:
Verifies that the Kubernetes Node becomes
Ready.Verifies that all OpenStack components on a given node are
Healthy, which means that the pods are up and running, and the readiness probes are passing.Ensures that the OpenStack components are connected to RabbitMQ. For example, the Neutron Agents become alive on the node, and compute instances are in the
UPstate.
Note
The OpenStack Controller enables you to have only one
nodeworkloadlock object at a time in the inactive state. Therefore,
the update process for nodes is sequential.
NodeMaintenanceRequest object removal flow¶
When the cluster maintenance is over, the OpenStack Controller sets the
ClusterWorkloadLock object to back active and the update completes.
CLusterMaintenanceRequest object removal flow¶
Tungsten Fabric Controller maintenance API¶
The Tungsten Fabric (TF) Controller creates and uses both types of
workloadlocks that include ClusterWorkloadLock and NodeWorkloadLock.
When the ClusterMaintenanceRequest object is created, the TF Controller
verifies the TF cluster health status and proceeds as follows:
If the cluster is
Ready, the TF Controller moves theClusterWorkloadLockobject to the inactive state.Otherwise, the TF Controller keeps the
ClusterWorkloadLockobject in the active state.
When the NodeMaintenanceRequest object is created, the TF Controller
verifies the vRouter pod state on the corresponding node and proceeds as
follows:
If all containers are
Ready, the TF Controller moves theNodeWorkloadLockobject to the inactive state.Otherwise, the TF Controller keeps the
NodeWorkloadLockin the active state.
Note
If there is a NodeWorkloadLock object in the inactive state
present in the cluster, the TF Controller does not process the
NodeMaintenanceRequest object for other nodes until this inactive
NodeWorkloadLock object becomes active.
When the cluster LCM removes the MaintenanceRequest object, the TF
Controller waits for the vRouter pods to become ready and proceeds as follows:
If all containers are in the
Readystate, the TF Controller moves theNodeWorkloadLockobject to the active state.Otherwise, the TF Controller keeps the
NodeWorkloadLockobject in the inactive state.
Cluster update flow¶
This section describes the MOSK cluster update flow to the product releases that contain major updates and require node reboot such as support for new Linux kernel, and similar.
The diagram below illustrates the sequence of operations controlled by
LCM and taking place during the update under the hood. We assume that the
ClusterWorkloadLock and NodeWrokloadLock objects present in the cluster
are in the active state before the cloud operator triggers the update.
See also
For details about the Application Controllers flow during different maintenance stages, refer to:
Phase 1: The Operator triggers the update¶
The Operator sets appropriate annotations on nodes and selects suitable migration mode for workloads.
The Operator triggers the managed cluster update through the Mirantis Container Cloud web UI as described in Step 2. Initiate MOSK cluster update.
LCM creates the
ClusterMaintenanceobject and notifies the application controllers about planned maintenance.
Phase 2: LCM triggers the OpenStack and Ceph update¶
The OpenStack update starts.
Ceph is waiting for the OpenStack
ClusterWorkloadLockobject to become inactive.When the OpenStack update is finalized, the OpenStack Controller marks
ClusterWorkloadLockas inactive.The Ceph Controller triggers an update of the Ceph cluster.
When the Ceph update is finalized, Ceph marks the
ClusterWorkloadLockobject as inactive.
Phase 3: LCM initiates the Kubernetes master nodes update¶
If a master node has collocated roles, LCM creates
NodeMainteananceRequestfor the node.All Application Controllers mark their
NodeWorkloadLockobjects for this node as inactive.LCM starts draining the node by gracefully moving out all pods from the node. The DaemonSet pods are not evacuated and left running.
LCM downloads the new version of the LCM Agent and runs its states.
Note
While running Ansible states, the services on the node may be restarted.
The above flow is applied to all Kubernetes master nodes one by one.
LCM removes
NodeMainteananceRequest.
Phase 4: LCM initiates the Kubernetes worker nodes update¶
LCM creates
NodeMaintenanceRequestfor the node with specifying scope.Application Controllers start preparing the node according to the scope.
LCM waits until all Application Controllers mark their
NodeWorkloadLockobjects for this node as inactive.All pods are evacuated from the node by draining it. This does not apply to the DaemonSet pods, which cannot be removed.
LCM downloads the new version of the LCM Agent and runs its states.
Note
While running Ansible states, the services on the node may be restarted.
The above flow is applied to all Kubernetes worker nodes one by one.
LCM removes
NodeMainteananceRequest.
Phase 5: Finalization¶
LCM triggers the update for all other applications present in the cluster, such as StackLight, Tungsten Fabric, and others.
LCM removes
ClusterMaintenanceRequest.
After a while the cluster update completes and becomes fully operable again.
Parallelizing node update operations¶
Available since MOSK 23.2 TechPreview
OpenStack Controller (Rockoon)
Since MOSK 25.1, the OpenStack Controller has been open-sourced under the name Rockoon and is maintained as an independent open-source project going forward.
As part of this transition, all openstack-controller pods are named
rockoon pods across the MOSK documentation and deployments. This change
does not affect functionality, but this is the reminder for the users to
utilize the new naming for pods and other related artifacts accordingly.
MOSK enables you to parallelize node update operations, significantly improving the efficiency of your deployment. This capability applies to any operation that utilizes the Node Maintenance API, such as cluster updates or graceful node reboots.
The core implementation of parallel updates is handled by the LCM Controller
ensuring seamless execution of parallel operations. LCM starts performing an
operation on the node only when all NodeWorkloadLock objects for the node
are marked as inactive. By default, the LCM Controller creates one
NodeMaintenanceRequest at a time.
Each application controller, including Ceph, OpenStack, and Tungsten Fabric
Controllers, manages parallel NodeMaintenanceRequest objects independently.
The controllers determine how to handle and execute parallel node maintenance
requests based on specific requirements of their respective applications.
To understand the workflow of the Node Maintenance API, refer to
WorkloadLock objects.
Enhancing parallelism during node updates¶
- Set the nodes update order.
You can optimize parallel updates by setting the order in which nodes are updated. You can accomplish this by configuring
upgradeIndexof theMachineobject. For the procedure, refer to Change the upgrade order of a machine.
- Increase parallelism.
Boost parallelism by adjusting the maximum number of worker node updates that are allowed during LCM operations using the
spec.providerSpec.value.maxWorkerUpgradeCountconfiguration parameter, which is set to1by default.For configuration details, refer to Configure the parallel update of worker nodes.
- Execute LCM operations.
Run LCM operations, such as cluster updates, taking advantage of the increased parallelism.
OpenStack nodes update¶
By default, the OpenStack Controller handles the NodeMaintenanceRequest
objects as follows:
Updates the OpenStack controller nodes sequentially (one by one).
Updates the gateway nodes sequentially. Technically, you can increase the number of gateway nodes upgrades allowed in parallel using the
nwl_parallel_max_gatewayparameter but Mirantis does not recommend to do so.Updates the compute nodes in parallel. The default number of allowed parallel updates is
30. You can adjust this value through thenwl_parallel_max_computeparameter.Parallelism considerations for compute nodes
When considering parallelism for compute nodes, take into account that during certain pod restarts, for example, the
openvswitch-vswitchdpods, a brief instance downtime may occur. Select a suitable level of parallelism to minimize the impact on workloads and prevent excessive load on the control plane nodes.If your cloud environment is distributed across failure domains, which are represented by Nova availability zones, you can limit the parallel updates of nodes to only those within the same availability zone. This behavior is controlled by the
respect_nova_azoption in the OpenStack Controller.
The OpenStack Controller configuration is stored in the
rockoon-config configMap of the osh-system namespace.
The options are picked up automatically after update. To learn more about
the OpenStack Controller (Rockoon) configuration parameters,
refer to OpenStack Controller configuration.
Ceph nodes update¶
By default, the Ceph Controller handles the NodeMaintenanceRequest
objects as follows:
Updates the non-storage nodes sequentially. Non-storage nodes include all nodes that have
mon,mgr,rgw, ormdsroles.Updates storage nodes in parallel. The default number of allowed parallel updates is calculated automatically based on the minimal failure domain in a Ceph cluster.
Parallelism calculations for storage nodes
The Ceph Controller automatically calculates the parallelism number in the following way:
Finds the minimal failure domain for a Ceph cluster. For example, the minimal failure domain is
rack.Filters all currently requested nodes by minimal failure domain. For example, parallelism equals to 5, and LCM requests 3 nodes from the
rack1rack and 2 nodes from therack2rack.Handles each filtered node group one by one. For example, the controller handles in parallel all nodes from
rack1before processing nodes fromrack2.
The Ceph Controller handles non-storage nodes before the storage ones. If there are node requests for both node types, the Ceph Controller handles sequentially the non-storage nodes first. Therefore, Mirantis recommends setting the upgrade index of a higher priority for the non-storage nodes to decrease the total upgrade time.
If the minimal failure domain is host, the Ceph Controller updates only
one storage node per failure domain unit. This results in updating all Ceph
nodes sequentially, despite the potential for increased parallelism.
Tungsten Fabric nodes update¶
By default, the Tungsten Fabric Controller handles the
NodeMaintenanceRequest objects as follows:
Updates the Tungsten Fabric Controller and gateway nodes sequentially.
Updates the vRouter nodes in parallel. The Tungsten Fabric Controller allows updating up to 30 vRouter nodes in parallel.
Maximum amount of vRouter nodes in maintenance
While the Tungsten Fabric Controller has the capability to process up to 30
NodeMaintenanceRequestobjects targeted to vRouter nodes, the actual amount may be lower. This is due to a check that ensures OpenStack readiness to unlock the relevant nodes for maintenance. If OpenStack allows for maintenance, the Tungsten Fabric Controller verifies the vRouter pods. Upon successful verification, theNodeWorkloadLockobject is switched to the maintenance mode.
Deployment Guide¶
Mirantis OpenStack for Kubernetes (MOSK) enables the operator to create, scale, update, and upgrade OpenStack deployments on Kubernetes through a declarative API.
The Kubernetes built-in features, such as flexibility, scalability, and declarative resource definition make MOSK a robust solution.
Plan the deployment¶
The detailed plan of any Mirantis OpenStack for Kubernetes (MOSK) deployment is determined on a per-cloud basis. For the MOSK reference architecture and design overview, see Reference Architecture.
Also, read through Mirantis Container Cloud Reference Architecture: Container Cloud bare metal as a MOSK cluster is deployed on top of a bare metal cluster managed by Mirantis Container Cloud.
Note
One of the industry best practices is to verify every new update or configuration change in a non-customer-facing environment before applying it to production. Therefore, Mirantis recommends having a staging cloud, deployed and maintained along with the production clouds. The recommendation is especially applicable to the environments that:
Receive updates often and use continuous delivery. For example, any non-isolated deployment of Mirantis Container Cloud.
Have significant deviations from the reference architecture or third party extensions installed.
Are managed under the Mirantis OpsCare program.
Run business-critical workloads where even the slightest application downtime is unacceptable.
A typical staging cloud is a complete copy of the production environment including the hardware and software configurations, but with a bare minimum of compute and storage capacity.
Provision and deploy a management cluster¶
The bare metal management system enables the Infrastructure Operator to deploy a Container Cloud management cluster on a set of bare metal servers. It also enables Container Cloud to deploy MOSK clusters on bare metal servers without a pre-provisioned operating system.
This section instructs you on how to provision and deploy a Container Cloud management cluster.
Deploy a management cluster¶
Note
The deprecated bootstrap procedure using Bootstrap v1 was removed for the sake of Bootstrap v2 in Container Cloud 2.26.0.
Introduction¶
Mirantis Container Cloud Bootstrap v2 provides best user experience to set up Container Cloud. Using Bootstrap v2, you can provision and operate management clusters using required objects through the Container Cloud API.
Basic concepts and components of Bootstrap v2 include:
- Bootstrap cluster
Bootstrap cluster is any kind-based Kubernetes cluster that contains a minimal set of Container Cloud bootstrap components allowing the user to prepare the configuration for management cluster deployment and start the deployment. The list of these components includes:
- Bootstrap Controller
Controller that is responsible for:
Configuration of a bootstrap cluster with provider charts through the bootstrap Helm bundle.
Configuration and deployment of a management cluster and its related objects.
- Helm Controller
Operator that manages Helm chart releases. It installs the Container Cloud bootstrap and provider charts configured in the bootstrap Helm bundle.
- Public API charts
Helm charts that contain custom resource definitions for Container Cloud resources.
- Admission Controller
Controller that performs mutations and validations for the Container Cloud resources including cluster and machines configuration.
Currently one bootstrap cluster can be used for deployment of only one management cluster. For example, to add a new management cluster with different settings, a new bootstrap cluster must be created from scratch.
- Bootstrap region
BootstrapRegionis the first object to create in the bootstrap cluster for the Bootstrap Controller to identify and install provider components onto the bootstrap cluster. After, the user can prepare and deploy a management cluster with related resources.The bootstrap region is a starting point for the cluster deployment. The user needs to approve the
BootstrapRegionobject. Otherwise, the Bootstrap Controller will not be triggered for the cluster deployment.
- Bootstrap Helm bundle
Helm bundle that contains charts configuration for the bootstrap cluster. This object is managed by the Bootstrap Controller that updates the provider bundle in the
BootstrapRegionobject. The Bootstrap Controller always configures provider charts listed in theregionalsection of the Container Cloud release for the provider. Depending on the cluster configuration, the Bootstrap Controller may update or reconfigure this bundle even after the cluster deployment starts. For example, the Bootstrap Controller enables the provider in the bootstrap cluster only after the bootstrap region is approved for the deployment.
Overview of the deployment workflow¶
Management cluster deployment consists of several sequential stages. Each stage finishes when a specific condition is met or specific configuration applies to a cluster or its machines.
In case of issues at any deployment stage, you can identify the problem
and adjust it on the fly. The cluster deployment does not abort until all
stages complete by means of the infinite-timeout option enabled
by default in Bootstrap v2.
Infinite timeout prevents the bootstrap failure due to timeout. This option is useful in the following cases:
The network speed is slow for artifacts downloading
The infrastructure configuration does not allow fast booting
The inspection of a bare-metal node presupposes more than two HDDSATA disks to attach to a machine
You can track the status of each stage in the bootstrapStatus section of
the Cluster object that is updated by the Bootstrap Controller.
The Bootstrap Controller starts deploying the cluster after you approve the
BootstrapRegion configuration.
The following table describes deployment states of a management cluster that apply in the strict order.
Step |
State |
Description |
|---|---|---|
1 |
|
Verifies proxy configuration in the |
2 |
|
Verifies SSH configuration for the cluster and machines. You can provide any number of SSH public keys, which are added to
cluster machines. But the Bootstrap Controller always adds the
The |
3 |
|
Synchronizes the provider and settings of its components between the
|
4 |
|
Enables the provider and its components if any were disabled by the Bootstrap Controller during preparation of the bootstrap region. A cluster and machines deployment starts after the provider enablement. |
5 |
Nodes readiness |
Waits for the provider to complete nodes deployment that comprises VMs creation and MKE installation. |
6 |
|
Creates required namespaces and IAM secrets. |
7 |
|
Verifies the provider configuration in the provisioned cluster. |
8 |
|
Verifies the Helm bundle readiness for the provisioned cluster. |
9 |
|
Collects the list of deployment controllers and disables them to prepare for pivot. |
10 |
|
Moves all cluster-related objects from the bootstrap cluster to the
provisioned cluster. The copies of |
11 |
|
Enables controllers in the provisioned cluster. |
12 |
|
Updates the |
13 |
|
Disables the Helm Controller before reconfiguration. |
14 |
|
Updates the Helm Controller configuration for the provisioned cluster. |
15 |
Cluster readiness |
Contains information about the global cluster status. The Bootstrap Controller verifies that OIDC, Helm releases, and all Deployments are ready. Once the cluster is ready, the Bootstrap Controller stops managing the cluster. |
Set up a bootstrap cluster¶
The setup of a bootstrap cluster comprises preparation of the seed node, configuration of environment variables, acquisition of the Container Cloud license file, and execution of the bootstrap script.
To set up a bootstrap cluster:
Prepare the seed node:
Verify that the hardware allocated for the installation meets the minimal requirements described in Container Cloud Reference Architecture:.
Install basic Ubuntu 22.04 server using standard installation images of the operating system on the bare metal seed node.
Log in to the seed node that is running Ubuntu 22.04.
Configure the operating system and network:
Operating system and network configuration
Establish a virtual bridge using an IP address of the PXE network on the seed node. Use the following
netplan-based configuration file as an example:# cat /etc/netplan/config.yaml network: version: 2 renderer: networkd ethernets: ens3: dhcp4: false dhcp6: false bridges: br0: addresses: # Replace with IP address from PXE network to create a virtual bridge - 10.0.0.15/24 dhcp4: false dhcp6: false # Adjust for your environment gateway4: 10.0.0.1 interfaces: # Interface name may be different in your environment - ens3 nameservers: addresses: # Adjust for your environment - 8.8.8.8 parameters: forward-delay: 4 stp: false
Apply the new network configuration using
netplan:sudo netplan apply
Verify the new network configuration:
sudo apt update && sudo apt install -y bridge-utils sudo brctl show
Example of system response:
bridge name bridge id STP enabled interfaces br0 8000.fa163e72f146 no ens3
Verify that the interface connected to the PXE network belongs to the previously configured bridge.
Install the current Docker version available for Ubuntu 22.04:
sudo apt-get update sudo apt-get install docker.io
Verify that your logged
USERhas access to the Docker daemon:sudo usermod -aG docker $USER
Verify that the
br_netfilterkernel module is loaded:grep -q br_netfilter /proc/modules || sudo modprobe br_netfilter
Log out and log in again to the seed node to apply the changes.
Verify that Docker is configured correctly and has access to Container Cloud CDN. For example:
docker run --rm alpine sh -c "apk add --no-cache curl; \ curl https://binary.mirantis.com"
The system output must contain a
jsonfile with no error messages. In case of errors, follow the steps provided in Troubleshoot the bootstrap node configuration.Note
If you require all Internet access to go through a proxy server for security and audit purposes, configure Docker proxy settings as described in the official Docker documentation.
To verify that Docker is configured correctly and has access to Container Cloud CDN:
docker run --rm alpine sh -c "export http_proxy=http://<proxy_ip:proxy_port>; \ sed -i ‘s/https/http/g' /etc/apk/repositories; \ apk add --no-cache wget ; \ wget http://binary.mirantis.com; \ cat index.html
Verify that the seed node has direct access to the Baseboard Management Controller (BMC) of each bare metal host. All target hardware nodes must be in the
power offstate.For example, using the IPMI tool:
apt install ipmitool ipmitool -I lanplus -H 'IPMI IP' -U 'IPMI Login' -P 'IPMI password' \ chassis power status
Example of system response:
Chassis Power is off
Prepare the bootstrap script:
Download and run the Container Cloud bootstrap script:
sudo apt-get update sudo apt-get install wget wget https://binary.mirantis.com/releases/get_container_cloud.sh chmod 0755 get_container_cloud.sh ./get_container_cloud.sh
Change the directory to the
kaas-bootstrapfolder created by the script.
Obtain a Container Cloud license file required for the bootstrap:
Obtain a Container Cloud license
Select from the following options:
Open the email from support@mirantis.com with the subject Mirantis Container Cloud License File or Mirantis OpenStack License File
In the Mirantis CloudCare Portal, open the Account or Cloud page
Download the License File and save it as
mirantis.licunder thekaas-bootstrapdirectory on the bootstrap node.Verify that
mirantis.liccontains the previously downloaded Container Cloud license by decoding the license JWT token, for example, using jwt.io.Example of a valid decoded Container Cloud license data with the mandatory
licensefield:{ "exp": 1652304773, "iat": 1636669973, "sub": "demo", "license": { "dev": false, "limits": { "clusters": 10, "workers_per_cluster": 10 }, "openstack": null } }
Warning
The MKE license does not apply to
mirantis.lic. For details about MKE license, see MKE documentation.Export mandatory parameters:
Bare metal network mandatory parameters
Export the following mandatory parameters using the commands and table below:
export KAAS_BM_ENABLED="true" # export KAAS_BM_PXE_IP="172.16.59.5" export KAAS_BM_PXE_MASK="24" export KAAS_BM_PXE_BRIDGE="br0"
Bare metal prerequisites data¶ Parameter
Description
Example value
KAAS_BM_PXE_IPThe provisioning IP address in the PXE network. This address will be assigned on the seed node to the interface defined by the
KAAS_BM_PXE_BRIDGEparameter described below. The PXE service of the bootstrap cluster uses this address to network boot bare metal hosts.172.16.59.5KAAS_BM_PXE_MASKThe PXE network address prefix length to be used with the
KAAS_BM_PXE_IPaddress when assigning it to the seed node interface.24KAAS_BM_PXE_BRIDGEThe PXE network bridge name that must match the name of the bridge created on the seed node during preparation of the system and network configuration described earlier in this procedure.
br0Optional. Configure proxy settings to bootstrap the cluster using proxy:
Proxy configuration
Add the following environment variables:
HTTP_PROXYHTTPS_PROXYNO_PROXYPROXY_CA_CERTIFICATE_PATH
Example snippet:
export HTTP_PROXY=http://proxy.example.com:3128 export HTTPS_PROXY=http://user:pass@proxy.example.com:3128 export NO_PROXY=172.18.10.0,registry.internal.lan export PROXY_CA_CERTIFICATE_PATH="/home/ubuntu/.mitmproxy/mitmproxy-ca-cert.cer"
The following formats of variables are accepted:
Proxy configuration data¶ Variable
Format
HTTP_PROXYHTTPS_PROXYhttp://proxy.example.com:port- for anonymous access.http://user:password@proxy.example.com:port- for restricted access.
NO_PROXYComma-separated list of IP addresses or domain names.
PROXY_CA_CERTIFICATE_PATHOptional. Absolute path to the proxy CA certificate for man-in-the-middle (MITM) proxies. Must be placed on the bootstrap node to be trusted. For details, see Install a CA certificate for a MITM proxy on a bootstrap node.
Warning
If you require Internet access to go through a MITM proxy, ensure that the proxy has streaming enabled as described in Enable streaming for MITM.
For implementation details, see Container Cloud Reference Architecture: Proxy and cache support.
After the bootstrap cluster is set up, the
bootstrap-proxyobject is created with the provided proxy settings. You can use this object later for theClusterobject configuration.Deploy the bootstrap cluster:
./bootstrap.sh bootstrapv2Make sure that port 80 is open for
localhostto prevent security requirements for the seed node:Note
Kind uses port mapping for the master node.
telnet localhost 80
Example of a positive system response:
Connected to localhost.
Example of a negative system response:
telnet: connect to address ::1: Connection refused telnet: Unable to connect to remote host
To open port 80:
iptables -A INPUT -p tcp --dport 80 -j ACCEPT
Deploy a management cluster¶
This section contains an overview of the cluster-related objects along with the configuration procedure of these objects during deployment of a management cluster using Bootstrap v2 through the Container Cloud API.
The following procedure describes how to prepare and deploy a management
cluster using Bootstrap v2 by operating YAML templates available in the
kaas-bootstrap/templates/ folder.
To deploy a management cluster using CLI:
Export
kubeconfigof the kind cluster:export KUBECONFIG=<pathToKindKubeconfig>
By default,
<pathToKindKubeconfig>is$HOME/.kube/kind-config-clusterapi.Navigate to
kaas-bootstrap/templates/bm.Warning
The kubectl apply command automatically saves the applied data as plain text into the
kubectl.kubernetes.io/last-applied-configurationannotation of the corresponding object. This may result in revealing sensitive data in this annotation when creating or modifying objects containing credentials. Such Container Cloud objects include:BareMetalHostCredentialClusterOIDCConfiguration
LicenseProxy
ServiceUserTLSConfig
Therefore, do not use kubectl apply on these objects. Use kubectl create, kubectl patch, or kubectl edit instead.
If you used kubectl apply on these objects, you can remove the
kubectl.kubernetes.io/last-applied-configurationannotation from the objects using kubectl edit.Create the
BootstrapRegionobject by modifyingbootstrapregion.yaml.template.Configuration of bootstrapregion.yaml.template
Set
provider: baremetaland use the default<regionName>, which isregion-one.apiVersion: kaas.mirantis.com/v1alpha1 kind: BootstrapRegion metadata: name: region-one namespace: default spec: provider: baremetal
Create the object:
./kaas-bootstrap/bin/kubectl create -f \ kaas-bootstrap/templates/bm/bootstrapregion.yaml.template
Note
In the following steps, apply the changes to objects using the commands below with the required template name:
./kaas-bootstrap/bin/kubectl create -f \ kaas-bootstrap/templates/bm/<templateName>.yaml.template
Create the
ServiceUserobject by modifyingserviceusers.yaml.template.Configuration of serviceusers.yaml.template
Service user is the initial user to create in Keycloak for access to a newly deployed management cluster. By default, it has the
global-admin,operator(namespaced), andbm-pool-operator(namespaced) roles.You can delete
serviceuserafter setting up other required users with specific roles or after any integration with an external identity provider, such as LDAP.apiVersion: kaas.mirantis.com/v1alpha1 kind: ServiceUserList items: - apiVersion: kaas.mirantis.com/v1alpha1 kind: ServiceUser metadata: name: <USERNAME> spec: password: value: <PASSWORD>
Optional. Prepare any number of additional SSH keys using the following example:
apiVersion: kaas.mirantis.com/v1alpha1 kind: PublicKey metadata: name: <SSHKeyName> namespace: default spec: publicKey: | <insert your public key here>
Optional. Add the
Proxyobject using the example below:apiVersion: kaas.mirantis.com/v1alpha1 kind: Proxy metadata: name: <proxyName> namespace: default spec: ...
Inspect the default bare metal host profile definition in
baremetalhostprofiles.yaml.templateand adjust it to fit your hardware configuration. For details, see Customize the default bare metal host profile.Warning
All data will be wiped during cluster deployment on devices defined directly or indirectly in the
fileSystemslist ofBareMetalHostProfile. For example:A raw device partition with a file system on it
A device partition in a volume group with a logical volume that has a file system on it
An mdadm RAID device with a file system on it
An LVM RAID device with a file system on it
The
wipefield is always consideredtruefor these devices. Thefalsevalue is ignored.Therefore, to prevent data loss, move the necessary data from these file systems to another server beforehand, if required.
In
baremetalhosts.yaml.template, update the bare metal host definitions according to your environment configuration. Use the reference table below to manually set all parameters that start withSET_.Mandatory parameters for a bare metal host template
Parameter
Description
Example value
SET_MACHINE_0_IPMI_USERNAMEThe IPMI user name to access the BMC. 0
userSET_MACHINE_0_IPMI_PASSWORDThe IPMI password to access the BMC. 0
passwordSET_MACHINE_0_MACThe MAC address of the first master node in the PXE network.
ac:1f:6b:02:84:71SET_MACHINE_0_BMC_ADDRESSThe IP address of the BMC endpoint for the first master node in the cluster. Must be an address from the OOB network that is accessible through the management network gateway.
192.168.100.11SET_MACHINE_1_IPMI_USERNAMEThe IPMI user name to access the BMC. 0
userSET_MACHINE_1_IPMI_PASSWORDThe IPMI password to access the BMC. 0
passwordSET_MACHINE_1_MACThe MAC address of the second master node in the PXE network.
ac:1f:6b:02:84:72SET_MACHINE_1_BMC_ADDRESSThe IP address of the BMC endpoint for the second master node in the cluster. Must be an address from the OOB network that is accessible through the management network gateway.
192.168.100.12SET_MACHINE_2_IPMI_USERNAMEThe IPMI user name to access the BMC. 0
userSET_MACHINE_2_IPMI_PASSWORDThe IPMI password to access the BMC. 0
passwordSET_MACHINE_2_MACThe MAC address of the third master node in the PXE network.
ac:1f:6b:02:84:73SET_MACHINE_2_BMC_ADDRESSThe IP address of the BMC endpoint for the third master node in the cluster. Must be an address from the OOB network that is accessible through the management network gateway.
192.168.100.13Configure cluster network:
Important
Bootstrap V2 supports only separated PXE and LCM networks.
Update the network object definition in
ipam-objects.yaml.templateaccording to the environment configuration. By default, this template implies the use of separate PXE and life-cycle management (LCM) networks.Manually set all parameters that start with
SET_.To ensure successful bootstrap, enable asymmetric routing on the interfaces of the management cluster nodes. This is required because the seed node relies on one network by default, which can potentially cause traffic asymmetry.
In the
kernelParameterssection ofbaremetalhostprofiles.yaml.template, setrp_filterto2. This enables loose mode as defined in RFC3704.Example configuration of asymmetric routing
... kernelParameters: ... sysctl: # Enables the "Loose mode" for the "k8s-lcm" interface (management network) net.ipv4.conf.k8s-lcm.rp_filter: "2" # Enables the "Loose mode" for the "bond0" interface (PXE network) net.ipv4.conf.bond0.rp_filter: "2" ...
Note
More complicated solutions that are not described in this manual include getting rid of traffic asymmetry, for example:
Configure source routing on management cluster nodes.
Plug the seed node into the same networks as the management cluster nodes, which requires custom configuration of the seed node.
For configuration details of bond network interface for the PXE and management network, see Configure NIC bonding.
Example of the default L2 template snippet for a management cluster
bonds: bond0: interfaces: - {{ nic 0 }} - {{ nic 1 }} parameters: mode: active-backup primary: {{ nic 0 }} dhcp4: false dhcp6: false addresses: - {{ ip "bond0:mgmt-pxe" }} vlans: k8s-lcm: id: SET_VLAN_ID link: bond0 addresses: - {{ ip "k8s-lcm:kaas-mgmt" }} nameservers: addresses: {{ nameservers_from_subnet "kaas-mgmt" }} routes: - to: 0.0.0.0/0 via: {{ gateway_from_subnet "kaas-mgmt" }}
In this example, the following configuration applies:
A bond of two NIC interfaces
A static address in the PXE network set on the bond
An isolated L2 segment for the LCM network is configured using the
k8s-lcmVLAN with the static address in the LCM networkThe default gateway address is in the LCM network
For general concepts of configuring separate PXE and LCM networks for a management cluster, see Separate PXE and management networks. For current object templates and variable names to use, see the following tables.
Network parameters mapping overview
Deployment file name
Parameters list to update manually
ipam-objects.yaml.templateSET_LB_HOSTSET_MGMT_ADDR_RANGESET_MGMT_CIDRSET_MGMT_DNSSET_MGMT_NW_GWSET_MGMT_SVC_POOLSET_PXE_ADDR_POOLSET_PXE_ADDR_RANGESET_PXE_CIDRSET_PXE_SVC_POOLSET_VLAN_ID
bootstrap.envKAAS_BM_PXE_IPKAAS_BM_PXE_MASKKAAS_BM_PXE_BRIDGE
Mandatory network parameters of the IPAM object template
The following table contains examples of mandatory parameter values to set in
ipam-objects.yaml.templatefor the network scheme that has the following networks:172.16.59.0/24- PXE network172.16.61.0/25- LCM network
Parameter
Description
Example value
SET_PXE_CIDRThe IP address of the PXE network in the CIDR notation. The minimum recommended network size is 256 addresses (
/24prefix length).172.16.59.0/24SET_PXE_SVC_POOLThe IP address range to use for endpoints of load balancers in the PXE network for the Container Cloud services: Ironic-API, DHCP server, HTTP server, and caching server. The minimum required range size is 5 addresses.
172.16.59.6-172.16.59.15SET_PXE_ADDR_POOLThe IP address range in the PXE network to use for dynamic address allocation for hosts during inspection and provisioning.
The minimum recommended range size is 30 addresses for management cluster nodes if it is located in a separate PXE network segment. Otherwise, it depends on the number of managed cluster nodes to deploy in the same PXE network segment as the management cluster nodes.
172.16.59.51-172.16.59.200SET_PXE_ADDR_RANGEThe IP address range in the PXE network to use for static address allocation on each management cluster node. The minimum recommended range size is 6 addresses.
172.16.59.41-172.16.59.50SET_MGMT_CIDRThe IP address of the LCM network for the management cluster in the CIDR notation. If managed clusters will have their separate LCM networks, those networks must be routable to the LCM network. The minimum recommended network size is 128 addresses (
/25prefix length).172.16.61.0/25SET_MGMT_NW_GWThe default gateway address in the LCM network. This gateway must provide access to the OOB network of the Container Cloud cluster and to the Internet to download the Mirantis artifacts.
172.16.61.1SET_LB_HOSTThe IP address of the externally accessible MKE API endpoint of the cluster in the CIDR notation. This address must be within the management
SET_MGMT_CIDRnetwork but must NOT overlap with any other addresses or address ranges within this network. External load balancers are not supported.172.16.61.5/32SET_MGMT_DNSAn external (non-Kubernetes) DNS server accessible from the LCM network.
8.8.8.8SET_MGMT_ADDR_RANGEThe IP address range that includes addresses to be allocated to bare metal hosts in the LCM network for the management cluster.
When this network is shared with managed clusters, the size of this range limits the number of hosts that can be deployed in all clusters sharing this network.
When this network is solely used by a management cluster, the range must include at least 6 addresses for bare metal hosts of the management cluster.
172.16.61.30-172.16.61.40SET_MGMT_SVC_POOLThe IP address range to use for the externally accessible endpoints of load balancers in the LCM network for the Container Cloud services, such as Keycloak, web UI, and so on. The minimum required range size is 19 addresses.
172.16.61.10-172.16.61.29SET_VLAN_IDThe VLAN ID used for isolation of LCM network. The bootstrap.sh process and the seed node must have routable access to the network in this VLAN.
3975While using separate PXE and LCM networks, the management cluster services are exposed in different networks using two separate MetalLB address pools:
Services exposed through the PXE network are as follows:
Ironic API as a bare metal provisioning server
HTTP server that provides images for network boot and server provisioning
Caching server for accessing the Container Cloud artifacts deployed on hosts
Services exposed through the LCM network are all other Container Cloud services, such as Keycloak, web UI, and so on.
The default MetalLB configuration described in the
MetalLBConfigobject template ofmetallbconfig.yaml.templateuses two separate MetalLB address pools. Also, it uses theinterfacesselector in itsl2Advertisementstemplate.Caution
When you change the
L2Templateobject template inipam-objects.yaml.template, ensure that interfaces listed in theinterfacesfield of theMetalLBConfig.spec.l2Advertisementssection match those used in yourL2Template. For details about theinterfacesselector, see MetalLBConfig spec.See Configure and verify MetalLB for details on MetalLB configuration.
In
cluster.yaml.template:Set the mandatory label:
labels: kaas.mirantis.com/provider: baremetal
Update the cluster-related settings to fit your deployment.
Optional. Technology Preview. Deprecated since Container Cloud 2.29.0 (Cluster release 16.4.0). Enable WireGuard for traffic encryption on the Kubernetes workloads network.
WireGuard configuration
Ensure that the Calico MTU size is at least 60 bytes smaller than the interface MTU size of the workload network. IPv4 WireGuard uses a 60-byte header. For details, see Set the MTU size for Calico.
In
cluster.yaml.template, enable WireGuard by adding thesecureOverlayparameter:spec: ... providerSpec: value: ... secureOverlay: true
Caution
Changing this parameter on a running cluster causes a downtime that can vary depending on the cluster size.
For more details about WireGuard, see Calico documentation: Encrypt in-cluster pod traffic.
Configure StackLight. For parameters description, see StackLight configuration parameters.
Optional. Configure additional cluster settings as described in Configure optional settings.
In
machines.yaml.template:Add the following mandatory machine labels:
labels: kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: <clusterName> cluster.sigs.k8s.io/control-plane: "true"
Adjust
specandlabelssections of each entry according to your deployment.Adjust the
spec.providerSpec.value.hostSelectorvalues to matchBareMetalHostInventorycorresponding to each machine. For details, see spec:providerSpec for instance configuration.
Monitor the inspecting process of the baremetal hosts and wait until all hosts are in the
availablestate:kubectl get bmh -o go-template='{{- range .items -}} {{.status.provisioning.state}}{{"\n"}} {{- end -}}'
Example of system response:
available available available
Monitor the
BootstrapRegionobject status and wait until it is ready.kubectl get bootstrapregions -o go-template='{{(index .items 0).status.ready}}{{"\n"}}'
To obtain more granular status details, monitor
status.conditions:kubectl get bootstrapregions -o go-template='{{(index .items 0).status.conditions}}{{"\n"}}'
For a more user-friendly system response, consider using dedicated tools such as jq or yq and adjust the
-oflag to output in thejsonoryamlformat accordingly.Change the directory to
/kaas-bootstrap/.Approve the
BootstrapRegionobject to start the cluster deployment:./container-cloud bootstrap approve all
Caution
Once you approve the
BootstrapRegionobject, no cluster or machine modification is allowed.Warning
Do not manually restart or power off any of the bare metal hosts during the bootstrap process.
Monitor the deployment progress. For description of deployment stages, see Overview of the deployment workflow.
Verify that network addresses used on your clusters do not overlap with the following default MKE network addresses for Swarm and MCR:
10.0.0.0/16is used for Swarm networks. IP addresses from this network are virtual.10.99.0.0/16is used for MCR networks. IP addresses from this network are allocated on hosts.
Verification of Swarm and MCR network addresses
To verify Swarm and MCR network addresses, run on any master node:
docker infoExample of system response:
Server: ... Swarm: ... Default Address Pool: 10.0.0.0/16 SubnetSize: 24 ... Default Address Pools: Base: 10.99.0.0/16, Size: 20 ...
Not all of Swarm and MCR addresses are usually in use. One Swarm Ingress network is created by default and occupies the
10.0.0.0/24address block. Also, three MCR networks are created by default and occupy three address blocks:10.99.0.0/20,10.99.16.0/20,10.99.32.0/20.To verify the actual networks state and addresses in use, run:
docker network ls docker network inspect <networkName>
Optional. If you plan to use multiple L2 segments for provisioning of managed cluster nodes, consider the requirements specified in Configure multiple DHCP address ranges.
Configure bare metal settings¶
During creation of a bare metal management cluster using Bootstrap v2, configure several cluster settings to fit your deployment.
Note
Before update of the management cluster to Container Cloud 2.29.0
(Cluster release 16.4.0), instead of BareMetalHostInventory, use the
BareMetalHost object. For details, see BareMetalHost resource.
Caution
While the Cluster release of the management cluster is 16.4.0,
BareMetalHostInventory operations are allowed to
m:kaas@management-admin only. This limitation is lifted once the
management cluster is updated to the Cluster release 16.4.1 or later.
Before adding new BareMetalHostInventory objects, configure hardware hosts
to correctly boot them over the PXE network.
Important
Consider the following common requirements for hardware hosts configuration:
Update firmware for BIOS and Baseboard Management Controller (BMC) to the latest available version, especially if you are going to apply the UEFI configuration.
Container Cloud uses the
ipxe.efibinary loader that might be not compatible with old firmware and have vendor-related issues with UEFI booting. For example, the Supermicro issue. In this case, we recommend using thelegacybooting format.Configure all or at least the
PXENIC on switches.If the hardware host has more than one PXE NIC to boot, we strongly recommend setting up only one in the boot order. It speeds up the provisioning phase significantly.
Some hardware vendors require a host to be rebooted during BIOS configuration changes from
legacytoUEFIor vice versa for the extra option with NIC settings to appear in the menu.Connect only one Ethernet port on a host to the PXE network at any given time. Collect the physical address (MAC) of this interface and use it to configure the
BareMetalHostInventoryobject describing the host.
To configure BIOS on a bare metal host:
Enable the global BIOS mode using BIOS > Boot > boot mode select > legacy. Reboot the host if required.
Enable the
LAN-PXE-OPROMsupport using the following menus:BIOS > Advanced > PCI/PCIe Configuration > LAB OPROM TYPE > legacy
BIOS > Advanced > PCI/PCIe Configuration > Network Stack > enabled
BIOS > Advanced > PCI/PCIe Configuration > IPv4 PXE Support > enabled
Set up the configured boot order:
BIOS > Boot > Legacy-Boot-Order#1 > Hard Disk
BIOS > Boot > Legacy-Boot-Order#2 > NIC
Save changes and power off the host.
Enable the global BIOS mode using BIOS > Boot > boot mode select > UEFI. Reboot the host if required.
Enable the
LAN-PXE-OPROMsupport using the following menus:BIOS > Advanced > PCI/PCIe Configuration > LAB OPROM TYPE > uefi
BIOS > Advanced > PCI/PCIe Configuration > Network Stack > enabled
BIOS > Advanced > PCI/PCIe Configuration > IPv4 PXE Support > enabled
Note
UEFI support might not apply to all NICs. But at least built-in network interfaces should support it.
Set up the configured boot order:
BIOS > Boot > UEFI-Boot-Order#1 > UEFI Hard Disk
BIOS > Boot > UEFI-Boot-Order#2 > UEFI Network
Save changes and power off the host.
This section provides description of the bare metal host profile settings and provides instructions on how to configure this profile before deploying Mirantis Container Cloud on physical servers.
The bare metal host profile is a Kubernetes custom resource. It allows the infrastructure operator to define how the storage devices and the operating system are provisioned and configured.
The bootstrap templates for a bare metal deployment include the template for
the default BareMetalHostProfile object in the following file that defines
the default bare metal host profile:
templates/bm/baremetalhostprofiles.yaml.template
Note
Using BareMetalHostProfile, you can configure LVM or mdadm-based
software RAID support during a management or managed cluster creation. For
details, see Configure RAID support.
Warning
All data will be wiped during cluster deployment on devices
defined directly or indirectly in the fileSystems list of
BareMetalHostProfile. For example:
A raw device partition with a file system on it
A device partition in a volume group with a logical volume that has a file system on it
An mdadm RAID device with a file system on it
An LVM RAID device with a file system on it
The wipe field is always considered true for these devices.
The false value is ignored.
Therefore, to prevent data loss, move the necessary data from these file systems to another server beforehand, if required.
The customization procedure of BareMetalHostProfile is almost the same for
the management and managed clusters, with the following differences:
For a management cluster, the customization automatically applies to machines during bootstrap. And for a managed cluster, you apply the changes using kubectl before creating a managed cluster.
For a management cluster, you edit the default
baremetalhostprofiles.yaml.template. And for a managed cluster, you create a newBareMetalHostProfilewith the necessary configuration.
For the procedure details, see Create a custom bare metal host profile. Use this procedure for both types of clusters considering the differences described above.
You can configure L2 templates for the management cluster to set up a bond network interface for the PXE and management network.
This configuration must be applied to the bootstrap templates, before you run the bootstrap script to deploy the management cluster.
Configuration requirements for NIC bonding
Add at least two physical interfaces to each host in your management cluster.
Connect at least two interfaces per host to an Ethernet switch that supports Link Aggregation Control Protocol (LACP) port groups and LACP fallback.
Configure an LACP group on the ports connected to the NICs of a host.
Configure the LACP fallback on the port group to ensure that the host can boot over the PXE network before the bond interface is set up on the host operating system.
Configure server BIOS for both NICs of a bond to be PXE-enabled.
If the server does not support booting from multiple NICs, configure the port of the LACP group that is connected to the PXE-enabled NIC of a server to be the primary port. With this setting, the port becomes active in the fallback mode.
Configure the ports that connect servers to the PXE network with the PXE VLAN as native or untagged.
For reference configuration of network fabric in a baremetal-based cluster, see Container Cloud Reference Architecture: Network fabric.
To configure a bond interface that aggregates two interfaces for the PXE and management network:
In
kaas-bootstrap/templates/bm/ipam-objects.yaml.template:Verify that only the following parameters for the declaration of
{{nic 0}}and{{nic 1}}are set, as shown in the example below:dhcp4dhcp6
matchset-name
Remove other parameters.
Verify that the declaration of the bond interface
bond0has theinterfacesparameter listing both Ethernet interfaces.Verify that the node address in the PXE network (
ip "bond0:mgmt-pxe"in the below example) is bound to the bond interface or to the virtual bridge interface tied to that bond.Caution
No VLAN ID must be configured for the PXE network from the host side.
Configure bonding options using the
parametersfield. The only mandatory option ismode. See the example below for details.Note
You can set any mode supported by netplan and your hardware.
Important
Bond monitoring is disabled in Ubuntu by default. However, Mirantis highly recommends enabling it using the Media Independent Interface (MII) monitoring by setting the
mii-monitor-intervalparameter to a non-zero value. For details, see Linux documentation: bond monitoring.
Verify your configuration using the following example:
kind: L2Template metadata: name: kaas-mgmt ... spec: ... l3Layout: - subnetName: kaas-mgmt scope: namespace npTemplate: | version: 2 ethernets: {{nic 0}}: dhcp4: false dhcp6: false match: macaddress: {{mac 0}} set-name: {{nic 0}} {{nic 1}}: dhcp4: false dhcp6: false match: macaddress: {{mac 1}} set-name: {{nic 1}} bonds: bond0: interfaces: - {{nic 0}} - {{nic 1}} parameters: mode: 802.3ad mii-monitor-interval: 100 dhcp4: false dhcp6: false addresses: - {{ ip "bond0:mgmt-pxe" }} vlans: k8s-lcm: id: SET_VLAN_ID link: bond0 addresses: - {{ ip "k8s-lcm:kaas-mgmt" }} nameservers: addresses: {{ nameservers_from_subnet "kaas-mgmt" }} routes: - to: 0.0.0.0/0 via: {{ gateway_from_subnet "kaas-mgmt" }} ...
Proceed to bootstrap your management cluster as described in Deploy a management cluster using CLI.
This section describes how to configure a dedicated PXE network for a management bare metal cluster. A separate PXE network allows isolating sensitive bare metal provisioning process from the end users. The users still have access to Container Cloud services, such as Keycloak, to authenticate workloads in managed clusters, such as Horizon in a Mirantis OpenStack for Kubernetes cluster.
Note
This additional configuration procedure must be completed as part of the main Deploy a management cluster using CLI procedure. It substitutes or appends some configuration parameters and templates that are used in the main procedure for the management cluster to use two networks, PXE and management, instead of one PXE/management network. Mirantis recommends considering the main procedure first.
The following table describes the overall network mapping scheme with all
L2/L3 parameters, for example, for two networks, PXE (CIDR 10.0.0.0/24)
and management (CIDR 10.0.11.0/24):
Deployment file name |
Network |
Parameters and values |
|---|---|---|
|
Management |
|
|
PXE |
|
|
Management |
|
|
PXE |
|
When using separate PXE and management networks, the management cluster services are exposed in different networks using two separate MetalLB address pools:
Services exposed through the PXE network are as follows:
Ironic API as a bare metal provisioning server
HTTP server that provides images for network boot and server provisioning
Caching server for accessing the Container Cloud artifacts deployed on hosts
Services exposed through the management network are all other Container Cloud services, such as Keycloak, web UI, and so on.
To configure separate PXE and management networks:
Inspect guidelines to follow during configuration of the
Subnetobject as a MetalLB address pool as described MetalLB configuration guidelines for subnets.To ensure successful bootstrap, enable asymmetric routing on the interfaces of the management cluster nodes. This is required because the seed node relies on one network by default, which can potentially cause traffic asymmetry.
In the
kernelParameterssection ofbaremetalhostprofiles.yaml.template, setrp_filterto2. This enables loose mode as defined in RFC3704.Example configuration of asymmetric routing
... kernelParameters: ... sysctl: # Enables the "Loose mode" for the "k8s-lcm" interface (management network) net.ipv4.conf.k8s-lcm.rp_filter: "2" # Enables the "Loose mode" for the "bond0" interface (PXE network) net.ipv4.conf.bond0.rp_filter: "2" ...
Note
More complicated solutions that are not described in this manual include getting rid of traffic asymmetry, for example:
Configure source routing on management cluster nodes.
Plug the seed node into the same networks as the management cluster nodes, which requires custom configuration of the seed node.
In
kaas-bootstrap/templates/bm/ipam-objects.yaml.template:Substitute all
Subnetobject templates with the new ones as described in the example template belowUpdate the L2 template
spec.l3Layoutandspec.npTemplatefields as described in the example template below
Example of the Subnet object templates
# Subnet object that provides IP addresses for bare metal hosts of # management cluster in the PXE network. apiVersion: "ipam.mirantis.com/v1alpha1" kind: Subnet metadata: name: mgmt-pxe namespace: default labels: kaas.mirantis.com/provider: baremetal kaas-mgmt-pxe-subnet: "" spec: cidr: SET_IPAM_CIDR gateway: SET_PXE_NW_GW nameservers: - SET_PXE_NW_DNS includeRanges: - SET_IPAM_POOL_RANGE excludeRanges: - SET_METALLB_PXE_ADDR_POOL --- # Subnet object that provides IP addresses for bare metal hosts of # management cluster in the management network. apiVersion: "ipam.mirantis.com/v1alpha1" kind: Subnet metadata: name: mgmt-lcm namespace: default labels: kaas.mirantis.com/provider: baremetal kaas-mgmt-lcm-subnet: "" ipam/SVC-k8s-lcm: "1" ipam/SVC-ceph-cluster: "1" ipam/SVC-ceph-public: "1" cluster.sigs.k8s.io/cluster-name: CLUSTER_NAME spec: cidr: {{ SET_LCM_CIDR }} includeRanges: - {{ SET_LCM_RANGE }} excludeRanges: - SET_LB_HOST - SET_METALLB_ADDR_POOL --- # Deprecated since 2.27.0. Subnet object that provides configuration # for "services-pxe" MetalLB address pool that will be used to expose # services LB endpoints in the PXE network. apiVersion: "ipam.mirantis.com/v1alpha1" kind: Subnet metadata: name: mgmt-pxe-lb namespace: default labels: kaas.mirantis.com/provider: baremetal metallb/address-pool-name: services-pxe metallb/address-pool-protocol: layer2 metallb/address-pool-auto-assign: "false" cluster.sigs.k8s.io/cluster-name: CLUSTER_NAME spec: cidr: SET_IPAM_CIDR includeRanges: - SET_METALLB_PXE_ADDR_POOL
Example of the L2 template spec
kind: L2Template ... spec: ... l3Layout: - scope: namespace subnetName: kaas-mgmt-pxe labelSelector: kaas.mirantis.com/provider: baremetal kaas-mgmt-pxe-subnet: "" - scope: namespace subnetName: kaas-mgmt-lcm labelSelector: kaas.mirantis.com/provider: baremetal kaas-mgmt-lcm-subnet: "" npTemplate: | version: 2 renderer: networkd ethernets: {{nic 0}}: dhcp4: false dhcp6: false match: macaddress: {{mac 0}} set-name: {{nic 0}} {{nic 1}}: dhcp4: false dhcp6: false match: macaddress: {{mac 1}} set-name: {{nic 1}} bridges: bm-pxe: interfaces: - {{ nic 0 }} dhcp4: false dhcp6: false addresses: - {{ ip "bm-pxe:kaas-mgmt-pxe" }} nameservers: addresses: {{ nameservers_from_subnet "kaas-mgmt-pxe" }} routes: - to: 0.0.0.0/0 via: {{ gateway_from_subnet "kaas-mgmt-pxe" }} k8s-lcm: interfaces: - {{ nic 1 }} dhcp4: false dhcp6: false addresses: - {{ ip "k8s-lcm:kaas-mgmt-lcm" }} nameservers: addresses: {{ nameservers_from_subnet "kaas-mgmt-lcm" }}
Deprecated since Container Cloud 2.27.0 (Cluster releases 17.2.0 and 16.2.0): the last
Subnettemplate namedmgmt-pxe-lbin the example above will be used to configure the MetalLB address pool in the PXE network. The bare metal provider will automatically configure MetalLB with address pools using theSubnetobjects identified by specific labels.Warning
The
bm-pxeaddress must have a separate interface with only one address on this interface.Verify the current MetalLB configuration that is stored in
MetalLBobjects:kubectl -n metallb-system get ipaddresspools,l2advertisements
For the example configuration described above, the system outputs a similar content:
NAME AGE ipaddresspool.metallb.io/default 129m ipaddresspool.metallb.io/services-pxe 129m NAME AGE l2advertisement.metallb.io/default 129m l2advertisement.metallb.io/services-pxe 129m
To verify the
MetalLBobjects:kubectl -n metallb-system get <object> -o json | jq '.spec'
For the example configuration described above, the system outputs a similar content for
ipaddresspoolobjects:{ "addresses": [ "10.0.11.61-10.0.11.80" ], "autoAssign": true, "avoidBuggyIPs": false } $ kubectl -n metallb-system get ipaddresspool.metallb.io/services-pxe -o json | jq '.spec' { "addresses": [ "10.0.0.61-10.0.0.70" ], "autoAssign": false, "avoidBuggyIPs": false }
The
auto-assignparameter will be set tofalsefor all address pools except thedefaultone. So, a particular service will get an address from such an address pool only if theServiceobject has a specialmetallb.universe.tf/address-poolannotation that points to the specific address pool name.Note
- It is expected that every Container Cloud service on a management
cluster will be assigned to one of the address pools. Current consideration is to have two MetalLB address pools:
services-pxeis a reserved address pool name to use for the Container Cloud services in the PXE network (Ironic API, HTTP server, caching server).The bootstrap cluster also uses the
services-pxeaddress pool for its provision services for management cluster nodes to be provisioned from the bootstrap cluster. After the management cluster is deployed, the bootstrap cluster is deleted and that address pool is solely used by the newly deployed cluster.defaultis an address pool to use for all other Container Cloud services in the management network. No annotation is required on theServiceobjects in this case.
In addition to the network parameters defined in Deploy a management cluster using CLI, configure the following ones by replacing them in
templates/bm/ipam-objects.yaml.template:New subnet template parameters¶ Parameter
Description
Example value
SET_LCM_CIDRAddress of a management network for the management cluster in the CIDR notation. You can later share this network with managed clusters where it will act as the LCM network. If managed clusters have their separate LCM networks, those networks must be routable to the management network.
10.0.11.0/24SET_LCM_RANGEAddress range that includes addresses to be allocated to bare metal hosts in the management network for the management cluster. When this network is shared with managed clusters, the size of this range limits the number of hosts that can be deployed in all clusters that share this network. When this network is solely used by a management cluster, the range should include at least 3 IP addresses for bare metal hosts of the management cluster.
10.0.11.100-10.0.11.109SET_METALLB_PXE_ADDR_POOLAddress range to be used for LB endpoints of the Container Cloud services: Ironic-API, HTTP server, and caching server. This range must be within the PXE network. The minimum required range is 5 IP addresses.
10.0.0.61-10.0.0.70The following parameters will now be tied to the management network while their meaning remains the same as described in Deploy a management cluster using CLI:
Subnet template parameters migrated to management network¶ Parameter
Description
Example value
SET_LB_HOSTIP address of the externally accessible API endpoint of the management cluster. This address must NOT be within the
SET_METALLB_ADDR_POOLrange but within the management network. External load balancers are not supported.10.0.11.90SET_METALLB_ADDR_POOLThe address range to be used for the externally accessible LB endpoints of the Container Cloud services, such as Keycloak, web UI, and so on. This range must be within the management network. The minimum required range is 19 IP addresses.
10.0.11.61-10.0.11.80Proceed to further steps in Deploy a management cluster using CLI.
To facilitate multi-rack and other types of distributed bare metal datacenter topologies, the dnsmasq DHCP server used for host provisioning in Container Cloud supports working with multiple L2 segments through network routers that support DHCP relay.
Container Cloud has its own DHCP relay running on one of the management cluster nodes. That DHCP relay serves for proxying DHCP requests in the same L2 domain where the management cluster nodes are located.
Caution
Networks used for hosts provisioning of a managed cluster must have routes to the PXE network of the management cluster. This configuration enables hosts to have access to the management cluster services that are used during host provisioning.
Management cluster nodes must have routes through the PXE network to PXE network segments used on a managed cluster. The following example contains L2 template fragments for a management cluster node:
Configuration example extract
l3Layout:
# PXE/static subnet for a management cluster
- scope: namespace
subnetName: kaas-mgmt-pxe
labelSelector:
kaas-mgmt-pxe-subnet: "1"
# management (LCM) subnet for a management cluster
- scope: namespace
subnetName: kaas-mgmt-lcm
labelSelector:
kaas-mgmt-lcm-subnet: "1"
# PXE/dhcp subnets for a managed cluster
- scope: namespace
subnetName: managed-dhcp-rack-1
- scope: namespace
subnetName: managed-dhcp-rack-2
- scope: namespace
subnetName: managed-dhcp-rack-3
...
npTemplate: |
...
bonds:
bond0:
interfaces:
- {{ nic 0 }}
- {{ nic 1 }}
parameters:
mode: active-backup
primary: {{ nic 0 }}
mii-monitor-interval: 100
dhcp4: false
dhcp6: false
addresses:
# static address on management node in the PXE network
- {{ ip "bond0:kaas-mgmt-pxe" }}
routes:
# routes to managed PXE network segments
- to: {{ cidr_from_subnet "managed-dhcp-rack-1" }}
via: {{ gateway_from_subnet "kaas-mgmt-pxe" }}
- to: {{ cidr_from_subnet "managed-dhcp-rack-2" }}
via: {{ gateway_from_subnet "kaas-mgmt-pxe" }}
- to: {{ cidr_from_subnet "managed-dhcp-rack-3" }}
via: {{ gateway_from_subnet "kaas-mgmt-pxe" }}
...
To configure DHCP ranges for dnsmasq, create the Subnet objects
tagged with the ipam/SVC-dhcp-range label while setting up subnets
for a managed cluster using CLI.
Caution
Support of multiple DHCP ranges has the following limitations:
Using of custom DNS server addresses for servers that boot over PXE is not supported.
The
Subnetobjects for DHCP ranges cannot be associated with any specific cluster, as DHCP server configuration is only applicable to the management cluster where DHCP server is running. Thecluster.sigs.k8s.io/cluster-namelabel will be ignored.
Create the
Subnetobjects tagged with theipam/SVC-dhcp-rangelabel.Caution
For cluster-specific subnets, create
Subnetobjects in the same namespace as the relatedClusterobject project. For shared subnets, createSubnetobjects in thedefaultnamespace.To create the
Subnetobjects, refer to Create subnets.Use the following
Subnetobject example to specify DHCP ranges and DHCP options to pass the default route address:apiVersion: "ipam.mirantis.com/v1alpha1" kind: Subnet metadata: name: mgmt-dhcp-range namespace: default labels: ipam/SVC-dhcp-range: "" kaas.mirantis.com/provider: baremetal spec: cidr: 10.11.0.0/24 gateway: 10.11.0.1 includeRanges: - 10.11.0.121-10.11.0.125 - 10.11.0.191-10.11.0.199
Note
Setting of custom nameservers in the DHCP subnet is not supported.
After creation of the above
Subnetobject, the provided data will be utilized to render theDnsmasqobject used for configuration of the dnsmasq deployment. You do not have to manually edit theDnsmasqobject.Verify that the changes are applied to the
Dnsmasqobject:kubectl --kubeconfig <pathToMgmtClusterKubeconfig> \ -n kaas get dnsmasq dnsmasq-dynamic-config -o json
For servers to access the DHCP server across the L2 segment boundaries, for example, from another rack with a different VLAN for PXE network, you must configure DHCP relay (agent) service on the border switch of the segment. For example, on a top-of-rack (ToR) or leaf (distribution) switch, depending on the data center network topology.
Warning
To ensure predictable routing for the relay of DHCP packets, Mirantis strongly advises against the use of chained DHCP relay configurations. This precaution limits the number of hops for DHCP packets, with an optimal scenario being a single hop.
This approach is justified by the unpredictable nature of chained relay configurations and potential incompatibilities between software and hardware relay implementations.
The dnsmasq server listens on the PXE network of the management
cluster by using the dhcp-lb Kubernetes Service.
To configure the DHCP relay service, specify the external address of the
dhcp-lb Kubernetes Service as an upstream address for the relayed DHCP
requests, which is the IP helper address for DHCP. There is the dnsmasq
deployment behind this service that can only accept relayed DHCP requests.
Container Cloud has its own DHCP relay running on one of the management cluster nodes. That DHCP relay serves for proxying DHCP requests in the same L2 domain where the management cluster nodes are located.
To obtain the actual IP address issued to the dhcp-lb Kubernetes Service:
kubectl -n kaas get service dhcp-lb
Note
This section applies only to existing management clusters that are created before Container Cloud 2.24.0 (Cluster release 14.0.0).
Caution
Since Container Cloud 2.24.0, you can only remove the deprecated
dnsmasq.dhcp_range, dnsmasq.dhcp_ranges, dnsmasq.dhcp_routers,
and dnsmasq.dhcp_dns_servers values from the cluster spec.
The Admission Controller does not accept any other changes in these values.
This configuration is completely superseded by the Subnet object.
The DHCP configuration automatically migrated from the cluster spec to
Subnet objects after cluster upgrade to Container Cloud 2.21.0 (Cluster
release 11.5.0).
To remove the deprecated dnsmasq parameters from the cluster spec:
Open the management cluster
specfor editing.In the
baremetal-operatorrelease values, remove thednsmasq.dhcp_range,dnsmasq.dhcp_ranges,dnsmasq.dhcp_routers, anddnsmasq.dhcp_dns_serversparameters. For example:regional: - helmReleases: - name: baremetal-operator values: dnsmasq: dhcp_range: 10.204.1.0,10.204.5.255,255.255.255.0
Caution
The
dnsmasq.dhcp_<name>parameters of thebaremetal-operatorHelm chart values in theClusterspecare deprecated since the Cluster release 11.5.0 and removed in the Cluster release 14.0.0.Ensure that the required DHCP ranges and options are set in the
Subnetobjects. For configuration details, see Configure DHCP ranges for dnsmasq.
The dnsmasq configuration options dhcp-option=3 and dhcp-option=6
are absent in the default configuration. So, by default, dnsmasq
will send the DNS server and default route to DHCP clients as defined in the
dnsmasq official documentation:
The netmask and broadcast address are the same as on the host running dnsmasq.
The DNS server and default route are set to the address of the host running dnsmasq.
If the domain name option is set, this name is sent to DHCP clients.
Available since MCC 2.26.0 (Cluster release 16.1.0)
This section instructs you on how to enable dynamic IP allocation feature to increase the amount of baremetal hosts to be provisioned in parallel on managed clusters.
Using this feature, you can effortlessly deploy a large managed cluster by provisioning up to 100 hosts simultaneously. In addition to dynamic IP allocation, this feature disables the ping check in the DHCP server. Therefore, if you plan to deploy large managed clusters, enable this feature during the management cluster bootstrap.
Caution
Before using this feature, familiarize yourself with Container Cloud Reference Architecture: DHCP range requirements for PXE.
To enable dynamic IP allocation for large managed clusters:
In the Cluster object of the management cluster, modify the configuration
of baremetal-operator by setting dynamic_bootp to true:
spec:
...
providerSpec:
value:
kaas:
...
regional:
- helmReleases:
- name: baremetal-operator
values:
dnsmasq:
dynamic_bootp: true
provider: baremetal
...
Available since MCC 2.25.0 (Cluster release 16.0.0)
This section instructs you on how to set a custom external IP address for
the dhcp-lb service so that it remains the same during management cluster
upgrades and other LCM operations.
The changes of dhcp-lb service address may lead to the necessity of
changing configuration for DHCP relays on ToR switches.
The described procedure allows you to avoid such unwanted changes.
This configuration makes sense when you use multiple DHCP address ranges
on your deployment. See Configure multiple DHCP address ranges for details.
To set a custom external IP address for the dhcp-lb service:
In the
Clusterobject of the management cluster, modify the configuration of thebaremetal-operatorrelease by settingdnsmasq.dedicated_udp_service_address_pooltotrue:spec: ... providerSpec: value: kaas: ... regional: - helmReleases: ... - name: baremetal-operator values: dnsmasq: dedicated_udp_service_address_pool: true ... provider: baremetal ...
In the
MetalLBConfigobject of the management cluster, modify theipAddressPoolsobject list by adding thedhcp-lbobject and theserviceAllocationparameters for thedefaultobject:ipAddressPools: - name: default spec: addresses: - 112.181.11.41-112.181.11.60 autoAssign: true avoidBuggyIPs: false serviceAllocation: serviceSelectors: - matchExpressions: - key: app.kubernetes.io/name operator: NotIn values: - dhcp-lb - name: services-pxe spec: addresses: - 10.0.24.122-10.0.24.140 autoAssign: false avoidBuggyIPs: false - name: dhcp-lb spec: addresses: - 10.0.24.121/32 autoAssign: true avoidBuggyIPs: false serviceAllocation: namespaces: - kaas serviceSelectors: - matchExpressions: - key: app.kubernetes.io/name operator: In values: - dhcp-lb
Select non-overlapping IP addresses for all the
ipAddressPoolsthat you use:default,services-pxe, anddhcp-lb.In the
MetalLBConfigobject of the management cluster, modify thel2Advertisementsobject list by addingdhcp-lbto theipAddressPoolssection in thepxeobject spec:Note
A cluster may have a different
L2Advertisementobject name instead ofpxe.l2Advertisements: ... - name: pxe spec: ipAddressPools: - services-pxe - dhcp-lb ...
Configure optional settings¶
Note
Consider this section as part of the Bootstrap v2 CLI procedure.
During creation of a management cluster, you can configure optional cluster
settings using the Container Cloud API by modifying cluster.yaml.template.
To configure optional cluster settings:
Technology Preview. Enable custom host names for cluster machines. When enabled, any machine host name in a particular region matches the related
Machineobject name. For example, instead of the defaultkaas-node-<UID>, a machine host name will bemaster-0. The custom naming format is more convenient and easier to operate with.Configuration for custom host names on the management and its future managed clusters
In
cluster.yaml.template, find thespec.providerSpec.value.kaas.regional.helmReleases.name: baremetal-providersection.Under
values.config, addcustomHostnamesEnabled: true:regional: - helmReleases: - name: baremetal-provider values: config: allInOneAllowed: false customHostnamesEnabled: true internalLoadBalancers: false provider: baremetal-provider
Optional. Technology Preview. Enable the Linux Audit daemon auditd to monitor activity of cluster processes and prevent potential malicious activity.
Configuration for auditd
In the
Clusterobject orcluster.yaml.template, add the auditd parameters:spec: providerSpec: value: audit: auditd: enabled: <bool> enabledAtBoot: <bool> backlogLimit: <int> maxLogFile: <int> maxLogFileAction: <string> maxLogFileKeep: <int> mayHaltSystem: <bool> presetRules: <string> customRules: <string> customRulesX32: <text> customRulesX64: <text>
Configuration parameters for auditd:
enabledBoolean, default -
false. Enables theauditdrole to install the auditd packages and configure rules. CIS rules: 4.1.1.1, 4.1.1.2.enabledAtBootBoolean, default -
false. Configures grub to audit processes that can be audited even if they start up prior to auditd startup. CIS rule: 4.1.1.3.backlogLimitInteger, default - none. Configures the backlog to hold records. If during boot
audit=1is configured, the backlog holds 64 records. If more than 64 records are created during boot, auditd records will be lost with a potential malicious activity being undetected. CIS rule: 4.1.1.4.maxLogFileInteger, default - none. Configures the maximum size of the audit log file. Once the log reaches the maximum size, it is rotated and a new log file is created. CIS rule: 4.1.2.1.
maxLogFileActionString, default - none. Defines handling of the audit log file reaching the maximum file size. Allowed values:
keep_logs- rotate logs but never delete themrotate- add a cron job to compress rotated log files and keep maximum 5 compressed files.compress- compress log files and keep them under the/var/log/auditd/directory. Requiresauditd_max_log_file_keepto be enabled.
CIS rule: 4.1.2.2.
maxLogFileKeepInteger, default -
5. Defines the number of compressed log files to keep under the/var/log/auditd/directory. Requiresauditd_max_log_file_action=compress. CIS rules - none.mayHaltSystemBoolean, default -
false. Halts the system when the audit logs are full. Applies the following configuration:space_left_action = emailaction_mail_acct = rootadmin_space_left_action = halt
CIS rule: 4.1.2.3.
customRulesString, default - none. Base64-encoded content of the
60-custom.rulesfile for any architecture. CIS rules - none.customRulesX32String, default - none. Base64-encoded content of the
60-custom.rulesfile for thei386architecture. CIS rules - none.customRulesX64String, default - none. Base64-encoded content of the
60-custom.rulesfile for thex86_64architecture. CIS rules - none.presetRulesString, default - none. Comma-separated list of the following built-in preset rules:
accessactionsdeletedocker
identityimmutableloginsmac-policy
modulesmountsperm-modprivileged
scopesessionsystem-localetime-change
Since Container Cloud 2.28.0 (Cluster releases 17.3.0 and 16.3.0) in the Technology Preview scope, you can collect some of the preset rules indicated above as groups and use them in
presetRules:ubuntu-cis-rules- this group contains rules to comply with the Ubuntu CIS Benchmark recommendations, including the following CIS Ubuntu 20.04 v2.0.1 rules:scope- 5.2.3.1actions- same as 5.2.3.2time-change- 5.2.3.4system-locale- 5.2.3.5privileged- 5.2.3.6access- 5.2.3.7identity- 5.2.3.8
perm-mod- 5.2.3.9mounts- 5.2.3.10session- 5.2.3.11logins- 5.2.3.12delete- 5.2.3.13mac-policy- 5.2.3.14modules- 5.2.3.19
docker-cis-rules- this group contains rules to comply with Docker CIS Benchmark recommendations, including thedockerDocker CIS v1.6.0 rules 1.1.3 - 1.1.18.
You can also use two additional keywords inside
presetRules:none- select no built-in rules.all- select all built-in rules. When using this keyword, you can add the!prefix to a rule name to exclude some rules. You can use the!prefix for rules only if you add theallkeyword as the first rule. Place a rule with the!prefix only after theallkeyword.
Example configurations:
presetRules: none- disable all preset rulespresetRules: docker- enable only thedockerrulespresetRules: access,actions,logins- enable only theaccess,actions, andloginsrulespresetRules: ubuntu-cis-rules- enable all rules from theubuntu-cis-rulesgrouppresetRules: docker-cis-rules,actions- enable all rules from thedocker-cis-rulesgroup and theactionsrulepresetRules: all- enable all preset rulespresetRules: all,!immutable,!sessions- enable all preset rules exceptimmutableandsessions
CIS controls
4.1.3 (time-change)4.1.4 (identity)4.1.5 (system-locale)4.1.6 (mac-policy)4.1.7 (logins)4.1.8 (session)4.1.9 (perm-mod)4.1.10 (access)4.1.11 (privileged)4.1.12 (mounts)4.1.13 (delete)4.1.14 (scope)4.1.15 (actions)4.1.16 (modules)4.1.17 (immutable)Docker CIS controls
1.1.41.1.81.1.101.1.121.1.131.1.151.1.161.1.171.1.181.2.31.2.41.2.51.2.61.2.71.2.101.2.11
Configure OIDC integration with LDAP or Google OAuth. For details, see Configure LDAP for IAM or Configure Google OAuth IdP for IAM.
Configure NTP server. You can disable NTP that is enabled by default. This option disables the management of
chronyconfiguration by MOSK to use your own system forchronymanagement. Otherwise, configure the regional NTP server parameters as described below.NTP configuration
Configure the regional NTP server parameters to be applied to all machines of managed clusters.
In
cluster.yaml.templateor theClusterobject, add thentp:serverssection with the list of required server names:spec: ... providerSpec: value: kaas: ... ntpEnabled: true regional: - helmReleases: - name: baremetal-provider values: config: lcm: ... ntp: servers: - 0.pool.ntp.org ... provider: baremetal ...
To disable NTP:
spec: ... providerSpec: value: ... ntpEnabled: false ...
Applies since Container Cloud 2.26.0 (Cluster release 16.1.0). If you plan to deploy large managed clusters, enable dynamic IP allocation to increase the amount of baremetal hosts to be provisioned in parallel. For details, see Enable dynamic IP allocation.
Now, proceed with completing the bootstrap process using the Container Cloud Bootstrap API as described in Deploy a management cluster.
Post-deployment steps¶
After bootstrapping the management cluster, collect and save the following cluster details in a secure location:
Obtain the management cluster
kubeconfig:./container-cloud get cluster-kubeconfig \ --kubeconfig <pathToKindKubeconfig> \ --cluster-name <clusterName>
By default,
pathToKindKubeconfigis$HOME/.kube/kind-config-clusterapi.Obtain the Keycloak credentials as described in Access the Keycloak Admin Console.
Remove the kind cluster:
./bin/kind delete cluster -n <kindClusterName>
By default,
kindClusterNameisclusterapi.
Now, you can proceed with operating your management cluster through the Container Cloud web UI and deploying MOSK clusters as described in Operations Guide.
Create initial users after a management cluster bootstrap¶
Once you bootstrap your management cluster, create Keycloak users for access to the Container Cloud web UI.
Mirantis recommends creating at least two users, user and operator,
that are required for a typical MOSK deployment.
To create the user for access to the Container Cloud web UI:
./container-cloud bootstrap user add \
--username <userName> \
--roles <roleName> \
--kubeconfig <pathToMgmtKubeconfig>
Note
You will be asked for the user password interactively.
Flag |
Description |
|---|---|
|
Required. Name of the user to create. |
|
Required. Comma-separated list of roles to assign to the user.
|
|
Required. Path to the management cluster |
|
Optional. Name of the Container Cloud project where the user will be created. If not set, a global user will be created for all Container Cloud projects with the corresponding role access to view or manage all public objects. |
|
Optional. Flag to provide the user password through echo '$PASSWORD' | ./container-cloud bootstrap user add \
--username <userName> \
--roles <roleName> \
--kubeconfig <pathToMgmtKubeconfig> \
--password-stdin
|
To delete the user:
./container-cloud bootstrap user delete --username <userName> --kubeconfig <pathToMgmtKubeconfig>
Requirements for a MITM proxy¶
Note
For MOSK clusters, the feature is generally available since MOSK 23.1.
While bootstrapping a Container Cloud management cluster using proxy, you may require Internet access to go through a man-in-the-middle (MITM) proxy. Such configuration requires that you enable streaming and install a CA certificate on a bootstrap node.
Enable streaming for MITM¶
Ensure that the MITM proxy is configured with enabled streaming. For example, if you use mitmproxy, enable the stream_large_bodies=1 option:
./mitmdump --set stream_large_bodies=1
Install a CA certificate for a MITM proxy on a bootstrap node¶
Log in to the bootstrap node.
Install
ca-certificates:apt install ca-certificates
Copy your CA certificate to the
/usr/local/share/ca-certificates/directory. For example:sudo cp ~/.mitmproxy/mitmproxy-ca-cert.cer /usr/local/share/ca-certificates/mitmproxy-ca-cert.crt
Replace
~/.mitmproxy/mitmproxy-ca-cert.cerwith the path to your CA certificate.Caution
The target CA certificate file must be in the
PEMformat with the.crtextension.Apply the changes:
sudo update-ca-certificates
Now, proceed with bootstrapping your management cluster.
Configure external identity provider for IAM¶
This section describes how to configure authentication for management cluster depending on the external identity provider type integrated to your deployment.
Configure LDAP for IAM¶
If you integrate LDAP for IAM to Mirantis OpenStack for Kubernetes, add the required LDAP
configuration to cluster.yaml.template during the management cluster
bootstrap.
Note
The example below defines the recommended non-anonymous
authentication type. If you require anonymous authentication, replace the
following parameters with authType: "none":
authType: "simple"
bindCredential: ""
bindDn: ""
To configure LDAP for IAM:
Open
templates/bm/cluster.yaml.template.Configure the
keycloak:userFederation:providers:andkeycloak:userFederation:mappers:sections as required:spec: providerSpec: value: kaas: management: helmReleases: - name: iam values: keycloak: userFederation: providers: - displayName: "<LDAP_NAME>" providerName: "ldap" priority: 1 fullSyncPeriod: -1 changedSyncPeriod: -1 config: pagination: "true" debug: "false" searchScope: "1" connectionPooling: "true" usersDn: "<DN>" # "ou=People, o=<ORGANIZATION>, dc=<DOMAIN_COMPONENT>" userObjectClasses: "inetOrgPerson,organizationalPerson" usernameLDAPAttribute: "uid" rdnLDAPAttribute: "uid" vendor: "ad" editMode: "READ_ONLY" uuidLDAPAttribute: "uid" connectionUrl: "ldap://<LDAP_DNS>" syncRegistrations: "false" authType: "simple" bindCredential: "" bindDn: "" mappers: - name: "username" federationMapperType: "user-attribute-ldap-mapper" federationProviderDisplayName: "<LDAP_NAME>" config: ldap.attribute: "uid" user.model.attribute: "username" is.mandatory.in.ldap: "true" read.only: "true" always.read.value.from.ldap: "false" - name: "full name" federationMapperType: "full-name-ldap-mapper" federationProviderDisplayName: "<LDAP_NAME>" config: ldap.full.name.attribute: "cn" read.only: "true" write.only: "false" - name: "last name" federationMapperType: "user-attribute-ldap-mapper" federationProviderDisplayName: "<LDAP_NAME>" config: ldap.attribute: "sn" user.model.attribute: "lastName" is.mandatory.in.ldap: "true" read.only: "true" always.read.value.from.ldap: "true" - name: "email" federationMapperType: "user-attribute-ldap-mapper" federationProviderDisplayName: "<LDAP_NAME>" config: ldap.attribute: "mail" user.model.attribute: "email" is.mandatory.in.ldap: "false" read.only: "true" always.read.value.from.ldap: "true"
Verify that the
userFederationsection is located on the same level as theinitUserssection.Verify that all attributes set in the
mapperssection are defined for users in the specified LDAP system. Missing attributes may cause authorization issues.
Now, return to the bootstrap instruction for your management cluster.
Configure Google OAuth IdP for IAM¶
Caution
The instruction below applies to the DNS-based management clusters. If you bootstrap a non-DNS-based management cluster, configure Google OAuth IdP for Keycloak after bootstrap using the official Keycloak documentation.
If you integrate Google OAuth external identity provider for IAM to
Mirantis OpenStack for Kubernetes, create the authorization credentials for IAM in your
Google OAuth account and configure cluster.yaml.template during the
bootstrap of the management cluster.
To configure Google OAuth IdP for IAM:
Create Google OAuth credentials for IAM:
Log in to your https://console.developers.google.com.
Navigate to Credentials.
In the APIs Credentials menu, select OAuth client ID.
In the window that opens:
In the Application type menu, select Web application.
In the Authorized redirect URIs field, type in
<keycloak-url>/auth/realms/iam/broker/google/endpoint, where<keycloak-url>is the corresponding DNS address.Press Enter to add the URI.
Click Create.
A page with your client ID and client secret opens. Save these credentials for further usage.
Log in to the bootstrap node.
Open
templates/bm/cluster.yaml.template.In the
keycloak:externalIdP:section, add the following snippet with your credentials created in previous steps:keycloak: externalIdP: google: enabled: true config: clientId: <Google_OAuth_client_ID> clientSecret: <Google_OAuth_client_secret>
Now, return to the bootstrap instruction for your management cluster.
Create a managed cluster¶
After bootstrapping your baremetal-based Mirantis Container Cloud management cluster, you can create a baremetal-based managed cluster to deploy Mirantis OpenStack for Kubernetes using the Container Cloud API.
Create a project for MOSK clusters¶
Note
The procedure below applies only to the Container Cloud web UI
users with the m:kaas@global-admin or m:kaas@writer access role
assigned by the infrastructure operator.
The default project (Kubernetes namespace) is dedicated for management
clusters only. MOSK clusters require a separate project.
You can create as many projects as required by your company infrastructure.
To create a project for MOSK clusters:
Log in to the Container Cloud web UI as
m:kaas@global-adminorm:kaas@writer.In the Projects tab, click Create.
Type the new project name.
Click Create.
Note
Due to the known issue 50168, access to the newly created project becomes available in five minutes after project creation.
Add a bare metal host¶
Before creating a bare metal managed cluster, add the required number of bare metal hosts either using the Container Cloud web UI for a default configuration or using CLI for an advanced configuration.
Add a bare metal host using web UI¶
This section describes how to add bare metal hosts using the Container Cloud web UI during a MOSK cluster creation.
Before you proceed with adding a bare metal host:
Verify that the physical network on the server has been configured correctly. For details, see Physical networks layout and Container Cloud Reference Architecture: Network fabric.
Prepare the host as described in Configure BIOS on a bare metal host.
To add a bare metal host to a MOSK cluster:
Optional. Create a custom bare metal host profile depending on your needs as described in Create a custom bare metal host profile.
Note
You can view the created profiles in the BM Host Profiles tab of the Container Cloud web UI.
Log in to the Container Cloud web UI with the
m:kaas@operatororm:kaas:namespace@bm-pool-operatorpermissions.Switch to the required non-
defaultproject using the Switch Project action icon located on top of the main left-side navigation panel.Caution
Do not create a MOSK cluster in the
defaultproject (Kubernetes namespace), which is dedicated for the management cluster only. If no projects are defined, first create a newmoskproject as described in Create a project for MOSK clusters.Optional. Available since Container Cloud 2.24.0 (Cluster releases 15.0.1 and 14.0.1). In the Credentials tab, click Add Credential and add the IPMI user name and password of the bare metal host to access the Baseboard Management Controller (BMC).
Select one of the following options:
In the Baremetal tab, click Create Host.
Fill out the Create baremetal host form as required:
- Name
Specify the name of the new bare metal host.
- Boot Mode
Specify the BIOS boot mode. Available options: Legacy, UEFI, or UEFISecureBoot.
- MAC Address
Specify the MAC address of the PXE network interface.
- Baseboard Management Controller (BMC)
Specify the following BMC details:
- IP Address
Specify the IP address to access the BMC.
- Credential Name
Specify the name of the previously added bare metal host credentials to associate with the current host.
- Cert Validation
Enable validation of the BMC API certificate. Applies only to the
redfish+httpBMC protocol. Disabled by default.
- Power off host after creation
Experimental. Select to power off the bare metal host after creation.
Caution
This option is experimental and intended only for testing and evaluation purposes. Do not use it for production deployments.
In the Baremetal tab, click Add BM host.
Fill out the Add new BM host form as required:
- Baremetal host name
Specify the name of the new bare metal host.
- Provider Credential
Optional. Available since Container Cloud 2.24.0 (Cluster releases 15.0.1 and 14.0.1). Specify the name of the previously added bare metal host credentials to associate with the current host.
- Add New Credential
Optional. Available since Container Cloud 2.24.0 (Cluster releases 15.0.1 and 14.0.1). Applies if you did not add bare metal host credentials using the Credentials tab. Add the bare metal host credentials:
- Username
Specify the name of the IPMI user to access the BMC.
- Password
Specify the IPMI password of the user to access the BMC.
- Boot MAC address
Specify the MAC address of the PXE network interface.
- IP Address
Specify the IP address to access the BMC.
- Label
Assign the machine label to the new host that defines which type of machine may be deployed on this bare metal host. Only one label can be assigned to a host. The supported labels include:
- Manager
This label is selected and set by default. Assign this label t the bare metal hosts that can be used to deploy machines with the
managertype. These hosts must match the CPU and RAM requirements described in MOSK cluster hardware requirements.
- Worker
The host with this label may be used to deploy the
workermachine type. Assign this label to the bare metal hosts that have sufficient CPU and RAM resources, as described in MOSK cluster hardware requirements.
- Storage
Assign this label to the bare metal hosts that have sufficient storage devices to match MOSK cluster hardware requirements. Hosts with this label will be used to deploy machines with the storage type that run Ceph OSDs.
Click Create.
While adding the bare metal host, Container Cloud discovers and inspects the hardware of the bare metal host and adds it to
BareMetalHost.statusfor future references.During provisioning,
baremetal-operatorinspects the bare metal host and moves it to thePreparingstate. The host becomes ready to be linked to a bare metal machine.Verify the results of the hardware inspection to avoid unexpected errors during the host usage:
Select one of the following options:
In the left sidebar, click Baremetal. The Hosts page opens.
In the left sidebar, click BM Hosts.
Verify that the bare metal host is registered and switched to one of the following statuses:
Preparing for a newly added host
Ready for a previously used host or for a host that is already linked to a machine
Select one of the following options:
On the Hosts page, click the host kebab menu and select Host info.
On the BM Hosts page, click the name of the newly added bare metal host.
In the window with the host details, scroll down to the Hardware section.
Review the section and make sure that the number and models of disks, network interface cards, and CPUs match the hardware specification of the server.
If the hardware details are consistent with the physical server specifications for all your hosts, proceed to Create a MOSK cluster.
If you find any discrepancies in the hardware inspection results, it might indicate that the server has hardware issues or is not compatible with Container Cloud.
Add a bare metal host using CLI¶
This section describes how to add bare metal hosts using the Container Cloud CLI during a managed cluster creation.
To add a bare metal host:
Log in to the host where your management cluster
kubeconfigis located and where kubectl is installed.Describe the unique credentials of the new bare metal host:
Create a YAML file that describes the unique credentials of the new bare metal host as a
BareMetalHostCredentialobject.apiVersion: kaas.mirantis.com/v1alpha1 kind: BareMetalHostCredential metadata: labels: kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: <bare-metal-host-credential-unique-name> namespace: <managed-cluster-project-name> spec: username: <ipmi-user-name> password: value: <ipmi-user-password>
In the
metadatasection, add a unique credentialsnameand the name of the non-defaultproject (namespace) dedicated for the managed cluster being created.In the
specsection, add the IPMI user name and password in plain text to access the Baseboard Management Controller (BMC). The password will not be stored in theBareMetalHostCredentialobject but will be erased and saved in an underlyingSecretobject.
Caution
Each bare metal host must have a unique
BareMetalHostCredential. For details about theBareMetalHostCredentialobject, refer to BareMetalHostCredential resource.Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.Create a secret YAML file that describes the unique credentials of the new bare metal host. Example of the bare metal host secret:
apiVersion: v1 data: password: <credentials-password> username: <credentials-user-name> kind: Secret metadata: labels: kaas.mirantis.com/credentials: "true" kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: <credentials-name> namespace: <managed-cluster-project-name> type: Opaque
In the
datasection, add the IPMI user name and password in the base64 encoding to access the BMC. To obtain the base64-encoded credentials, you can use the following command in your Linux console:echo -n <username|password> | base64
Caution
Each bare metal host must have a unique
Secret.In the
metadatasection, add the uniquenameof credentials and the name of the non-defaultproject (namespace) dedicated for the managed cluster being created.
Apply this secret YAML file to your deployment:
Warning
The kubectl apply command automatically saves the applied data as plain text into the
kubectl.kubernetes.io/last-applied-configurationannotation of the corresponding object. This may result in revealing sensitive data in this annotation when creating or modifying the object.Therefore, do not use kubectl apply on this object. Use kubectl create, kubectl patch, or kubectl edit instead.
If you used kubectl apply on this object, you can remove the
kubectl.kubernetes.io/last-applied-configurationannotation from the object using kubectl edit.kubectl create -n <managedClusterProjectName> -f ${<bmh-cred-file-name>}.yaml
Create a YAML file that contains a description of the new bare metal host:
m:kaas@management-adminonly. This limitation is lifted once the management cluster is updated to the Cluster release 16.4.1 or later.apiVersion: kaas.mirantis.com/v1alpha1 kind: BareMetalHostInventory metadata: annotations: inspect.metal3.io/hardwaredetails-storage-sort-term: hctl ASC, wwn ASC, by_id ASC, name ASC labels: kaas.mirantis.com/baremetalhost-id: <unique-bare-metal-host-hardware-node-id> kaas.mirantis.com/provider: baremetal name: <bare-metal-host-unique-name> namespace: <managed-cluster-project-name> spec: bmc: address: <ip-address-for-bmc-access> bmhCredentialsName: <bare-metal-host-credential-unique-name> bootMACAddress: <bare-metal-host-boot-mac-address> online: true
Note
If you have a limited amount of free and unused IP addresses for server provisioning, you can add the
baremetalhost.metal3.io/detachedannotation that pauses automatic host management to manually allocate an IP address for the host. For details, see Manually allocate IP addresses for bare metal hosts.apiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: annotations: kaas.mirantis.com/baremetalhost-credentials-name: <bare-metal-host-credential-unique-name> labels: kaas.mirantis.com/baremetalhost-id: <unique-bare-metal-host-hardware-node-id> hostlabel.bm.kaas.mirantis.com/worker: "true" kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: <bare-metal-host-unique-name> namespace: <managed-cluster-project-name> spec: bmc: address: <ip-address-for-bmc-access> credentialsName: '' bootMACAddress: <bare-metal-host-boot-mac-address> online: true
Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.Note
If you have a limited amount of free and unused IP addresses for server provisioning, you can add the
baremetalhost.metal3.io/detachedannotation that pauses automatic host management to manually allocate an IP address for the host. For details, see Manually allocate IP addresses for bare metal hosts.apiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: labels: kaas.mirantis.com/baremetalhost-id: <unique-bare-metal-host-hardware-node-id> hostlabel.bm.kaas.mirantis.com/worker: "true" kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: <bare-metal-host-unique-name> namespace: <managed-cluster-project-name> spec: bmc: address: <ip-address-for-bmc-access> credentialsName: <credentials-name> bootMACAddress: <bare-metal-host-boot-mac-address> online: true
For a detailed fields description, see API Reference: BareMetalHostInventory resource and BareMetalHost resource.
Apply this configuration YAML file to your deployment:
kubectl create -n <managedClusterProjectName> -f ${<bare-metal-host-config-file-name>}.yaml
During provisioning,
baremetal-operatorinspects the bare metal host and moves it to thePreparingstate. The host becomes ready to be linked to a bare metal machine.Caution
If changing or adding of DHCP subnets is required to bootstrap new nodes, wait after changing or adding of DHCP subnets until the
dnsmasqpod becomes ready, then create bare metal host objects as described above.For details about the related issue, refer to Inspection error on bare metal hosts after dnsmasq restart.
Verify the new bare metal host object status:
kubectl -n <managed-cluster-project-name> get bmh -o wide <bare-metal-host-unique-name>
Example of system response:
NAMESPACE NAME STATUS STATE CONSUMER BMC BOOTMODE ONLINE ERROR REGION my-project bmh1 OK preparing ip-address-for-bmc-access legacy true region-one
During provisioning, the status changes as follows:
registeringinspectingpreparing
After the bare metal host object switches to the
preparingstage, theinspectingphase finishes and you can verify that hardware information is available in the object status and matches the MOSK cluster hardware requirements. For example:Verify the status of hardware NICs:
kubectl -n <managed-cluster-project-name> get bmh <bare-metal-host-unique-name> -o json | jq -r '[.status.hardware.nics]'
Example of system response:
[ [ { "ip": "172.18.171.32", "mac": "ac:1f:6b:02:81:1a", "model": "0x8086 0x1521", "name": "eno1", "pxe": true }, { "ip": "fe80::225:90ff:fe33:d5ac%ens1f0", "mac": "00:25:90:33:d5:ac", "model": "0x8086 0x10fb", "name": "ens1f0" }, ...
Verify the status of RAM:
kubectl -n <managed-cluster-project-name> get bmh <bare-metal-host-unique-name> -o json | jq -r '[.status.hardware.ramMebibytes]'
Example of system response:
[ 98304 ]
Now, proceed with Create a custom bare metal host profile.
Create a custom bare metal host profile¶
The bare metal host profile is a Kubernetes custom resource. It enables the operator to define how the storage devices and the operating system are provisioned and configured.
This section describes the bare metal host profile default settings and configuration of custom profiles for managed clusters using Container Cloud API. The section also applies to a management cluster with a few differences described in Customize the default bare metal host profile.
Default configuration of the host system storage¶
The default host profile requires three storage devices in the following strict order:
- Boot device and operating system storage
This device contains boot data and operating system data. It is partitioned using the GUID Partition Table (GPT) labels. The root file system is an
ext4file system created on top of an LVM logical volume. For a detailed layout, refer to the table below.
- Local volumes device
This device contains an
ext4file system with directories mounted as persistent volumes to Kubernetes. These volumes are used by the Mirantis Container Cloud services to store its data, including monitoring and identity databases.
- Ceph storage device
This device is used as a Ceph datastore or Ceph OSD on managed clusters.
The following table summarizes the default configuration of the host system storage set up by the Container Cloud bare metal management.
Device/partition |
Name/Mount point |
Recommended size, GB |
Description |
|---|---|---|---|
|
|
4 MiB |
The mandatory GRUB boot partition required for non-UEFI systems. |
|
|
0.2 GiB |
The boot partition required for the UEFI boot mode. |
|
|
64 MiB |
The mandatory partition for the |
|
|
100% of the remaining free space in the LVM volume group |
The main LVM physical volume that is used to create the root file system. |
|
|
100% of the remaining free space in the LVM volume group |
The LVM physical volume that is used to create the file system
for |
|
|
100% of the remaining free space in the LVM volume group |
Clean raw disk that will be used for the Ceph storage backend on managed clusters. |
If required, you can customize the default host storage configuration. For details, see Create MOSK host profiles.
Wipe a device or partition¶
Available since MCC 2.26.0 (17.1.0 and 16.1.0)
Before deploying a cluster, you may need to erase existing data from hardware
devices to be used for deployment. You can either erase an existing partition
or remove all existing partitions from a physical device. For this purpose,
use the wipeDevice structure that configures cleanup behavior during
configuration of a custom bare metal host profile described in
Create MOSK host profiles.
The wipeDevice structure contains the following options:
eraseMetadataConfigures metadata cleanup of a device
eraseDeviceConfigures a complete cleanup of a device
When you enable the eraseMetadata option, which is disabled by default,
the Ansible provisioner attempts to clean up the existing metadata from
the target device. Examples of metadata include:
Existing file system
Logical Volume Manager (LVM) or Redundant Array of Independent Disks (RAID) configuration
The behavior of metadata erasure varies depending on the target device:
If a device is part of other logical devices, for example, a partition, logical volume, or MD RAID volume, such logical device is disassembled and its file system metadata is erased. On the final erasure step, the file system metadata of the target device is erased as well.
If a device is a physical disk, then all its nested partitions along with their nested logical devices, if any, are erased and disassembled. On the final erasure step, all partitions and metadata of the target device are removed.
Caution
None of the eraseMetadata actions include overwriting the
target device with data patterns. For this purpose, use the eraseDevice
option as described in Erase a device.
To enable the eraseMetadata option, use the wipeDevice field in the
spec:devices section of the BareMetalHostProfile object. For a
detailed description of the option, see BareMetalHostProfile resource.
If you require not only disassembling of existing logical volumes but also
removing of all data ever written to the target device, configure the
eraseDevice option, which is disabled by default. This option is not
applicable to paritions, LVM, or MD RAID logical volumes because such volumes
may use caching that prevents a physical device from being erased properly.
Important
The eraseDevice option does not replace the secure erase.
To configure the eraseDevice option, use the wipeDevice field in the
spec:devices section of the BareMetalHostProfile object. For a
detailed description of the option, see BareMetalHostProfile resource.
Create MOSK host profiles¶
Different types of MOSK nodes require differently configured host storage. This section describes how to create custom host profiles for different types of MOSK nodes.
You can create custom profiles for managed clusters using Container Cloud API.
Note
The procedure below also applies to management clusters.
You can use flexible size units throughout bare metal host profiles.
For example, you can now use either sizeGiB: 0.1 or size: 100Mi
when specifying a device size.
Mirantis recommends using only one parameter name type and units throughout
the configuration files. If both sizeGiB and size are used,
sizeGiB is ignored during deployment and the suffix is adjusted
accordingly. For example, 1.5Gi will be serialized as 1536Mi. The size
without units is counted in bytes. For example, size: 120 means 120 bytes.
Warning
All data will be wiped during cluster deployment on devices
defined directly or indirectly in the fileSystems list of
BareMetalHostProfile. For example:
A raw device partition with a file system on it
A device partition in a volume group with a logical volume that has a file system on it
An mdadm RAID device with a file system on it
An LVM RAID device with a file system on it
The wipe field is always considered true for these devices.
The false value is ignored.
Therefore, to prevent data loss, move the necessary data from these file systems to another server beforehand, if required.
To create MOSK bare metal host profiles:
Select from the following options:
For a management cluster, log in to the bare metal seed node that will be used to bootstrap the management cluster.
For a managed cluster, log in to the local machine where you management cluster
kubeconfigis located and wherekubectlis installed.
Note
The management cluster
kubeconfigis created automatically during the last stage of the management cluster bootstrap.Select from the following options:
For a management cluster, open
templates/bm/baremetalhostprofiles.yaml.templatefor editing.For a managed cluster, create a new bare metal host profile for MOSK compute nodes in a YAML file under the
templates/bm/directory.
Edit the host profile using the example template below to meet your hardware configuration requirements:
apiVersion: metal3.io/v1alpha1 kind: BareMetalHostProfile metadata: name: <PROFILE_NAME> namespace: <PROJECT_NAME> spec: devices: # From the HW node, obtain the first device, which size is at least 60Gib - device: workBy: "by_id,by_wwn,by_path,by_name" minSize: 60Gi type: ssd wipe: true partitions: - name: bios_grub partflags: - bios_grub size: 4Mi wipe: true - name: uefi partflags: - esp size: 200Mi wipe: true - name: config-2 size: 64Mi wipe: true # This partition is only required on compute nodes if you plan to # use LVM ephemeral storage. - name: lvm_nova_part wipe: true size: 100Gi - name: lvm_root_part size: 0 wipe: true # From the HW node, obtain the second device, which size is at least 60Gib # If a device exists but does not fit the size, # the BareMetalHostProfile will not be applied to the node - device: workBy: "by_id,by_wwn,by_path,by_name" minSize: 60Gi type: ssd wipe: true # From the HW node, obtain the disk device with the exact name - device: workBy: "by_id,by_wwn,by_path,by_name" minSize: 60Gi wipe: true partitions: - name: lvm_lvp_part size: 0 wipe: true # Example of wiping a device w\o partitioning it. # Mandatory for the case when a disk is supposed to be used for Ceph backend # later - device: workBy: "by_id,by_wwn,by_path,by_name" wipe: true fileSystems: - fileSystem: vfat partition: config-2 - fileSystem: vfat mountPoint: /boot/efi partition: uefi - fileSystem: ext4 logicalVolume: root mountPoint: / - fileSystem: ext4 logicalVolume: lvp mountPoint: /mnt/local-volumes/ logicalVolumes: - name: root size: 0 vg: lvm_root - name: lvp size: 0 vg: lvm_lvp postDeployScript: | #!/bin/bash -ex echo $(date) 'post_deploy_script done' >> /root/post_deploy_done preDeployScript: | #!/bin/bash -ex echo $(date) 'pre_deploy_script done' >> /root/pre_deploy_done volumeGroups: - devices: - partition: lvm_root_part name: lvm_root - devices: - partition: lvm_lvp_part name: lvm_lvp grubConfig: defaultGrubOptions: - GRUB_DISABLE_RECOVERY="true" - GRUB_PRELOAD_MODULES=lvm - GRUB_TIMEOUT=20 kernelParameters: sysctl: # For the list of options prohibited to change, refer to # https://docs.mirantis.com/mke/3.7/install/predeployment/set-up-kernel-default-protections.html kernel.dmesg_restrict: "1" kernel.core_uses_pid: "1" fs.file-max: "9223372036854775807" fs.aio-max-nr: "1048576" fs.inotify.max_user_instances: "4096" vm.max_map_count: "262144"
Add or edit the mandatory parameters in the new
BareMetalHostProfileobject. For the parameters description, see BareMetalHostProfile spec <bmhprofile-spec>.Note
If asymmetric traffic is expected on some of the managed cluster nodes, enable the loose mode for the corresponding interfaces on those nodes by setting the
net.ipv4.conf.<interface-name>.rp_filterparameter to"2"in thekernelParameters.sysctlsection. For example:kernelParameters: sysctl: net.ipv4.conf.k8s-lcm.rp_filter: "2"
Configure required disks for the Ceph cluster as described in Configure Ceph disks in a host profile.
Optional. Configure wiping of the target device or partition to be used for cluster deployment as described in Wipe a device or partition.
Optional. Configure multiple devices for LVM volume using the example template extract below for reference.
Caution
The following template extract contains only sections relevant to LVM configuration with multiple PVs. Expand the main template described in the previous step with the configuration below if required.
spec: devices: ... - device: ... partitions: - name: lvm_lvp_part1 size: 0 wipe: true - device: ... partitions: - name: lvm_lvp_part2 size: 0 wipe: true volumeGroups: ... - devices: - partition: lvm_lvp_part1 - partition: lvm_lvp_part2 name: lvm_lvp logicalVolumes: ... - name: root size: 0 vg: lvm_lvp fileSystems: ... - fileSystem: ext4 logicalVolume: root mountPoint: /
Optional. Technology Preview. Configure support of the Redundant Array of Independent Disks (RAID) that allows, for example, installing a cluster operating system on a RAID device, refer to Configure RAID support.
Optional. Configure the RX/TX buffer size for physical network interfaces and
txqueuelenfor any network interfaces.This configuration can greatly benefit high-load and high-performance network interfaces. You can configure these parameters using the
udevrules. For example:postDeployScript: | #!/bin/bash -ex ... echo 'ACTION=="add|change", SUBSYSTEM=="net", KERNEL=="eth*|en*", RUN+="/sbin/ethtool -G $name rx 4096 tx 4096"' > /etc/udev/rules.d/59-net.ring.rules echo 'ACTION=="add|change", SUBSYSTEM=="net", KERNEL=="eth*|en*|bond*|k8s-*|v*" ATTR{tx_queue_len}="10000"' > /etc/udev/rules.d/58-net.txqueue.rules
Select from the following options:
For a management cluster, proceed with the cluster bootstrap procedure as described in Deploy a management cluster.
For a managed cluster, select from the following options:
Log in to the Container Cloud web UI with the
operatorpermissions.Switch to the required non-
defaultproject using the Switch Project action icon located on top of the main left-side navigation panel.Caution
Do not create a MOSK cluster in the
defaultproject (Kubernetes namespace), which is dedicated for the management cluster only. If no projects are defined, first create a newmoskproject as described in Create a project for MOSK clusters.In the left sidebar, navigate to Baremetal and click the Host Profiles tab.
Click Create Host Profile.
Fill out the Create host profile form:
- Name
Name of the bare metal host profile.
- Specification
BareMetalHostProfileobject specification in the YAML format that you have previously created. Click Edit to edit theBareMetalHostProfileobject if required.Note
Before Container Cloud 2.28.0 (Cluster releases 17.3.0 and 16.3.0), the field name is YAML file, and you can upload the required YAML file instead of inserting and editing it.
- Labels
Available since Container Cloud 2.28.0 (Cluster releases 17.3.0 and 16.3.0). Key-value pairs attached to
BareMetalHostProfile.
Add the bare metal host profile to your management cluster:
kubectl --kubeconfig <pathToManagementClusterKubeconfig> -n <managedClusterProjectName> apply -f <pathToBareMetalHostProfileFile>
If required, further modify the host profile:
kubectl --kubeconfig <pathToManagementClusterKubeconfig> -n <managedClusterProjectName> edit baremetalhostprofile <hostProfileName>
Repeat the steps above to create host profiles for other OpenStack node roles such as control plane nodes and storage nodes.
Now, proceed to Enable huge pages in a host profile.
Configure Ceph disks in a host profile¶
This section describes how to configure devices for the Ceph cluster in the
BareMetalHostProfile object of a managed cluster.
To configure disks for a Ceph cluster:
Open the
BareMetalHostProfileobject of a managed cluster for editing.In the
spec.devicessection, add each disk intended for use as a Ceph OSD data device withsize: 0andwipe: true.Example configuration for
sde-sdhdisks to use as Ceph OSDs:spec: devices: ... - device: byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:1 size: 0 wipe: true - device: byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:2 size: 0 wipe: true - device: byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:3 size: 0 wipe: true - device: byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:4 size: 0 wipe: true
Since MOSK 23.2, if you plan to use a separate metadata device for Ceph OSD, configure the
spec.devicessection as described below.Important
Mirantis highly recommends configuring disk partitions for Ceph OSD metadata using
BareMetalHostProfile.Configuration of a separate metadata device for Ceph OSD
Add the device to
spec.deviceswith a single partition that will use the entire disk size.For example, if you plan to use four Ceph OSDs with a separate metadata device for each Ceph OSD, configure the
spec.devicessection as follows:spec: devices: ... - device: byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:5 wipe: true partitions: - name: ceph_meta size: 0 wipe: true
Create a volume group on top of the defined partition and create the required number of logical volumes (LVs) on top of the created volume group (VG). Add one logical volume per one Ceph OSD on the node.
Example snippet of an LVM configuration for a Ceph metadata disk:
spec: ... volumeGroups: ... - devices: - partition: ceph_meta name: bluedb logicalVolumes: ... - name: meta_1 size: 25%VG vg: bluedb - name: meta_2 size: 25%VG vg: bluedb - name: meta_3 size: 25%VG vg: bluedb - name: meta_4 size: 25%VG vg: bluedb
Important
Plan LVs of a separate metadata device thoroughly. Any logical volume misconfiguration causes redeployment of all Ceph OSDs that use this disk as metadata devices.
Note
General Ceph recommendation is to have a metadata device in between 1% to 4% of the Ceph OSD data size. Mirantis highly recommends having at least 4% of Ceph OSD data size.
If you plan using a disk as a separate metadata device for 10 Ceph OSDs, define the size of an LV for each Ceph OSD in between 1% to 4% of the corresponding Ceph OSD data size. If RADOS Gateway is enabled, the minimum data size must be 4%. For details, see Ceph documentation: Bluestore config reference.
For example, if the total data size of 10 Ceph OSDs equals
1Tbwith100Gbeach, assign a metadata disk less than10Gbwith1Gbper each LV. The recommended size is40Gbwith4Gbper each LV.After applying
BareMetalHostProfile, the bare metal provider creates an LVM partitioning for the metadata disk and places these volumes as/devpaths, for example,/dev/bluedb/meta_1or/dev/bluedb/meta_3.Example template of a host profile configuration for Ceph
spec: ... devices: ... - device: byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:1 wipe: true - device: byName: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:2 wipe: true - device: byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:3 wipe: true - device: byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:4 wipe: true - device: byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:5 wipe: true partitions: - name: ceph_meta size: 0 wipe: true volumeGroups: ... - devices: - partition: ceph_meta name: bluedb logicalVolumes: ... - name: meta_1 size: 25%VG vg: bluedb - name: meta_2 size: 25%VG vg: bluedb - name: meta_3 size: 25%VG vg: bluedb - name: meta_4 size: 25%VG vg: bluedb
After applying such
BareMetalHostProfileto a node, thenodesspec of theKaaSCephClusterobject contains the followingstorageDevicessection:spec: cephClusterSpec: ... nodes: ... machine-1: ... storageDevices: - fullPath: /dev/disk/by-id/scsi-SATA_ST4000NM002A-2HZ_WS20NNKC config: metadataDevice: /dev/bluedb/meta_1 - fullPath: /dev/disk/by-id/ata-ST4000NM002A-2HZ101_WS20NEGE config: metadataDevice: /dev/bluedb/meta_2 - fullPath: /dev/disk/by-id/scsi-0ATA_ST4000NM002A-2HZ_WS20LEL3 config: metadataDevice: /dev/bluedb/meta_3 - fullPath: /dev/disk/by-id/ata-HGST_HUS724040ALA640_PN1334PEDN9SSU config: metadataDevice: /dev/bluedb/meta_4
spec: cephClusterSpec: ... nodes: ... machine-1: ... storageDevices: - name: sde config: metadataDevice: /dev/bluedb/meta_1 - name: sdf config: metadataDevice: /dev/bluedb/meta_2 - name: sdg config: metadataDevice: /dev/bluedb/meta_3 - name: sdh config: metadataDevice: /dev/bluedb/meta_4
Enable huge pages in a host profile¶
The BareMetalHostProfile API allows configuring a host to use the
huge pages feature of the Linux kernel on managed clusters. The procedure
included in this section applies to both new and existing cluster
deployments.
Note
Huge pages is a mode of operation of the Linux kernel. With huge pages enabled, the kernel allocates the RAM in bigger chunks, or pages. This allows kernel-based virtual machines and virtual machines running on it to use the host RAM more efficiently and improves the performance of the virtual machines.
To enable huge pages in a custom bare metal host profile for a managed cluster:
Log in to the local machine where you management cluster
kubeconfigis located and wherekubectlis installed.Note
The management cluster
kubeconfigis created automatically during the last stage of the management cluster bootstrap.Open for editing or create a new bare metal host profile under the
templates/bm/directory.Edit the
grubConfigsection of the host profilespecusing the example below to configure the kernel boot parameters and enable huge pages:spec: grubConfig: defaultGrubOptions: - GRUB_DISABLE_RECOVERY="true" - GRUB_PRELOAD_MODULES=lvm - GRUB_TIMEOUT=20 - GRUB_CMDLINE_LINUX_DEFAULT="hugepagesz=1G hugepages=N"
The example configuration above will allocate
Nhuge pages of 1 GB each on the server boot. The lasthugepageszparameter value is default unlessdefault_hugepageszis defined. For details about possible values, see official Linux kernel documentation.Add the bare metal host profile to your management cluster:
kubectl --kubeconfig <pathToManagementClusterKubeconfig> -n <projectName> apply -f <pathToBareMetalHostProfileFile>
If required, further modify the host profile:
kubectl --kubeconfig <pathToManagementClusterKubeconfig> -n <projectName> edit baremetalhostprofile <hostProfileName>
Proceed to Create a MOSK cluster.
Configure RAID support¶
TechPreview
During a management or MOSK cluster creation, you can
configure the support of the software-based Redundant Array of Independent
Disks (RAID) using BareMetalHosProfile to set up an LVM-based RAID level 1
(raid1) or an mdadm-based RAID level 0, 1, or 10 (raid0, raid1, or
raid10).
If required, you can further configure RAID in the same profile, for example, to install a cluster operating system onto a RAID device.
Caution
RAID configuration on already provisioned bare metal machines or on an existing cluster is not supported.
To start using any kind of RAIDs, reprovisioning of machines with a new
BaremetalHostProfileis required.Mirantis supports the
raid1type of RAID devices both for LVM and mdadm.Mirantis supports the
raid0type for the mdadm RAID to be on par with the LVMlineartype.Mirantis recommends having at least two physical disks for
raid0andraid1devices to prevent unnecessary complexity.Mirantis supports the
raid10type for mdadm RAID. At least four physical disks are required for this type of RAID.Only an even number of disks can be used for a
raid1orraid10device.
TechPreview
Warning
The EFI system partition partflags: ['esp'] must be
a physical partition in the main partition table of the disk, not under
LVM or mdadm software RAID.
During configuration of your custom bare metal host profile,
you can create an LVM-based software RAID device raid1 by adding
type: raid1 to the logicalVolume spec in BaremetalHostProfile.
For the LVM RAID parameters description, refer to BareMetalHostProfile spec.
For a bare metal host profile configuration, refer to Create a custom bare metal host profile.
Caution
The logicalVolume spec of the raid1 type requires at least
two devices (partitions) in volumeGroup where you build a logical
volume. For an LVM of the linear type, one device is enough.
You can use flexible size units throughout bare metal host profiles.
For example, you can now use either sizeGiB: 0.1 or size: 100Mi
when specifying a device size.
Mirantis recommends using only one parameter name type and units throughout
the configuration files. If both sizeGiB and size are used,
sizeGiB is ignored during deployment and the suffix is adjusted
accordingly. For example, 1.5Gi will be serialized as 1536Mi. The size
without units is counted in bytes. For example, size: 120 means 120 bytes.
Note
The LVM raid1 requires additional space to store the raid1
metadata on a volume group, roughly 4 MB for each partition.
Therefore, you cannot create a logical volume of exactly the same
size as the partitions it works on.
For example, if you have two partitions of 10 GiB, the corresponding
raid1 logical volume size will be less than 10 GiB. For that
reason, you can either set size: 0 to use all available
space on the volume group, or set a smaller size than the partition
size. For example, use size: 9.9Gi instead of
size: 10Gi for the logical volume.
The following example illustrates an extract of BaremetalHostProfile
with / on the LVM raid1.
...
devices:
- device:
workBy: "by_id,by_wwn,by_path,by_name"
minSize: 200Gi
type: hdd
wipe: true
partitions:
- name: root_part1
size: 120Gi
partitions:
- name: rest_sda
size: 0
- device:
workBy: "by_id,by_wwn,by_path,by_name"
minSize: 200Gi
type: hdd
wipe: true
partitions:
- name: root_part2
size: 120Gi
partitions:
- name: rest_sdb
size: 0
volumeGroups:
- name: vg-root
devices:
- partition: root_part1
- partition: root_part2
- name: vg-data
devices:
- partition: rest_sda
- partition: rest_sdb
logicalVolumes:
- name: root
type: raid1 ## <-- LVM raid1
vg: vg-root
size: 119.9Gi
- name: data
type: linear
vg: vg-data
size: 0
fileSystems:
- fileSystem: ext4
logicalVolume: root
mountPoint: /
mountOpts: "noatime,nodiratime"
- fileSystem: ext4
logicalVolume: data
mountPoint: /mnt/data
Warning
All data will be wiped during cluster deployment on devices
defined directly or indirectly in the fileSystems list of
BareMetalHostProfile. For example:
A raw device partition with a file system on it
A device partition in a volume group with a logical volume that has a file system on it
An mdadm RAID device with a file system on it
An LVM RAID device with a file system on it
The wipe field is always considered true for these devices.
The false value is ignored.
Therefore, to prevent data loss, move the necessary data from these file systems to another server beforehand, if required.
TechPreview
You can configure an LVM volume group on top of mdadm-based RAID devices as
physical volumes using the BareMetalHostProfile resource. List the
required RAID devices in a separate field of the volumeGroups definition
within the storage configuration of BareMetalHostProfile.
You can use flexible size units throughout bare metal host profiles.
For example, you can now use either sizeGiB: 0.1 or size: 100Mi
when specifying a device size.
Mirantis recommends using only one parameter name type and units throughout
the configuration files. If both sizeGiB and size are used,
sizeGiB is ignored during deployment and the suffix is adjusted
accordingly. For example, 1.5Gi will be serialized as 1536Mi. The size
without units is counted in bytes. For example, size: 120 means 120 bytes.
Warning
All data will be wiped during cluster deployment on devices
defined directly or indirectly in the fileSystems list of
BareMetalHostProfile. For example:
A raw device partition with a file system on it
A device partition in a volume group with a logical volume that has a file system on it
An mdadm RAID device with a file system on it
An LVM RAID device with a file system on it
The wipe field is always considered true for these devices.
The false value is ignored.
Therefore, to prevent data loss, move the necessary data from these file systems to another server beforehand, if required.
The following example illustrates an extract of BaremetalHostProfile with
a volume group named lvm_nova to be created on top of an mdadm-based
RAID device raid1:
...
devices:
- device:
workBy: "by_id,by_wwn,by_path,by_name"
minSize: 60Gi
type: ssd
wipe: true
partitions:
- name: bios_grub
partflags:
- bios_grub
size: 4Mi
- name: uefi
partflags:
- esp
size: 200Mi
- name: config-2
size: 64Mi
- device:
workBy: "by_id,by_wwn,by_path,by_name"
minSize: 30Gi
type: ssd
wipe: true
partitions:
- name: md0_part1
- device:
workBy: "by_id,by_wwn,by_path,by_name"
minSize: 30Gi
type: ssd
wipe: true
partitions:
- name: md0_part2
softRaidDevices:
- devices:
- partition: md0_part1
- partition: md0_part2
level: raid1
metadata: "1.0"
name: /dev/md0
volumeGroups:
- devices:
- softRaidDevice: /dev/md0
name: lvm_nova
...
TechPreview
Warning
The EFI system partition partflags: ['esp'] must be
a physical partition in the main partition table of the disk, not under
LVM or mdadm software RAID.
During configuration of your custom bare metal host profile as described in
Create a custom bare metal host profile, you can create an mdadm-based software RAID
device raid0 and raid1 by describing the mdadm devices under the
softRaidDevices field in BaremetalHostProfile. For example:
...
softRaidDevices:
- name: /dev/md0
devices:
- partition: sda1
- partition: sdb1
- name: raid-name
devices:
- partition: sda2
- partition: sdb2
...
You can also use the raid10 type for the mdadm-based software RAID devices.
This type requires at least four and in total an even number of storage devices
available on your servers. For example:
softRaidDevices:
- name: /dev/md0
level: raid10
devices:
- partition: sda1
- partition: sdb1
- partition: sdd1
The following fields in softRaidDevices describe RAID devices:
nameName of the RAID device to refer to throughout the
baremetalhostprofile.
levelType or level of RAID used to create a device, defaults to
raid1. Set toraid0orraid10to create a device of the corresponding type.
devicesList of physical devices or partitions used to build a software RAID device. It must include at least two partitions or devices to build a
raid0andraid1devices and at least four forraid10.
For the rest of the mdadm RAID parameters, see BareMetalHostProfile spec.
Caution
The mdadm RAID devices cannot be created on top of LVM devices.
You can use flexible size units throughout bare metal host profiles.
For example, you can now use either sizeGiB: 0.1 or size: 100Mi
when specifying a device size.
Mirantis recommends using only one parameter name type and units throughout
the configuration files. If both sizeGiB and size are used,
sizeGiB is ignored during deployment and the suffix is adjusted
accordingly. For example, 1.5Gi will be serialized as 1536Mi. The size
without units is counted in bytes. For example, size: 120 means 120 bytes.
Warning
All data will be wiped during cluster deployment on devices
defined directly or indirectly in the fileSystems list of
BareMetalHostProfile. For example:
A raw device partition with a file system on it
A device partition in a volume group with a logical volume that has a file system on it
An mdadm RAID device with a file system on it
An LVM RAID device with a file system on it
The wipe field is always considered true for these devices.
The false value is ignored.
Therefore, to prevent data loss, move the necessary data from these file systems to another server beforehand, if required.
The following example illustrates an extract of BaremetalHostProfile
with / on the mdadm raid1 and some data storage on raid0:
Example with / on the mdadm raid1 and data storage on raid0
...
devices:
- device:
workBy: "by_id,by_wwn,by_path,by_name"
type: nvme
wipe: true
partitions:
- name: root_part1
size: 120Gi
partitions:
- name: rest_sda
size: 0
- device:
workBy: "by_id,by_wwn,by_path,by_name"
type: nvme
wipe: true
partitions:
- name: root_part2
size: 120Gi
partitions:
- name: rest_sdb
size: 0
softRaidDevices:
- name: root
level: raid1 ## <-- mdadm raid1
devices:
- partition: root_part1
- partition: root_part2
- name: raid-name
level: raid0 ## <-- mdadm raid0
devices:
- partition: rest_sda
- partition: rest_sdb
fileSystems:
- fileSystem: ext4
softRaidDevice: root
mountPoint: /
mountOpts: "noatime,nodiratime"
- fileSystem: ext4
softRaidDevice: data
mountPoint: /mnt/data
...
The following example illustrates an extract of BaremetalHostProfile
with data storage on a raid10 device:
Example with data storage on the mdadm raid10
...
devices:
- device:
workBy: "by_id,by_wwn,by_path,by_name"
minSize: 60Gi
type: ssd
wipe: true
partitions:
- name: bios_grub1
partflags:
- bios_grub
size: 4Mi
wipe: true
- name: uefi
partflags:
- esp
size: 200Mi
wipe: true
- name: config-2
size: 64Mi
wipe: true
- name: lvm_root
size: 0
wipe: true
- device:
workBy: "by_id,by_wwn,by_path,by_name"
minSize: 60Gi
type: nvme
wipe: true
partitions:
- name: md_part1
partflags:
- raid
size: 40Gi
wipe: true
- device:
workBy: "by_id,by_wwn,by_path,by_name"
minSize: 60Gi
type: nvme
wipe: true
partitions:
- name: md_part2
partflags:
- raid
size: 40Gi
wipe: true
- device:
workBy: "by_id,by_wwn,by_path,by_name"
minSize: 60Gi
type: nvme
wipe: true
partitions:
- name: md_part3
partflags:
- raid
size: 40Gi
wipe: true
- device:
workBy: "by_id,by_wwn,by_path,by_name"
minSize: 60Gi
type: nvme
wipe: true
partitions:
- name: md_part4
partflags:
- raid
size: 40Gi
wipe: true
fileSystems:
- fileSystem: vfat
partition: config-2
- fileSystem: vfat
mountPoint: /boot/efi
partition: uefi
- fileSystem: ext4
mountOpts: rw,noatime,nodiratime,lazytime,nobarrier,commit=240,data=ordered
mountPoint: /
partition: root
- filesystem: ext4
mountPoint: /var
softRaidDevice: /dev/md0
softRaidDevices:
- devices:
- partition: md_root_part1
- partition: md_root_part2
- partition: md_root_part3
- partition: md_root_part4
level: raid10
metadata: "1.2"
name: /dev/md0
...
TechPreview
Warning
The EFI system partition partflags: ['esp'] must be
a physical partition in the main partition table of the disk, not under
LVM or mdadm software RAID.
You can deploy MOSK on local software-based Redundant Array of Independent Disks (RAID) devices to withstand failure of one device at a time.
Using a custom bare metal host profile, you can configure and create
an mdadm-based software RAID device of type raid10 if you have
an even number of devices available on your servers. At least four
storage devices are required for such RAID device.
During configuration of your custom bare metal host profile as described in
Create a custom bare metal host profile, create an mdadm-based software RAID device
raid10 by describing the mdadm devices under the softRaidDevices
field. For example:
...
softRaidDevices:
- name: /dev/md0
level: raid10
devices:
- partition: sda1
- partition: sdb1
- partition: sdc1
- partition: sdd1
...
The following fields in softRaidDevices describe RAID devices:
nameName of the RAID device to refer to throughout the
baremetalhostprofile.
devicesList of physical devices or partitions used to build a software RAID device. It must include at least four partitions or devices to build a
raid10device.
levelType or level of RAID used to create device. Set to
raid10orraid1to create a device of the corresponding type.
For the rest of the mdadm RAID parameters, see BareMetalHostProfile spec.
Caution
The mdadm RAID devices cannot be created on top of an LVM device.
The following example illustrates an extract of baremetalhostprofile
with data storage on a raid10 device:
...
devices:
- device:
minSize: 60Gi
wipe: true
partitions:
- name: bios_grub1
partflags:
- bios_grub
size: 4Mi
wipe: true
- name: uefi
partflags:
- esp
size: 200Mi
wipe: true
- name: config-2
size: 64Mi
wipe: true
- name: lvm_root
size: 0
wipe: true
- device:
minSize: 60Gi
wipe: true
partitions:
- name: md_part1
partflags:
- raid
size: 40Gi
wipe: true
- device:
minSize: 60Gi
wipe: true
partitions:
- name: md_part2
partflags:
- raid
size: 40Gi
wipe: true
- device:
minSize: 60Gi
wipe: true
partitions:
- name: md_part3
partflags:
- raid
size: 40Gi
wipe: true
- device:
minSize: 60Gi
wipe: true
partitions:
- name: md_part4
partflags:
- raid
size: 40Gi
wipe: true
fileSystems:
- fileSystem: vfat
partition: config-2
- fileSystem: vfat
mountPoint: /boot/efi
partition: uefi
- fileSystem: ext4
mountOpts: rw,noatime,nodiratime,lazytime,nobarrier,commit=240,data=ordered
mountPoint: /
partition: root
- filesystem: ext4
mountPoint: /var
softRaidDevice: /dev/md0
softRaidDevices:
- devices:
- partition: md_root_part1
- partition: md_root_part2
- partition: md_root_part3
- partition: md_root_part4
level: raid10
metadata: "1.2"
name: /dev/md0
...
Warning
When building the raid10 array on top of device partitions,
make sure that only one partition per device is used for a given array.
Although having two partitions located on the same physical device as array members is technically possible, it may lead to data loss if mdadm selects both partitions of the same drive to be mirrored. In such case, redundancy against entire drive failure is lost.
Warning
All data will be wiped during cluster deployment on devices
defined directly or indirectly in the fileSystems list of
BareMetalHostProfile. For example:
A raw device partition with a file system on it
A device partition in a volume group with a logical volume that has a file system on it
An mdadm RAID device with a file system on it
An LVM RAID device with a file system on it
The wipe field is always considered true for these devices.
The false value is ignored.
Therefore, to prevent data loss, move the necessary data from these file systems to another server beforehand, if required.
Create a MOSK cluster¶
With L2 networking templates, you can create MOSK clusters with advanced host networking configurations. For example, you can create bond interfaces on top of physical interfaces on the host or use multiple subnets to separate different types of network traffic.
You can use several host-specific L2 templates per one cluster to support different hardware configurations. For example, you can create L2 templates with a different number and layout of NICs to be applied to specific machines of one cluster.
You can also use multiple L2 templates to support different roles for nodes in a MOSK installation. You can create L2 templates with different logical interfaces and assign them to individual machines based on their roles in a MOSK cluster.
Caution
Modification of L2 templates in use is only allowed with a mandatory validation step from the infrastructure operator to prevent accidental cluster failures due to unsafe changes. The list of risks posed by modifying L2 templates includes:
Services running on hosts cannot reconfigure automatically to switch to the new IP addresses and/or interfaces.
Connections between services are interrupted unexpectedly, which can cause data loss.
Incorrect configurations on hosts can lead to irrevocable loss of connectivity between services and unexpected cluster partition or disassembly.
For details, see Modify network configuration on an existing machine.
Since MOSK 23.2.2, in the Technology Preview scope, you can create a MOSK cluster with the multi-rack topology, where cluster nodes including Kubernetes masters are distributed across multiple racks without L2 layer extension between them, and use BGP for announcement of the cluster API load balancer address and external addresses of Kubernetes load-balanced services.
Implementation of the multi-rack topology implies the use of Rack and
MultiRackCluster objects that support configuration of BGP announcement
of the cluster API load balancer address. For the configuration procedure,
refer to Configure BGP announcement for cluster API LB address. For configuring the BGP announcement of
external addresses of Kubernetes load-balanced services, refer to
Configure and verify MetalLB.
Follow the procedures described in the below subsections to configure initial settings and advanced network objects for your managed clusters.
Create a managed bare metal cluster¶
This section instructs you on how to configure and deploy a managed cluster that is based on the baremetal-based management cluster through the Mirantis Container Cloud web UI.
Note
Due to the known issue 50181, creation of a compact managed cluster or addition of any labels to the control plane nodes is not available through the Container Cloud web UI.
To create a managed cluster on bare metal:
Available since the Cluster release 16.1.0 on the management cluster. If you plan to deploy a large managed cluster, enable dynamic IP allocation to increase the amount of baremetal hosts to be provisioned in parallel. For details, see Enable dynamic IP allocation.
Available since Container Cloud 2.24.0 (Cluster release 14.0.0). Optional. Technology Preview. Enable custom host names for cluster machines. When enabled, any machine host name in a particular region matches the related
Machineobject name. For example, instead of the defaultkaas-node-<UID>, a machine host name will bemaster-0. The custom naming format is more convenient and easier to operate with.For details, see Configure host names for cluster machines.
Skip this step if you enabled this feature during management cluster bootstrap, because custom host names will be automatically enabled on the related managed cluster as well.
Log in to the Container Cloud web UI with the
writerpermissions.Switch to the required non-
defaultproject using the Switch Project action icon located on top of the main left-side navigation panel.Caution
Do not create a MOSK cluster in the
defaultproject (Kubernetes namespace), which is dedicated for the management cluster only. If no projects are defined, first create a newmoskproject as described in Create a project for MOSK clusters.In the SSH keys tab, click Add SSH Key to upload the public SSH key that will be used for the SSH access to VMs.
Optional. In the Proxies tab, enable proxy access to the managed cluster:
Click Add Proxy.
In the Add New Proxy wizard, fill out the form with the following parameters:
Proxy configuration¶ Parameter
Description
Proxy Name
Name of the proxy server to use during a managed cluster creation.
Region Removed in MOSK 24.1
From the drop-down list, select the required region.
HTTP Proxy
Add the HTTP proxy server domain name in the following format:
http://proxy.example.com:port- for anonymous accesshttp://user:password@proxy.example.com:port- for restricted access
HTTPS Proxy
Add the HTTPS proxy server domain name in the same format as for HTTP Proxy.
No Proxy
Comma-separated list of IP addresses or domain names.
Note
The possibility to use a MITM proxy with a CA certificate is available since MOSK 23.1.
For the list of Mirantis resources and IP addresses to be accessible from the Container Cloud clusters, see Reference Architecture: Requirements.
In the Clusters tab, click Create Cluster.
Configure the new cluster in the Create New Cluster wizard that opens:
Define general and Kubernetes parameters:
Create new cluster: General, Provider, and Kubernetes¶ Section
Parameter name
Description
General settings
Cluster name
The cluster name.
Provider
Select Baremetal.
Region Removed since MOSK 24.1
From the drop-down list, select Baremetal.
Release version
Select a Container Cloud version with the OpenStack label tag. Otherwise, you will not be able to deploy MOSK on this managed cluster.
Proxy
Optional. From the drop-down list, select the proxy server name that you have previously created.
SSH keys
From the drop-down list, select the SSH key name that you have previously added for SSH access to the bare metal hosts.
Container Registry
From the drop-down list, select the Docker registry name that you have previously added using the Container Registries tab. For details, see Define a custom CA certificate for a private Docker registry.
Enable WireGuard
Optional. Technology Preview. Deprecated since Container Cloud 2.29.0 (Cluster releases 17.4.0 and 16.4.0). Available since Container Cloud 2.24.0 (Cluster release 14.0.0). Enable WireGuard for traffic encryption on the Kubernetes workloads network.
WireGuard configuration
Ensure that the Calico MTU size is at least 60 bytes smaller than the interface MTU size of the workload network. IPv4 WireGuard uses a 60-byte header. For details, see Set the MTU size for Calico.
Enable WireGuard by selecting the Enable WireGuard check box.
Caution
Changing this parameter on a running cluster causes a downtime that can vary depending on the cluster size.
Note
This parameter was renamed from Enable Secure Overlay to Enable WireGuard in Container Cloud 2.25.0 (Cluster releases 17.0.0 and 16.0.0).
For more details about WireGuard, see Calico documentation: Encrypt in-cluster pod traffic.
Parallel Upgrade Of Worker Machines
Optional. Available since Container Cloud 2.25.0 (Cluster releases 17.0.0 and 16.0.0).
The maximum number of the worker nodes to update simultaneously. It serves as an upper limit on the number of machines that are drained at a given moment of time. Defaults to
1.You can also configure this option after deployment before the cluster update.
Parallel Preparation For Upgrade Of Worker Machines
Optional. Available since Container Cloud 2.25.0 (Cluster releases 17.0.0 and 16.0.0)
The maximum number of worker nodes being prepared at a given moment of time, which includes downloading of new artifacts. It serves as a limit for the network load that can occur when downloading the files to the nodes. Defaults to
50.You can also configure this option after deployment before the cluster update.
Provider
LB host IP
The IP address of the load balancer endpoint that will be used to access the Kubernetes API of the new cluster. This IP address must be in the LCM network if a separate LCM network is in use and if L2 (ARP) announcement of cluster API load balancer IP is in use.
LB address range Removed in 24.3
The range of IP addresses that can be assigned to load balancers for Kubernetes Services by MetalLB. For a more flexible MetalLB configuration, refer to Configure and verify MetalLB.
Note
Since MOSK 24.3, MetalLB configuration must be added after cluster creation.
Kubernetes
Services CIDR blocks
The Kubernetes Services CIDR blocks. For example,
10.233.0.0/18.Pods CIDR blocks
The Kubernetes pods CIDR blocks. For example,
10.233.64.0/18.Note
The network subnet size of Kubernetes pods influences the number of nodes that can be deployed in the cluster.
The default subnet size
/18is enough to create a cluster with up to 256 nodes. Each node uses the/26address blocks (64 addresses), at least one address block is allocated per node. These addresses are used by the Kubernetes pods withhostNetwork: false. The cluster size may be limited further when some nodes use more than one address block.Configure StackLight:
Note
If StackLight is enabled in non-HA mode but Ceph is not deployed yet, StackLight will not be installed and will be stuck in the
Yellowstate waiting for a successful Ceph installation. Once the Ceph cluster is deployed, the StackLight installation resumes. To deploy a Ceph cluster, refer to Add a Ceph cluster.StackLight configuration
Section
Parameter name
Description
StackLight
Enable Monitoring
Selected by default. Deselect to skip StackLight deployment.
Note
You can also enable, disable, or configure StackLight parameters after deploying a managed cluster. For details, see Change a cluster configuration and StackLight configuration procedure.
Enable Logging
Select to deploy the StackLight logging stack. For details about the logging components, see Deployment architecture.
Note
The logging mechanism performance depends on the cluster log load. In case of a high load, you may need to increase the default resource requests and limits for
fluentdLogs. For details, see StackLight resource limits.HA Mode
Select to enable StackLight monitoring in High Availability (HA) mode. For differences between HA and non-HA modes, see Deployment architecture. If disabled, StackLight requires a Ceph cluster. To deploy a Ceph cluster, refer to Add a Ceph cluster.
StackLight Default Logs Severity Level
Log severity (verbosity) level for all StackLight components. The default value for this parameter is Default component log level that respects original defaults of each StackLight component. For details about severity levels, see StackLight log verbosity.
StackLight Component Logs Severity Level
The severity level of logs for a specific StackLight component that overrides the value of the StackLight Default Logs Severity Level parameter. For details about severity levels, see StackLight log verbosity. Expand the drop-down menu for a specific component to display its list of available log levels.
OpenSearch
Logstash Retention Time Removed in MOSK 24.1
Available if you select Enable Logging. Specifies the
logstash-*index retention time.Events Retention Time
Available if you select Enable Logging. Specifies the
kubernetes_events-*index retention time.Notifications Retention Time
Available if you select Enable Logging. Specifies the
notification-*index retention time.Persistent Volume Claim Size
Available if you select Enable Logging. The OpenSearch persistent volume claim size.
Collected Logs Severity Level
Available if you select Enable Logging. The minimum severity of all Container Cloud components logs collected in OpenSearch. For details about severity levels, see StackLight logging.
Prometheus
Retention Time
The Prometheus database retention period.
Retention Size
The Prometheus database retention size.
Persistent Volume Claim Size
The Prometheus persistent volume claim size.
Enable Watchdog Alert
Select to enable the Watchdog alert that fires as long as the entire alerting pipeline is functional.
Custom Alerts
Specify alerting rules for new custom alerts or upload a YAML file in the following exemplary format:
- alert: HighErrorRate expr: job:request_latency_seconds:mean5m{job="myjob"} > 0.5 for: 10m labels: severity: page annotations: summary: High request latency
For details, see Official Prometheus documentation: Alerting rules. For the list of the predefined StackLight alerts, see Operations Guide: StackLight alerts.
StackLight Email Alerts
Enable Email Alerts
Select to enable the StackLight email alerts.
Send Resolved
Select to enable notifications about resolved StackLight alerts.
Require TLS
Select to enable transmitting emails through TLS.
Email alerts configuration for StackLight
Fill out the following email alerts parameters as required:
To - the email address to send notifications to.
From - the sender address.
SmartHost - the SMTP host through which the emails are sent.
Authentication username - the SMTP user name.
Authentication password - the SMTP password.
Authentication identity - the SMTP identity.
Authentication secret - the SMTP secret.
StackLight Slack Alerts
Enable Slack alerts
Select to enable the StackLight Slack alerts.
Send Resolved
Select to enable notifications about resolved StackLight alerts.
Slack alerts configuration for StackLight
Fill out the following Slack alerts parameters as required:
API URL - The Slack webhook URL.
Channel - The channel to send notifications to, for example, #channel-for-alerts.
Click Create.
To monitor cluster readiness, see Verify cluster status.
Available since Container Cloud 2.24.0 (Cluster releases 14.0.0 and 15.0.1). Optional. Technology Preview. Enable the Linux Audit daemon auditd to monitor activity of cluster processes and prevent potential malicious activity.
Configuration for auditd
In the
Clusterobject orcluster.yaml.template, add the auditd parameters:spec: providerSpec: value: audit: auditd: enabled: <bool> enabledAtBoot: <bool> backlogLimit: <int> maxLogFile: <int> maxLogFileAction: <string> maxLogFileKeep: <int> mayHaltSystem: <bool> presetRules: <string> customRules: <string> customRulesX32: <text> customRulesX64: <text>
Configuration parameters for auditd:
enabledBoolean, default -
false. Enables theauditdrole to install the auditd packages and configure rules. CIS rules: 4.1.1.1, 4.1.1.2.enabledAtBootBoolean, default -
false. Configures grub to audit processes that can be audited even if they start up prior to auditd startup. CIS rule: 4.1.1.3.backlogLimitInteger, default - none. Configures the backlog to hold records. If during boot
audit=1is configured, the backlog holds 64 records. If more than 64 records are created during boot, auditd records will be lost with a potential malicious activity being undetected. CIS rule: 4.1.1.4.maxLogFileInteger, default - none. Configures the maximum size of the audit log file. Once the log reaches the maximum size, it is rotated and a new log file is created. CIS rule: 4.1.2.1.
maxLogFileActionString, default - none. Defines handling of the audit log file reaching the maximum file size. Allowed values:
keep_logs- rotate logs but never delete themrotate- add a cron job to compress rotated log files and keep maximum 5 compressed files.compress- compress log files and keep them under the/var/log/auditd/directory. Requiresauditd_max_log_file_keepto be enabled.
CIS rule: 4.1.2.2.
maxLogFileKeepInteger, default -
5. Defines the number of compressed log files to keep under the/var/log/auditd/directory. Requiresauditd_max_log_file_action=compress. CIS rules - none.mayHaltSystemBoolean, default -
false. Halts the system when the audit logs are full. Applies the following configuration:space_left_action = emailaction_mail_acct = rootadmin_space_left_action = halt
CIS rule: 4.1.2.3.
customRulesString, default - none. Base64-encoded content of the
60-custom.rulesfile for any architecture. CIS rules - none.customRulesX32String, default - none. Base64-encoded content of the
60-custom.rulesfile for thei386architecture. CIS rules - none.customRulesX64String, default - none. Base64-encoded content of the
60-custom.rulesfile for thex86_64architecture. CIS rules - none.presetRulesString, default - none. Comma-separated list of the following built-in preset rules:
accessactionsdeletedocker
identityimmutableloginsmac-policy
modulesmountsperm-modprivileged
scopesessionsystem-localetime-change
Since Container Cloud 2.28.0 (Cluster releases 17.3.0 and 16.3.0) in the Technology Preview scope, you can collect some of the preset rules indicated above as groups and use them in
presetRules:ubuntu-cis-rules- this group contains rules to comply with the Ubuntu CIS Benchmark recommendations, including the following CIS Ubuntu 20.04 v2.0.1 rules:scope- 5.2.3.1actions- same as 5.2.3.2time-change- 5.2.3.4system-locale- 5.2.3.5privileged- 5.2.3.6access- 5.2.3.7identity- 5.2.3.8
perm-mod- 5.2.3.9mounts- 5.2.3.10session- 5.2.3.11logins- 5.2.3.12delete- 5.2.3.13mac-policy- 5.2.3.14modules- 5.2.3.19
docker-cis-rules- this group contains rules to comply with Docker CIS Benchmark recommendations, including thedockerDocker CIS v1.6.0 rules 1.1.3 - 1.1.18.
You can also use two additional keywords inside
presetRules:none- select no built-in rules.all- select all built-in rules. When using this keyword, you can add the!prefix to a rule name to exclude some rules. You can use the!prefix for rules only if you add theallkeyword as the first rule. Place a rule with the!prefix only after theallkeyword.
Example configurations:
presetRules: none- disable all preset rulespresetRules: docker- enable only thedockerrulespresetRules: access,actions,logins- enable only theaccess,actions, andloginsrulespresetRules: ubuntu-cis-rules- enable all rules from theubuntu-cis-rulesgrouppresetRules: docker-cis-rules,actions- enable all rules from thedocker-cis-rulesgroup and theactionsrulepresetRules: all- enable all preset rulespresetRules: all,!immutable,!sessions- enable all preset rules exceptimmutableandsessions
CIS controls
4.1.3 (time-change)4.1.4 (identity)4.1.5 (system-locale)4.1.6 (mac-policy)4.1.7 (logins)4.1.8 (session)4.1.9 (perm-mod)4.1.10 (access)4.1.11 (privileged)4.1.12 (mounts)4.1.13 (delete)4.1.14 (scope)4.1.15 (actions)4.1.16 (modules)4.1.17 (immutable)Docker CIS controls
1.1.41.1.81.1.101.1.121.1.131.1.151.1.161.1.171.1.181.2.31.2.41.2.51.2.61.2.71.2.101.2.11
Optional. Colocate the OpenStack control plane with the managed cluster Kubernetes manager nodes by adding the following field to the
Clusterobject spec:spec: providerSpec: value: dedicatedControlPlane: false
Note
This feature is available as technical preview. Use such configuration for testing and evaluation purposes only.
Optional. Customize MetalLB speakers that are deployed on all Kubernetes nodes except master nodes by default. For details, see Configure node selectors for MetalLB speakers.
Configure the MetalLB parameters related to IP address allocation and announcement for load-balanced cluster services. For details, see Configure and verify MetalLB.
Proceed to Obtain and use details about network interfaces.
Note
Once you have created a MOSK cluster, some StackLight alerts may raise as false-positive until you deploy the Mirantis OpenStack environment.
Configure MetalLB¶
Before configuring subnets for a MOSK cluster, set up and verify MetalLB parameters as described in the following subsections.
This section describes how to set up and verify MetalLB parameters before configuring subnets for a MOSK cluster.
Caution
This section also applies to the bootstrap procedure of a management cluster with the following differences:
Instead of the
Clusterobject, configuretemplates/bm/cluster.yaml.template.Instead of the
MetalLBConfigobject, configuretemplates/bm/metallbconfig.yaml.template.Instead of creating specific IPAM objects such as
SubnetandL2Template(as well asRackandMultiRackClusterwhen using BGP configuration), add their settings totemplates/bm/ipam-objects.yaml.template.
Caution
The use of the MetalLBConfig object is mandatory after your
management cluster upgrade to the Cluster release 16.0.0.
The following rules and requirements apply to configuration of the
MetalLBConfig object:
Define one
MetalLBConfigobject per cluster.Define the following mandatory labels:
cluster.sigs.k8s.io/cluster-nameSpecifies the cluster name where the MetalLB address pool is used.
kaas.mirantis.com/regionSpecifies the region name of the cluster where the MetalLB address pool is used.
kaas.mirantis.com/providerSpecifies the provider of the cluster where the MetalLB address pool is used.
Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.Intersection of IP address ranges within any single MetalLB address pool is not allowed.
At least one MetalLB address pool must have the auto-assign policy enabled so that unannotated services can have load balancer IP addresses allocated for them.
When configuring multiple address pools with the auto-assign policy enabled, keep in mind that it is not determined in advance which pool of those multiple address pools is used to allocate an IP address for a particular unannotated service.
Note
You can optimize address announcement for load-balanced services
using the interfaces selector for the l2Advertisements object. This
selector allows for address announcement only on selected host interfaces.
For details, see MetalLBConfig spec.
Configuration rules for MetalLBConfigTemplate (obsolete since 24.2)
Caution
The MetalLBConfigTemplate object is deprecated since
MOSK 24.2 and unsupported since
MOSK 24.3. For details, see
Deprecation notes.
All rules described above for
MetalLBConfigalso apply toMetalLBConfigTemplate.Optional. Define one
MetalLBConfigTemplateobject per cluster. The use of this object withoutMetalLBConfigis not allowed.When using
MetalLBConfigTemplate:MetalLBConfigmust referenceMetalLBConfigTemplateby name:spec: templateName: <managed-metallb-template>
You can use
Subnetobjects for defining MetalLB address pools. Refer to MetalLB configuration guidelines for subnets for guidelines on configuring MetalLB address pools usingSubnetobjects.You can optimize address announcement for load-balanced services using the
interfacesselector for thel2Advertisementsobject. This selector allows for address announcement only on selected host interfaces. For details, see MetalLBConfigTemplate spec.
Available since MOSK 24.3
Note
The BGP configuration is not yet supported in the Container Cloud web UI. Meantime, use the CLI for this purpose. For details, see Configure and verify MetalLB using the CLI.
Read the MetalLB configuration guidelines described in Configuration rules for the ‘MetalLBConfig’ object.
Optional. Configure parameters related to MetalLB components life cycle such as deployment and update using the
metallbHelm chart values in theClusterspecsection. For example:Increase Pod resource limits for MetalLB as described in Limits for common components of any cluster type.
Configure Pod node selectors or affinity for MetalLB speakers as described in Configure node selectors for MetalLB speakers.
Log in to the Container Cloud web UI with the
writerpermissions.Switch to the required non-
defaultproject using the Switch Project action icon located on top of the main left-side navigation panel.Caution
Do not create a MOSK cluster in the
defaultproject (Kubernetes namespace), which is dedicated for the management cluster only. If no projects are defined, first create a newmoskproject as described in Create a project for MOSK clusters.In the Networks section, click the MetalLB Configs tab.
Click Create MetalLB Config.
Fill out the Create MetalLB Config form as required:
- Name
Name of the
MetalLBobject being created.
- Cluster
Name of the cluster that the
MetalLBobject is being created for
- IP Address Pools
List of MetalLB IP address pool descriptions that will be used to create the MetalLB
IPAddressPoolobjects. Click the + button on the right side of the section to add more objects.- Name
IP address pool name.
- Addresses
Comma-separated ranges of the IP addresses included into the address pool.
- Auto Assign
Enable auto-assign policy for unannotated services to have load balancer IP addresses allocated to them. At least one MetalLB address pool must have the auto-assign policy enabled.
- Service Allocation
IP address pool allocation to services. Click Edit to insert a service allocation object with required label selectors for services in the YAML format. For example:
serviceSelectors: - matchExpressions: - key: app.kubernetes.io/name operator: NotIn values: - dhcp-lb
For details on the MetalLB
IPAddressPoolobject type, see MetalLB documentation.- L2 Advertisements
List of
MetalLBL2Advertisementobjects to create MetalLBL2Advertisementobjects.The
l2Advertisementsobject allows defining interfaces to optimize the announcement. When you use theinterfacesselector, LB addresses are announced only on selected host interfaces.Mirantis recommends using the
interfacesselector if nodes use separate host networks for different types of traffic. The pros of such configuration are as follows: less spam on other interfaces and networks and limited chances to reach IP addresses of load-balanced services from irrelevant interfaces and networks.Caution
Interface names in the
interfaceslist must match those on the corresponding nodes.Add the following parameters:
- Name
Name of the
l2Advertisementsobject.
- Interfaces
Optional. Comma-separated list of interface names that must match the ones on the corresponding nodes. These names are defined in L2 templates that are linked to the selected cluster.
- IP Address Pools
Select the IP adress pool to use for the
l2Advertisementsobject.
- Node Selectors
Optional. Match labels and values for the Kubernetes node selector to limit the nodes announced as next hops for the
LoadBalancerIP. If you do not provide any labels, all nodes are announced as next hops.
For details on the MetalLB
L2Advertisementsobject type, see MetalLB documentation.
Click Create.
In Networks > MetalLB Configs, verify the status of the created MetalLB object:
Ready - object is operational.
Error - object is non-operational. Hover over the status to obtain details of the issue.
Note
To verify the object details, in Networks > MetalLB Configs, click the More action icon in the last column of the required object section and select MetalLB Config info.
Proceed to creating cluster subnets as described in Create subnets.
Optional. Configure parameters related to MetalLB components life cycle such as deployment and update using the
metallbHelm chart values in theClusterspecsection. For example:Increase Pod resource limits for MetalLB as described in Limits for common components of any cluster type.
Configure Pod node selectors or affinity for MetalLB speakers as described in Configure node selectors for MetalLB speakers.
Configure the MetalLB parameters related to IP address allocation and announcement for load-balanced cluster services:
Since MOSK 24.2
Mandatory after a management cluster upgrade to the Cluster release 17.2.0. Recommended and default since MOSK 24.2.
Create the
MetalLBConfigobject:For configuration rules and requirements, see Configuration rules for the ‘MetalLBConfig’ object.
For the object description, see MetalLBConfig.
For configuration examples that use
MetalLBConfig, see the following sections in the Container Cloud documentation:Example of a complete template configuration for cluster creation (the MetalLB configuration objects step)
In the Technology Preview scope, you can use BGP for announcement of external addresses of Kubernetes load-balanced services for a MOSK cluster. To configure the BGP announcement mode for MetalLB, use
MetalLBConfigobject.The use of BGP is required to announce IP addresses for load-balanced services when using MetalLB on nodes that are distributed across multiple racks. In this case, setting of
rack-idlabels on nodes is required, they are used in node selectors forBGPPeer,BGPAdvertisement, or both MetalLB objects to properly configure BGP connections from each node.Configuration example of the
Machineobject for the BGP announcement modeapiVersion: cluster.k8s.io/v1alpha1 kind: Machine metadata: name: test-cluster-compute-1 namespace: mosk-ns labels: cluster.sigs.k8s.io/cluster-name: test-cluster ipam/RackRef: rack-1 # reference to the "rack-1" Rack kaas.mirantis.com/provider: baremetal spec: providerSpec: value: ... nodeLabels: - key: rack-id # node label can be used in "nodeSelectors" inside value: rack-1 # "BGPPeer" and/or "BGPAdvertisement" MetalLB objects ...
Configuration example of the
MetalLBConfigobject for the BGP announcement modeapiVersion: ipam.mirantis.com/v1alpha1 kind: MetalLBConfig metadata: name: test-cluster-metallb-config namespace: mosk-ns labels: cluster.sigs.k8s.io/cluster-name: test-cluster kaas.mirantis.com/provider: baremetal spec: ... bgpPeers: - name: svc-peer-1 spec: holdTime: 0s keepaliveTime: 0s peerAddress: 10.77.42.1 peerASN: 65100 myASN: 65101 nodeSelectors: - matchLabels: rack-id: rack-1 # references the nodes having # the "rack-id=rack-1" label bgpAdvertisements: - name: services spec: aggregationLength: 32 aggregationLengthV6: 128 ipAddressPools: - services peers: - svc-peer-1 ...
The
bgpPeersandbgpAdvertisementsfields are used to configure BGP announcement instead ofl2Advertisements.The use of BGP for announcement also allows for better balancing of service traffic between cluster nodes as well as gives more configuration control and flexibility for infrastructure administrators. For configuration examples, refer to MetalLB configuration examples. For configuration procedure, refer to Configure BGP announcement for cluster API LB address.
Since MOSK 23.2
Select from the following options:
Deprecated since MOSK 24.2 and unsupported since MOSK 24.3. Mandatory after a management cluster upgrade to the Cluster release 16.0.0. Recommended and default since MOSK 23.2 in the Technology Preview scope. Create
MetalLBConfigandMetalLBConfigTemplateobjects. This method allows using theSubnetobject to define MetalLB address pools.For configuration rules and requirements, see Configuration rules for the ‘MetalLBConfig’ object.
For objects description, see MetalLBConfig and MetalLBConfigTemplate.
For configuration examples that use
MetalLBConfigandMetalLBConfigTemplate, see the following sections in Container Cloud documentation:Example of a complete template configuration for cluster creation (the MetalLB configuration objects step)
Since MOSK 23.2.2, in the Technology Preview scope, you can use BGP for announcement of external addresses of Kubernetes load-balanced services for a MOSK cluster. To configure the BGP announcement mode for MetalLB, use
MetalLBConfigandMetalLBConfigTemplateobjects.The use of BGP is required to announce IP addresses for load-balanced services when using MetalLB on nodes that are distributed across multiple racks. In this case, setting of
rack-idlabels on nodes is required, they are used in node selectors forBGPPeer,BGPAdvertisement, or both MetalLB objects to properly configure BGP connections from each node.Configuration example of the
Machineobject for the BGP announcement modeapiVersion: cluster.k8s.io/v1alpha1 kind: Machine metadata: name: test-cluster-compute-1 namespace: mosk-ns labels: cluster.sigs.k8s.io/cluster-name: test-cluster ipam/RackRef: rack-1 # reference to the "rack-1" Rack kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one spec: providerSpec: value: ... nodeLabels: - key: rack-id # node label can be used in "nodeSelectors" inside value: rack-1 # "BGPPeer" and/or "BGPAdvertisement" MetalLB objects ...
Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.Configuration example of the
MetalLBConfigTemplateobject for the BGP announcement modeapiVersion: ipam.mirantis.com/v1alpha1 kind: MetalLBConfigTemplate metadata: name: test-cluster-metallb-config-template namespace: mosk-ns labels: cluster.sigs.k8s.io/cluster-name: test-cluster kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one spec: templates: ... bgpPeers: | - name: svc-peer-1 spec: peerAddress: 10.77.42.1 peerASN: 65100 myASN: 65101 nodeSelectors: - matchLabels: rack-id: rack-1 # references the nodes having # the "rack-id=rack-1" label bgpAdvertisements: | - name: services spec: ipAddressPools: - services peers: - svc-peer-1 ...
Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.The
bgpPeersandbgpAdvertisementsfields are used to configure BGP announcement instead ofl2Advertisements.The use of BGP for announcement also allows for better balancing of service traffic between cluster nodes as well as gives more configuration control and flexibility for infrastructure administrators. For configuration examples, refer to MetalLBConfigTemplate. For configuration procedure, refer to Configure BGP announcement for cluster API LB address.
Not recommended. Configure the
configInlinevalue in the MetalLB chart of theClusterobject.Warning
This option is deprecated since MOSK 23.2 and is removed during the management cluster upgrade to the Cluster release 16.0.0, which is introduced in Container Cloud 2.25.0.
Therefore, this option becomes unavailable on MOSK 23.2 clusters after the parent management cluster upgrade to 2.25.0.
Not recommended. Configure the
Subnetobjects withoutMetalLBConfigTemplate.Warning
This option is deprecated since MOSK 23.2 and is removed during the management cluster upgrade to the Cluster release 16.0.0, which is introduced in Container Cloud 2.25.0.
Therefore, this option becomes unavailable on MOSK 23.2 clusters after the parent management cluster upgrade to 2.25.0.
Caution
If the
MetalLBConfigobject is not used for MetalLB configuration related to address allocation and announcement for load-balanced services, then automated migration applies during cluster creation or update to MOSK 23.2.During automated migration, the
MetalLBConfigandMetalLBConfigTemplateobjects are created and contents of the MetalLB chartconfigInlinevalue is converted to the parameters of theMetalLBConfigTemplateobject.Any change to the
configInlinevalue made on a MOSK 23.2 cluster will be reflected in theMetalLBConfigTemplateobject.This automated migration is removed during your management cluster upgrade to the Cluster release 16.0.0, which is introduced in Container Cloud 2.25.0, together with the possibility to use the
configInlinevalue of the MetalLB chart. After that, any changes in MetalLB configuration related to address allocation and announcement for load-balanced services are applied using theMetalLBConfig,MetalLBConfigTemplate, andSubnetobjects only.Before MOSK 23.2
Select from the following options:
Configure
Subnetobjects. For details, see MetalLB configuration guidelines for subnets.Configure the
configInlinevalue for the MetalLB chart in theClusterobject.Configure both the
configInlinevalue for the MetalLB chart andSubnetobjects.The resulting MetalLB address pools configuration will contain address ranges from both cluster specification and
Subnetobjects. All address ranges for L2 address pools will be aggregated into a single L2 address pool and sorted as strings.
Changes to be applied since MOSK 23.2
The configuration options above become deprecated since 23.2, and automated migration of MetalLB parameters applies during cluster creation or update to MOSK 23.2.
During automated migration, the
MetalLBConfigandMetalLBConfigTemplateobjects are created and contents of the MetalLB chartconfigInlinevalue is converted to the parameters of theMetalLBConfigTemplateobject.Any change to the
configInlinevalue made on a MOSK 23.2 cluster will be reflected in theMetalLBConfigTemplateobject.This automated migration is removed during your management cluster upgrade to Container Cloud 2.25.0 together with the possibility to use the
configInlinevalue of the MetalLB chart. After that, any changes in MetalLB configuration related to address allocation and announcement for load-balanced services will be applied using theMetalLBConfigTemplateandSubnetobjects only.Verify the current MetalLB configuration:
Verify the MetalLB configuration that is stored in
MetalLBobjects:kubectl -n metallb-system get ipaddresspools,l2advertisements
The example system output:
NAME AGE ipaddresspool.metallb.io/default 129m ipaddresspool.metallb.io/services-pxe 129m NAME AGE l2advertisement.metallb.io/default 129m l2advertisement.metallb.io/services-pxe 129m
Verify one of the listed above
MetalLBobjects:kubectl -n metallb-system get <object> -o json | jq '.spec'
The example system output for
ipaddresspoolobjects:$ kubectl -n metallb-system get ipaddresspool.metallb.io/default -o json | jq '.spec' { "addresses": [ "10.0.11.61-10.0.11.80" ], "autoAssign": true, "avoidBuggyIPs": false } $ kubectl -n metallb-system get ipaddresspool.metallb.io/services-pxe -o json | jq '.spec' { "addresses": [ "10.0.0.61-10.0.0.70" ], "autoAssign": false, "avoidBuggyIPs": false }
Verify the MetalLB configuration that is stored in the
ConfigMapobject:kubectl -n metallb-system get cm metallb -o jsonpath={.data.config}
An example of a successful output:
address-pools: - name: default protocol: layer2 addresses: - 10.0.11.61-10.0.11.80 - name: services-pxe protocol: layer2 auto-assign: false addresses: - 10.0.0.61-10.0.0.70
The
auto-assignparameter will be set tofalsefor all address pools except thedefaultone. So, a particular service will get an address from such an address pool only if theServiceobject has a specialmetallb.universe.tf/address-poolannotation that points to the specific address pool name.Note
It is expected that every Kubernetes service on a management cluster will be assigned to one of the address pools. Current consideration is to have two MetalLB address pools:
services-pxeis a reserved address pool name to use for the Kubernetes services in the PXE network (Ironic API, HTTP server, caching server).defaultis an address pool to use for all other Kubernetes services in the management network. No annotation is required on theServiceobjects in this case.
Proceed to creating cluster subnets as described in Create subnets.
By default, MetalLB speakers are deployed on all Kubernetes nodes except master nodes. You can configure MetalLB to run its speakers on a particular set of nodes. This decreases the number of nodes that should be connected to external network. In this scenario, only a few nodes are exposed for ingress traffic from the outside world.
To customize a node selector for a MetalLB speaker:
Using
kubeconfigof the Container Cloud management cluster, open the MOSKClusterobject for editing:kubectl --kubeconfig <pathToManagementClusterKubeconfig> -n <OSClusterNamespace> edit cluster <OSClusterName>
In the
spec:providerSpec:value:helmReleasessection, add thespeaker.nodeSelectorfield formetallb:spec: ... providerSpec: value: ... helmReleases: - name: metallb values: ... speaker: nodeSelector: metallbSpeakerEnabled: "true"
The
metallbSpeakerEnabled: "true"parameter in this example is the label on Kubernetes nodes where MetalLB speakers will be deployed. It can be an already existing node label or a new one.Note
The issue [24435] MetalLB speaker fails to announce the LB IP for the Ingress service, which is related to collocation of MetalLB speakers and the OpenStack Ingress service pods is addressed in MOSK 22.5. For details, see Release Notes: Set externalTrafficPolicy=Local for the OpenStack Ingress service.
You can add user-defined labels to nodes using the
nodeLabelsfield.This field contains the list of node labels to be attached to a node for the user to run certain components on separate cluster nodes. The list of allowed node labels is located in the
Clusterobject statusproviderStatus.releaseRef.current.allowedNodeLabelsfield.If the
valuefield is not defined inallowedNodeLabels, a label can have any value. For example:allowedNodeLabels: - displayName: Stacklight key: stacklight
Before or after a machine deployment, add the required label from the allowed node labels list with the corresponding value to
spec.providerSpec.value.nodeLabelsinmachine.yaml. For example:nodeLabels: - key: stacklight value: enabled
Adding of a node label that is not available in the list of allowed node labels is restricted.
Note
Consider this section as obsolete since MOSK
24.2 due to the MetalLBConfigTemplate object deprecation. For details,
see Deprecation Notes: MetalLBConfigTemplate object.
Caution
This section also applies to the bootstrap procedure of a
management cluster with the following difference: instead of creating the
Subnet object, add its configuration to
ipam-objects.yaml.template located in kaas-bootstrap/templates/bm/.
The Kubernetes Subnet object is created for a management cluster from
templates during bootstrap.
Each Subnet object can define either a MetalLB address range or MetalLB
address pool. A MetalLB address pool may contain one or several address
ranges. The following rules apply to creation of address ranges or pools:
To designate a subnet as a MetalLB address pool or range, use the
ipam/SVC-MetalLBlabel key. Set the label value to"1".The object must contain the
cluster.sigs.k8s.io/<cluster-name>label to reference the name of the target cluster where the MetalLB address pool is used.You may create multiple subnets with the
ipam/SVC-MetalLBlabel to define multiple IP address ranges or multiple address pools for MetalLB in the cluster.The IP addresses of the MetalLB address pool are not assigned to the interfaces on hosts. This subnet is virtual. Do not include such subnets to the L2 template definitions for your cluster.
If a
Subnetobject defines a MetalLB address range, no additional object properties are required.You can use any number of
Subnetobjects that define a single MetalLB address range. In this case, all address ranges are aggregated into a single MetalLB L2 address pool namedserviceshaving the auto-assign policy enabled.Intersection of IP address ranges within any single MetalLB address pool is not allowed.
The bare metal provider verifies intersection of IP address ranges. If it detects intersection, the MetalLB configuration is blocked and the provider logs contain corresponding error messages.
Use the following labels to identify the Subnet object as a MetalLB
address pool and configure the name and protocol for that address pool.
All labels below are mandatory for the Subnet object that configures
a MetalLB address pool.
Label |
Description |
|---|---|
Labels to link |
Note The |
|
Defines that the |
|
Every address pool must have a distinct name. The A bootstrap cluster also uses the |
|
Configures the auto-assign policy of an address pool. Boolean. Caution For the address pools defined using the MetalLB Helm chart
values in the For any service that does not have a specific MetalLB annotation
configured, MetalLB allocates external IPs from arbitrary address
pools that have the auto-assign policy set to Only for the service that has a specific MetalLB annotation with
the address pool name, MetalLB allocates external IPs from the address
pool having the auto-assign policy set to |
|
Sets the address pool protocol.
The only supported value is |
Caution
Do not set the same address pool name for two or more
Subnet objects. Otherwise, the corresponding MetalLB address pool
configuration fails with a warning message in the bare metal provider log.
Caution
For the auto-assign policy, the following configuration rules apply:
At least one MetalLB address pool must have the auto-assign policy enabled so that unannotated services can have load balancer IPs allocated for them. To satisfy this requirement, either configure one of address pools using the
Subnetobject withmetallb/address-pool-auto-assign: "true"or configure address range(s) using theSubnetobject(s) withoutmetallb/address-pool-*labels.When configuring multiple address pools with the auto-assign policy enabled, keep in mind that it is not determined in advance which pool of those multiple address pools is used to allocate an IP for a particular unannotated service.
Configure BGP announcement for cluster API LB address¶
Available since MOSK 23.2.2 TechPreview
When you create a MOSK cluster with the multi-rack topology, where Kubernetes masters are distributed across multiple racks without an L2 layer extension between them, you must configure BGP announcement of the cluster API load balancer address.
For clusters where Kubernetes masters are in the same rack or with an L2 layer extension between masters, you can configure either BGP or L2 (ARP) announcement of the cluster API load balancer address. The L2 (ARP) announcement is used by default and its configuration is covered in Create a managed bare metal cluster.
Caution
Create Rack and MultiRackCluster objects, which are
described in the below procedure, before initiating the provisioning
of master nodes to ensure that both BGP and netplan configurations
are applied simultaneously during the provisioning process.
To enable the use of BGP announcement for the cluster API LB address:
In the
Clusterobject, set theuseBGPAnnouncementparameter totrue:spec: providerSpec: value: useBGPAnnouncement: true
Create the
MultiRackClusterobject that is mandatory when configuring BGP announcement for the cluster API LB address. This object enables you to set cluster-wide parameters for configuration of BGP announcement.In this scenario, the
MultiRackClusterobject must be bound to the correspondingClusterobject using thecluster.sigs.k8s.io/cluster-namelabel.Container Cloud uses the
birdBGP daemon for announcement of the cluster API LB address. For this reason, set the correspondingbgpdConfigFileNameandbgpdConfigFilePathparameters in theMultiRackClusterobject, so thatbirdcan locate the configuration file. For details, see the configuration example below.The
bgpdConfigTemplateobject contains the default configuration file template for thebirdBGP daemon, which you can override inRackobjects.The
defaultPeerparameter contains default parameters of the BGP connection from master nodes to infrastructure BGP peers, which you can override inRackobjects.Configuration example for
MultiRackClusterapiVersion: ipam.mirantis.com/v1alpha1 kind: MultiRackCluster metadata: name: multirack-test-cluster namespace: mosk-ns labels: cluster.sigs.k8s.io/cluster-name: test-cluster kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one spec: bgpdConfigFileName: bird.conf bgpdConfigFilePath: /etc/bird bgpdConfigTemplate: | ... defaultPeer: localASN: 65101 neighborASN: 65100 neighborIP: "" password: deadbeef
Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.For the object description, see MultiRackCluster.
Create the
Rackobject(s). This object is mandatory when configuring BGP announcement for the cluster API LB address and it allows you to configure BGP announcement parameters for each rack.In this scenario,
Rackobjects must be bound toMachineobjects corresponding to master nodes of the cluster. EachRackobject describes the configuration for thebirdBGP daemon used to announce the cluster API LB address from a particular master node or from several master nodes in the same rack.The
Rackobject fields are described in Rack.Set a reference to the
Rackobject used to configure thebirdBGP daemon for a particular master node to announce the cluster API LB IP:In the
Machineobjects for all master nodes, set theipam/RackReflabel with the value equal to the name of the correspondingRackobject. For example:apiVersion: cluster.k8s.io/v1alpha1 kind: Machine metadata: labels: ipam/RackRef: rack-master-1 # reference to the "rack-master-1" Rack ...
In the
BareMetalHostobjects for all cluster nodes, set theipam.mirantis.com/rack-refannotation with the value equal to the name of the correspondingRackobject. For example:apiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: annotations: ipam.mirantis.com/rack-ref: rack-master-1 # reference to the "rack-master-1" Rack ...
Optional. Using the
Machineobject, define therack-idnode label that is not used for BGP announcement of the cluster API LB IP but can be used for MetalLB.The
rack-idnode label is required for MetalLB node selectors when MetalLB is used to announce LB IP addresses on nodes that are distributed across multiple racks. In this scenario, the L2 (ARP) announcement mode cannot be used for MetalLB because master nodes are in different L2 segments. So, the BGP announcement mode must be used for MetalLB, and node selectors are required to properly configure BGP connections from each node. See Configure and verify MetalLB for details.The
L2Templateobject includes thelointerface configuration to set the IP address for thebirdBGP daemon that will be advertised as the cluster API LB address. The{{ cluster_api_lb_ip }}function is used innpTemplateto obtain the cluster API LB address value.Configuration example for
RackapiVersion: ipam.mirantis.com/v1alpha1 kind: Rack metadata: name: rack-master-1 namespace: mosk-ns labels: cluster.sigs.k8s.io/cluster-name: test-cluster kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one spec: bgpdConfigTemplate: | # optional ... peeringMap: lcm-rack-control-1: peers: - neighborIP: 10.77.31.2 # "localASN" & "neighborASN" are taken from - neighborIP: 10.77.31.3 # "MultiRackCluster.spec.defaultPeer" if # not set here
Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.Configuration example for
MachineapiVersion: cluster.k8s.io/v1alpha1 kind: Machine metadata: name: test-cluster-master-1 namespace: mosk-ns annotations: metal3.io/BareMetalHost: mosk-ns/test-cluster-master-1 labels: cluster.sigs.k8s.io/cluster-name: test-cluster cluster.sigs.k8s.io/control-plane: controlplane hostlabel.bm.kaas.mirantis.com/controlplane: controlplane ipam/RackRef: rack-master-1 # reference to the "rack-master-1" Rack kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one spec: providerSpec: value: kind: BareMetalMachineProviderSpec apiVersion: baremetal.k8s.io/v1alpha1 hostSelector: matchLabels: kaas.mirantis.com/baremetalhost-id: test-cluster-master-1 l2TemplateSelector: name: test-cluster-master-1 nodeLabels: # optional. it is not used for BGP announcement - key: rack-id # of the cluster API LB IP but it can be used value: rack-master-1 # for MetalLB if "nodeSelectors" are required ...
Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.Configuration example for
L2TemplateapiVersion: ipam.mirantis.com/v1alpha1 kind: L2Template metadata: labels: cluster.sigs.k8s.io/cluster-name: test-cluster kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: test-cluster-master-1 namespace: mosk-ns spec: ... l3Layout: - subnetName: lcm-rack-control-1 # this network is referenced scope: namespace # in the "rack-master-1" Rack - subnetName: ext-rack-control-1 # optional. this network is used scope: namespace # for k8s services traffic and # MetalLB BGP connections ... npTemplate: | ... ethernets: lo: addresses: - {{ cluster_api_lb_ip }} # function for cluster API LB IP dhcp4: false dhcp6: false ...
Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.The configuration example for the scenario where Kubernetes masters are in the same rack or with an L2 layer extension between masters is described in Single rack configuration example.
The configuration example for the scenario where Kubernetes masters are distributed across multiple racks without L2 layer extension between them is described in Multiple rack configuration example.
Obtain and use details about network interfaces¶
To simplify operations with L2 templates, before you start creating them, inspect the general workflow of a network interface name gathering and processing.
Network interface naming workflow:
The operator creates a
BareMetalHostInventoryobject.Note
Before update of the management cluster to Container Cloud 2.29.0 (Cluster release 16.4.0), instead of
BareMetalHostInventory, use theBareMetalHostobject. For details, see BareMetalHost resource.Caution
While the Cluster release of the management cluster is 16.4.0,
BareMetalHostInventoryoperations are allowed tom:kaas@management-adminonly. This limitation is lifted once the management cluster is updated to the Cluster release 16.4.1 or later.The
BareMetalHostInventoryobject executes theintrospectionstage and becomesready.The operator collects information about NIC count, naming, and so on for further changes in the mapping logic.
At this stage, the NICs order in the object may randomly change during each introspection, but the NICs names are always the same. For more details, see Predictable Network Interface Names.
For example:
# Example commands: # kubectl -n managed-ns get bmh baremetalhost1 -o custom-columns='NAME:.metadata.name,STATUS:.status.provisioning.state' # NAME STATE # baremetalhost1 ready # kubectl -n managed-ns get bmh baremetalhost1 -o yaml # Example output: apiVersion: metal3.io/v1alpha1 kind: BareMetalHost ... status: ... nics: - ip: fe80::ec4:7aff:fe6a:fb1f%eno2 mac: 0c:c4:7a:6a:fb:1f model: 0x8086 0x1521 name: eno2 pxe: false - ip: fe80::ec4:7aff:fe1e:a2fc%ens1f0 mac: 0c:c4:7a:1e:a2:fc model: 0x8086 0x10fb name: ens1f0 pxe: false - ip: fe80::ec4:7aff:fe1e:a2fd%ens1f1 mac: 0c:c4:7a:1e:a2:fd model: 0x8086 0x10fb name: ens1f1 pxe: false - ip: 192.168.1.151 # Temp. PXE network adress mac: 0c:c4:7a:6a:fb:1e model: 0x8086 0x1521 name: eno1 pxe: true ...
The operator selects from the following options:
Create an
L2Templateobject with theifMappingconfiguration. For details, see Create L2 templates.Create a
Machineobject, with thel2TemplateIfMappingOverrideconfiguration. For details, see Override network interfaces naming and order.
The operator creates a
MachineorSubnetobject.The
baremetal-providerservice links theMachineobject to theBareMetalHostInventoryobject.The
kaas-ipamandbaremetal-providerservices collect hardware information from theBareMetalHostInventoryobject and use it to configure host networking and services.The
kaas-ipamservice:Spawns the
IpamHostobject.Renders the
L2Templateobject.Spawns the
ipaddrobject.Updates the
IpamHostobject status with all rendered and linked information.
The
baremetal-providerservice collects the rendered networking information from theIpamHostobjectThe
baremetal-providerservice proceeds with theIpamHostobject provisioning.
Now proceed to Create subnets.
Create subnets¶
After creating the MetalLB configuration as described in Configure and verify MetalLB
and before creating L2 templates, ensure that you have the required subnets
that can be used in the L2 template to allocate IP addresses for the
MOSK cluster nodes.
Where required, create a number of subnets for a particular project
using the Subnet CR. A subnet has the following logical scopes:
Each subnet used in an L2 template has its logical scope that is set using the
scope parameter in the corresponding L2Template.spec.l3Layout section.
One of the following logical scopes is used for each subnet referenced in an
L2 template:
global - CR uses the
defaultnamespace. A subnet can be used for any cluster located in any project.namespaced - CR uses the namespace that corresponds to a particular project where MOSK clusters are located. A subnet can be used for any cluster located in the same project.
cluster - Unsupported since Container Cloud 2.28.0 (Cluster releases 17.3.0 and 16.3.0). CR uses the namespace where the referenced cluster is located. A subnet is only accessible to the cluster that
L2Template.metadata.labels:cluster.sigs.k8s.io/cluster-name(mandatory since MOSK 23.3) orL2Template.spec.clusterRef(deprecated since MOSK 23.3) refers to. TheSubnetobjects with theclusterscope will be created for every new cluster.
Note
The use of the ipam/SVC-MetalLB label in Subnet objects
is unsupported as part of the MetalLBConfigTemplate object deprecation
since Container Cloud 2.27.0 (Cluster releases 17.2.0 and 16.2.0). No
actions are required for existing objects. A Subnet object containing
this label will be ignored by baremetal-provider after cluster update
to the mentioned Cluster releases.
You can have subnets with the same name in different projects. In this case, the subnet that has the same project as the cluster will be used. One L2 template may often reference several subnets, those subnets may have different scopes in this case.
The IP address objects (IPaddr CR) that are allocated from subnets
always have the same project as their corresponding IpamHost objects,
regardless of the subnet scope.
You can create subnets using either the Container Cloud web UI or CLI.
Any Subnet object may contain ipam/SVC-<serviceName> labels.
All IP addresses allocated from the Subnet object that has service labels
defined inherit those labels.
When a particular IpamHost uses IP addresses allocated from such labeled
Subnet objects, the ServiceMap field in IpamHost.Status
contains information about which IPs and interfaces correspond to which service
labels (that have been set in the Subnet objects). Using ServiceMap,
you can understand what IPs and interfaces of a particular host are used
for network traffic of a given service.
Container Cloud uses the following service labels that allow using of the specific subnets for particular Container Cloud services:
ipam/SVC-k8s-lcmipam/SVC-ceph-clusteripam/SVC-ceph-publicipam/SVC-dhcp-rangeipam/SVC-MetalLBUnsupported since 24.3ipam/SVC-LBhost
Caution
The use of the ipam/SVC-k8s-lcm label is mandatory for every
cluster.
Important
A label value is not mandatory and can be empty but it must
match the value in the related L2Template object, in which the
corresponding subnet is used. Otherwise, network configuration for related
hosts will not be rendered due to not found subnets.
You can also add custom service labels to the Subnet objects the same way
you add Container Cloud service labels. The mapping of IPs and interfaces to
the defined services is displayed in IpamHost.Status.ServiceMap.
You can assign multiple service labels to one network. You can also assign the
ceph-* and dhcp-range services to multiple networks. In the latter
case, the system sorts the IP addresses in the ascending order:
serviceMap:
ipam/SVC-ceph-cluster:
- ifName: ceph-br2
ipAddress: 10.0.10.11
- ifName: ceph-br1
ipAddress: 10.0.12.22
ipam/SVC-ceph-public:
- ifName: ceph-public
ipAddress: 10.1.1.15
ipam/SVC-k8s-lcm:
- ifName: k8s-lcm
ipAddress: 10.0.1.52
You can add service labels during creation of subnets as described in Create subnets.
After creating the MetalLB configuration as described in Configure and verify MetalLB and before creating an L2 template, create the required subnets to use in the L2 template to allocate IP addresses for the managed cluster nodes.
To create subnets for a managed cluster using web UI:
Log in to the Container Cloud web UI with the
operatorpermissions.Switch to the required non-
defaultproject using the Switch Project action icon located on top of the main left-side navigation panel.Caution
Do not create a MOSK cluster in the
defaultproject (Kubernetes namespace), which is dedicated for the management cluster only. If no projects are defined, first create a newmoskproject as described in Create a project for MOSK clusters.Create basic cluster settings as described in Create a managed bare metal cluster.
Select one of the following options:
In the left sidebar, navigate to Networks. The Subnets tab opens.
Click Create Subnet.
Fill out the Create subnet form as required:
- Name
Subnet name.
- Subnet Type
Subnet type:
- DHCP Optional
DHCP subnet that configures DHCP address ranges used by the DHCP server on the management cluster. For details, see Configure multiple DHCP address ranges.
- LB
Cluster API LB subnet.
- LCM
LCM subnet(s).
- Storage access Optional
Available in the web UI since Container Cloud 2.28.0 (17.3.0 and 16.3.0). Storage access subnet.
- Storage replication Optional
Available in the web UI since Container Cloud 2.28.0 (17.3.0 and 16.3.0). Storage replication subnet.
- Custom Optional
Custom subnet. For example, external or for Kubernetes workloads. For details, see optional steps in Create subnets for a managed cluster using CLI.
- MetalLB
Services subnet(s).
Warning
Since Container Cloud 2.28.0 (Cluster releases 17.3.0 and 16.3.0), disregard this parameter during subnet creation. Configure MetalLB separately as described in Configure and verify MetalLB.
This parameter is removed from the Container Cloud web UI in Container Cloud 2.29.0 (Cluster releases 17.4.0 and 16.4.0).
For description of subnet types in a managed cluster, see MOSK cluster networking.
- Cluster
Cluster name that the subnet is being created for. Not required only for the DHCP subnet.
- CIDR
A valid IPv4 address of the subnet in the CIDR notation, for example, 10.11.0.0/24.
- Include Ranges Optional
A comma-separated list of IP address ranges within the given CIDR that should be used in the allocation of IPs for nodes. The gateway, network, broadcast, and DNSaddresses will be excluded (protected) automatically if they intersect with one of the range. The IPs outside the given ranges will not be used in the allocation. Each element of the list can be either an interval
10.11.0.5-10.11.0.70or a single address10.11.0.77.Warning
Do not use values that are out of the given CIDR.
- Exclude Ranges Optional
A comma-separated list of IP address ranges within the given CIDR that should not be used in the allocation of IPs for nodes. The IPs within the given CIDR but outside the given ranges will be used in the allocation. The gateway, network, broadcast, and DNS addresses will be excluded (protected) automatically if they are included in the CIDR. Each element of the list can be either an interval
10.11.0.5-10.11.0.70or a single address10.11.0.77.Warning
Do not use values that are out of the given CIDR.
- Gateway Optional
A valid IPv4 gateway address, for example, 10.11.0.9. Does not apply to the MetalLB subnet.
- Nameservers
IP addresses of nameservers separated by a comma. Does not apply to the DHCP and MetalLB subnet types.
- Use whole CIDR
Optional. Select to use the whole IPv4 address range that is set in the CIDR field. Useful when defining single IP address (/32), for example, in the Cluster API load balancer (LB) subnet.
If not set, the network address and broadcast address in the IP subnet are excluded from the address allocation.
- Labels
Key-value pairs attached to the selected subnet:
Caution
The values of the created subnet labels must match the ones in the
spec.l3Layoutsection of the correspondingL2Templateobject.Optional user-defined labels to distinguish different subnets of the same type. For an example of user-defined labels, see Expand IP addresses capacity in an existing cluster.
The following special values define the storage subnets:
ipam/SVC-ceph-clusteripam/SVC-ceph-public
For more examples of label usage, see Service labels and their life cycle and Create subnets for a managed cluster using CLI.
Click Add a label and assign the first custom label with the required name and value. To assign consecutive labels, use the + button located in the right side of the Labels section.
MetalLB:
Warning
Since Container Cloud 2.28.0 (Cluster releases 17.3.0 and 16.3.0), disregard this label during subnet creation. Configure MetalLB separately as described in Configure and verify MetalLB.
The label will be removed from the Container Cloud web UI in one of the following releases.
metallb/address-pool-nameName of the subnet address pool. Exemplary values:
services,default,external,services-pxe.The latter label is dedicated for management clusters only. For details about address pool names of a management cluster, see Separate PXE and management networks.
metallb/address-pool-auto-assignEnables automatic assignment of address pool. Boolean.
metallb/address-pool-protocolDefines the address pool protocol. Possible values:
layer2- announcement using the ARP protocol.bgp- announcement using the BGP protocol. Technology Preview.
For description of these protocols, refer to the MetalLB documentation.
Click Create.
In the Networks tab, verify the status of the created subnet:
Ready - object is operational.
Error - object is non-operational. Hover over the status to obtain details of the issue.
Note
To verify subnet details, in the Networks tab, click the More action icon in the last column of the required subnet and select Subnet info.
In the Clusters tab, click the required cluster and scroll down to the Subnets section.
Click Add Subnet.
Fill out the Add new subnet form as required:
- Subnet Name
Subnet name.
- CIDR
A valid IPv4 CIDR, for example, 10.11.0.0/24.
- Include Ranges Optional
A comma-separated list of IP address ranges within the given CIDR that should be used in the allocation of IPs for nodes. The gateway, network, broadcast, and DNSaddresses will be excluded (protected) automatically if they intersect with one of the range. The IPs outside the given ranges will not be used in the allocation. Each element of the list can be either an interval
10.11.0.5-10.11.0.70or a single address10.11.0.77.Warning
Do not use values that are out of the given CIDR.
- Exclude Ranges Optional
A comma-separated list of IP address ranges within the given CIDR that should not be used in the allocation of IPs for nodes. The IPs within the given CIDR but outside the given ranges will be used in the allocation. The gateway, network, broadcast, and DNS addresses will be excluded (protected) automatically if they are included in the CIDR. Each element of the list can be either an interval
10.11.0.5-10.11.0.70or a single address10.11.0.77.Warning
Do not use values that are out of the given CIDR.
- Gateway Optional
A valid gateway address, for example, 10.11.0.9.
Click Create.
Proceed to creating L2 templates as described in Create L2 templates.
After creating the MetalLB configuration as described in Configure and verify MetalLB and before creating an L2 template, create the required subnets to use in the L2 template to allocate IP addresses for the managed cluster nodes.
Ensure that the underlay network in your cluster satisfies the requirements listed in Networking.
Configure all required VLANs on the network switches as described in Network types.
Configure gateway IP addresses on the ToR switches. Set up static routes as described in Underlay networking: routing configuration.
Create a cluster using one of the following options:
Container Cloud web UI: Create a managed bare metal cluster
API Reference: Cluster resource
Log in to a local machine where your management cluster
kubeconfigis located and wherekubectlis installed.Note
The management cluster
kubeconfigis created during the last stage of the management cluster bootstrap.Create the
subnet.yamlfile with a number of global or namespaced subnets depending on the configuration of your cluster:kubectl --kubeconfig <pathToManagementClusterKubeconfig> apply -f <SubnetFileName.yaml>
Note
In the command above and in the steps below, substitute the parameters enclosed in angle brackets with the corresponding values.
Example of a
subnet.yamlfile:apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: demo namespace: demo-namespace labels: kaas.mirantis.com/provider: baremetal spec: cidr: 10.11.0.0/24 gateway: 10.11.0.9 includeRanges: - 10.11.0.5-10.11.0.70 nameservers: - 172.18.176.6
Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.Specification fields of the Subnet object¶ Parameter
Description
cidr(singular)A valid IPv4 CIDR, for example, 10.11.0.0/24.
includeRanges(list)A comma-separated list of IP address ranges within the given CIDR that should be used in the allocation of IPs for nodes. The gateway, network, broadcast, and DNSaddresses will be excluded (protected) automatically if they intersect with one of the range. The IPs outside the given ranges will not be used in the allocation. Each element of the list can be either an interval
10.11.0.5-10.11.0.70or a single address10.11.0.77.Warning
Do not use values that are out of the given CIDR.
excludeRanges(list)A comma-separated list of IP address ranges within the given CIDR that should not be used in the allocation of IPs for nodes. The IPs within the given CIDR but outside the given ranges will be used in the allocation. The gateway, network, broadcast, and DNS addresses will be excluded (protected) automatically if they are included in the CIDR. Each element of the list can be either an interval
10.11.0.5-10.11.0.70or a single address10.11.0.77.Warning
Do not use values that are out of the given CIDR.
useWholeCidr(boolean)If set to
true, the subnet address (10.11.0.0 in the example above) and the broadcast address (10.11.0.255 in the example above) are included into the address allocation for nodes. Otherwise, (falseby default), the subnet address and broadcast address are excluded from the address allocation.gateway(singular)A valid gateway address, for example, 10.11.0.9.
nameservers(list)A list of the IP addresses of name servers. Each element of the list is a single address, for example, 172.18.176.6.
Configuration rules:
The subnet for the LCM network must contain the
ipam/SVC-k8s-lcm: "1"label. For details, see Service labels and their life cycle.Each cluster must use at least one subnet for its LCM network. Every node must have the address allocated in the LCM network using such subnet(s).
Each node of every cluster must have one and only IP address in the LCM network that is allocated from one of the
Subnetobjects having theipam/SVC-k8s-lcmlabel defined. Therefore, allSubnetobjects used for LCM networks must have theipam/SVC-k8s-lcmlabel defined.You can use any interface name for the LCM network traffic. The
Subnetobjects for the LCM network must have theipam/SVC-k8s-lcmlabel. For details, see Service labels and their life cycle.
Note
You may use different subnets to allocate IP addresses to different Container Cloud components in your cluster. Add a label with the
ipam/SVC-prefix to each subnet that is used to configure a Container Cloud service. For details, see Service labels and their life cycle and the optional steps below.Configure DHCP relay agents on the edges of the broadcast domains in the provisioning network, as needed.
Make sure to assign the IP address ranges you want to allocate to the hosts using DHCP for discovery and inspection. Create subnets using these IP parameters. Specify the IP address of your DHCP relay as the default gateway in the corresponding
Subnetobject.Caution
Support of multiple DHCP ranges has the following limitations:
Using of custom DNS server addresses for servers that boot over PXE is not supported.
The
Subnetobjects for DHCP ranges cannot be associated with any specific cluster, as the DHCP server configuration is only applicable to the management cluster where the DHCP server is running. Thecluster.sigs.k8s.io/cluster-namelabel will be ignored.
Configuration examples:
Single-rack cluster
apiVersion: "ipam.mirantis.com/v1alpha1" kind: Subnet metadata: name: mgmt-dhcp namespace: default labels: kaas.mirantis.com/provider: baremetal ipam/SVC-dhcp-range: "presents" spec: cidr: 10.20.10.0/24 includeRanges: - 10.20.10.10-10.20.10.20
Multi-rack cluster
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-1-dhcp namespace: default labels: ipam/SVC-dhcp-range: "1" kaas.mirantis.com/provider: baremetal spec: cidr: 10.20.101.0/24 gateway: 10.20.101.1 includeRanges: - 10.20.101.16-10.20.101.127 --- apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-2-dhcp namespace: default labels: ipam/SVC-dhcp-range: "1" kaas.mirantis.com/provider: baremetal spec: cidr: 10.20.102.0/24 gateway: 10.20.102.1 includeRanges: - 10.20.102.16-10.20.102.127 --- apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-3-dhcp namespace: default labels: ipam/SVC-dhcp-range: "1" kaas.mirantis.com/provider: baremetal spec: cidr: 10.20.103.0/24 gateway: 10.20.103.1 includeRanges: - 10.20.103.16-10.20.103.127 --- # Add more Subnet object templates as required using the above example # (one subnet per rack)
Optional. Add subnets for configuring multiple DHCP ranges. For details, see Configure multiple DHCP address ranges.
Add one or more subnets for the LCM network:
Set the
ipam/SVC-k8s-lcmlabel with the value"1"to create a subnet that will be used to assign IP addresses in the LCM network.Optional. Set the
cluster.sigs.k8s.io/cluster-namelabel to the name of the target cluster during the subnet creation.Use this subnet in the L2 template for cluster nodes.
Using the L2 template, assign this subnet to the interface connected to your LCM network.
Precautions for the LCM network usage
Each cluster must use at least one subnet for its LCM network. Every node must have the address allocated in the LCM network using such subnet(s).
Each node of every cluster must have one and only IP address in the LCM network that is allocated from one of the
Subnetobjects having theipam/SVC-k8s-lcmlabel defined. Therefore, allSubnetobjects used for LCM networks must have theipam/SVC-k8s-lcmlabel defined.You can use any interface name for the LCM network traffic. The
Subnetobjects for the LCM network must have theipam/SVC-k8s-lcmlabel. For details, see Service labels and their life cycle.
Configuration examples:
Single-rack cluster
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: labels: kaas.mirantis.com/provider: baremetal ipam/SVC-k8s-lcm: "1" name: lcm-nw namespace: <MOSKClusterNamespace> spec: cidr: 172.16.43.0/24 gateway: 172.16.43.1 includeRanges: - 172.16.43.10-172.16.43.100 nameservers: - 8.8.8.8
Multi-rack cluster
Example
mosk-racks-lcm-subnets.yamlNote
Subnet labels such as
rack-x-lcm,rack-api-lcm, and so on are optional. You can use them in L2 templates to selectSubnetobjects by label.apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-1-lcm namespace: mosk-namespace-name labels: ipam/SVC-k8s-lcm: "1" kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-1-lcm: "true" spec: cidr: 10.20.111.0/24 gateway: 10.20.111.1 includeRanges: - 10.20.111.16-10.20.111.255 nameservers: - 8.8.8.8 --- apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-2-lcm namespace: mosk-namespace-name labels: ipam/SVC-k8s-lcm: "1" kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-2-lcm: "true" spec: cidr: 10.20.112.0/24 gateway: 10.20.112.1 includeRanges: - 10.20.112.16-10.20.112.255 nameservers: - 8.8.8.8 --- apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-3-lcm namespace: mosk-namespace-name labels: ipam/SVC-k8s-lcm: "1" kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-3-lcm: "true" spec: cidr: 10.20.113.0/24 gateway: 10.20.113.1 includeRanges: - 10.20.113.16-10.20.113.255 nameservers: - 8.8.8.8 --- # Add more subnet object templates as required using the above example # (one subnet per rack)
Example
mosk-racks-api-lcm-subnet.yamlNote
Since 23.2.2, MOSK supports full L3 networking topology in the Technology Preview scope. This enables deployment of specific cluster segments in dedicated racks without the need for L2 layer extension between them. For configuration procedure, see Configure BGP announcement for cluster API LB address and Configure BGP announcement of external addresses of Kubernetes load-balanced services in Deployment Guide.
If BGP announcement is configured for the MOSK cluster API LB address, the API/LCM network is not required. Announcement of the cluster API LB address is done using the LCM network.
If you configure ARP announcement of the load-balancer IP address for the MOSK cluster API, the API/LCM network must be configured on the Kubernetes manager nodes of the cluster. This network contains the Kubernetes API endpoint with the VRRP virtual IP address.
This network contains Kubernetes API endpoint with the VRRP virtual IP address. This is the IP address space that Container Cloud uses to ensure communication between the LCM agents and the management API. These addresses are also used by Kubernetes nodes for communication. The addresses from the subnet are assigned to all Kubernetes manager nodes of the MOSK cluster.
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-api-lcm namespace: mosk-namespace-name labels: ipam/SVC-k8s-lcm: "1" kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-api-lcm: "true" spec: cidr: 10.20.110.0/24 gateway: 10.20.110.1 includeRanges: - 10.20.110.16-10.20.110.25 nameservers: - 8.8.8.8
Optional. Add a subnet for external connection to the Kubernetes services exposed by the MOSK cluster. The network is used to expose the OpenStack, StackLight, and other MOSK services. Configuration examples:
Single-rack cluster
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: labels: kaas.mirantis.com/provider: baremetal name: k8s-ext-subnet namespace: <MOSKClusterNamespace> spec: cidr: 172.16.45.0/24 gateway: 172.16.45.1 includeRanges: - 172.16.45.10-172.16.45.100 nameservers: - 8.8.8.8
Multi-rack cluster
Note
Since 23.2.2, MOSK supports full L3 networking topology in the Technology Preview scope. This enables deployment of specific cluster segments in dedicated racks without the need for L2 layer extension between them. For configuration procedure, see Configure BGP announcement for cluster API LB address and Configure BGP announcement of external addresses of Kubernetes load-balanced services in Deployment Guide.
If you configure BGP announcement for IP addresses of load-balanced services of a MOSK cluster, the external network can consist of multiple VLAN segments connected to all nodes of a MOSK cluster where MetalLB speaker components are configured to announce IP addresses for Kubernetes load-balanced services. Mirantis recommends that you use OpenStack controller nodes for this purpose.
If you configure ARP announcement for IP addresses of load-balanced services of a MOSK cluster, the external network must consist of a single VLAN stretched to the ToR switches of all the racks where MOSK nodes connected to the external network are located. Those are the nodes where MetalLB speaker components are configured to announce IP addresses for Kubernetes load-balanced services. Mirantis recommends that you use OpenStack controller nodes for this purpose.
The subnets are used to assign addresses to the external interfaces of the MOSK controller nodes and will be used to assign the default gateway to these hosts. The default gateway for other hosts of the MOSK cluster is assigned using the LCM and optionally API/LCM subnets.
Example of a subnet where a single VLAN segment is stretched to all MOSK controller nodes:
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: k8s-external namespace: mosk-namespace-name labels: kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name k8s-external: true spec: cidr: 10.20.120.0/24 gateway: 10.20.120.1 # This will be the default gateway on hosts includeRanges: - 10.20.120.16-10.20.120.20 nameservers: - 8.8.8.8
Example of subnets where separate VLAN segments per rack are used:
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-1-k8s-ext namespace: mosk-namespace-name labels: kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-1-k8s-ext: true spec: cidr: 10.20.121.0/24 gateway: 10.20.121.1 # This will be the default gateway on hosts includeRanges: - 10.20.121.16-10.20.121.20 nameservers: - 8.8.8.8 --- apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-2-k8s-ext namespace: mosk-namespace-name labels: kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-2-k8s-ext: true spec: cidr: 10.20.122.0/24 gateway: 10.20.122.1 # This will be the default gateway on hosts includeRanges: - 10.20.122.16-10.20.122.20 nameservers: - 8.8.8.8 --- apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-3-k8s-ext namespace: mosk-namespace-name labels: kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-3-k8s-ext: true spec: cidr: 10.20.123.0/24 gateway: 10.20.123.1 # This will be the default gateway on hosts includeRanges: - 10.20.123.16-10.20.123.20 nameservers: - 8.8.8.8
Configuration rules:
Make sure that
loadBalancerHostis set to""(empty string) in theClusterspec.spec: providerSpec: value: apiVersion: baremetal.k8s.io/v1alpha1 kind: BaremetalClusterProviderSpec ... loadBalancerHost: ""
Create a subnet with the
ipam/SVC-LBhostlabel having the"1"value to make thebaremetal-provideruse this subnet for allocation of addresses for cluster API endpoints. One IP address will be allocated for each cluster to serve its Kubernetes/MKE API endpoint.Make sure that master nodes have host local-link addresses in the same subnet as the cluster API endpoint address. These host IP addresses will be used for VRRP traffic. The cluster API endpoint address will be assigned to the same interface on one of the master nodes where these host IP addresses are assigned.
Mirantis highly recommends that you assign the cluster API endpoint address from the LCM or external network. For details on cluster network types, refer to MOSK cluster networking.
To add an address allocation scope of API endpoints, create a subnet in the corresponding namespace with a reference to the target cluster using the
cluster.sigs.k8s.io/cluster-namelabel. For example:apiVersion: "ipam.mirantis.com/v1alpha1" kind: Subnet metadata: name: lbhost-mgmt-cluster namespace: default labels: kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mgmt-cluster ipam/SVC-LBhost: "presents" spec: cidr: "10.0.30.100/32" useWholeCidr: true
Optional. Add a subnet(s) for the storage access network. Ceph will automatically use this subnet for its external connections. A Ceph OSD will look for and bind to an address from this subnet when it is started on a machine. Configuration examples:
Single-rack cluster
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: labels: kaas.mirantis.com/provider: baremetal ipam/SVC-ceph-public: true cluster.sigs.k8s.io/cluster-name: <MOSKClusterName> name: ceph-public-subnet namespace: <MOSKClusterNamespace> spec: cidr: 10.12.0.0/24
Multi-rack cluster
This network may have per-rack VLANs and IP subnets. The addresses from the subnets are assigned to all MOSK cluster nodes besides Kubernetes manager nodes.
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-1-ceph-public namespace: mosk-namespace-name labels: ipam/SVC-ceph-public: "1" kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-1-ceph-public: true spec: cidr: 10.20.131.0/24 gateway: 10.20.131.1 includeRanges: - 10.20.131.16-10.20.131.255 --- apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-2-ceph-public namespace: mosk-namespace-name labels: ipam/SVC-ceph-public: "1" kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-2-ceph-public: true spec: cidr: 10.20.132.0/24 gateway: 10.20.132.1 includeRanges: - 10.20.132.16-10.20.132.255 --- apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-3-ceph-public namespace: mosk-namespace-name labels: ipam/SVC-ceph-public: "1" kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-3-ceph-public: true spec: cidr: 10.20.133.0/24 gateway: 10.20.133.1 includeRanges: - 10.20.133.16-10.20.133.255 --- # Add more Subnet object templates as required using the above example # (one subnet per rack)
Configuration rules:
Set the
ipam/SVC-ceph-publiclabel with the value"1"to create a subnet that will be used to configure the Ceph public network.Set the
cluster.sigs.k8s.io/cluster-namelabel to the name of the target cluster during the subnet creation.Use this subnet in the L2 template for all cluster nodes except Kubernetes manager nodes.
Assign this subnet to the interface connected to your storage access network.
Optional. Add a subnet(s) for the storage replication network. Ceph will automatically use this network for its internal replication traffic. Configuration examples:
Single-rack cluster
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: labels: kaas.mirantis.com/provider: baremetal ipam/SVC-ceph-cluster: true cluster.sigs.k8s.io/cluster-name: <MOSKClusterName> name: ceph-cluster-subnet namespace: <MOSKClusterNamespace> spec: cidr: 10.12.1.0/24
Multi-rack cluster
This network may have per-rack VLANs and IP subnets. The addresses from the subnets are assigned to storage nodes in the MOSK cluster.
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-1-ceph-cluster namespace: mosk-namespace-name labels: ipam/SVC-ceph-cluster: "1" kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-1-ceph-cluster: true spec: cidr: 10.20.141.0/24 gateway: 10.20.141.1 includeRanges: - 10.20.141.16-10.20.141.255 --- apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-2-ceph-cluster namespace: mosk-namespace-name labels: ipam/SVC-ceph-cluster: "1" kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-2-ceph-cluster: true spec: cidr: 10.20.142.0/24 gateway: 10.20.142.1 includeRanges: - 10.20.142.16-10.20.142.255 --- apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-3-ceph-cluster namespace: mosk-namespace-name labels: ipam/SVC-ceph-cluster: "1" kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-3-ceph-cluster: true spec: cidr: 10.20.143.0/24 gateway: 10.20.143.1 includeRanges: - 10.20.143.16-10.20.143.255 --- # Add more Subnet object templates as required using the above example # (one subnet per rack)
Configuration rules:
Set the
ipam/SVC-ceph-clusterlabel with the value"1"to create a subnet that will be used to configure the Ceph cluster network.Set the
cluster.sigs.k8s.io/cluster-namelabel to the name of the target cluster during the subnet creation.Use this subnet in the L2 template for storage nodes.
Assign this subnet to the interface connected to your storage replication network.
Optional. Add a subnet for the Kubernetes Pods traffic. The addresses from this subnet are assigned to interfaces connected to the Kubernetes workloads network and used by Calico CNI as underlay for traffic between the pods in the Kubernetes cluster. Configuration examples:
Single-rack cluster
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: labels: kaas.mirantis.com/provider: baremetal name: k8s-pods-subnet namespace: <MOSKClusterNamespace> spec: cidr: 10.12.3.0/24 includeRanges: - 10.12.3.10-10.12.3.100
Multi-rack cluster
This network may include multiple per-rack VLANs and IP subnets. The addresses from the subnets are assigned to all MOSK cluster nodes. For details, see Network types.
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-1-k8s-pods namespace: mosk-namespace-name labels: kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-1-k8s-pods: true spec: cidr: 10.20.151.0/24 gateway: 10.20.151.1 includeRanges: - 10.20.151.16-10.20.151.255 --- apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-2-k8s-pods namespace: mosk-namespace-name labels: kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-2-k8s-pods: true spec: cidr: 10.20.152.0/24 gateway: 10.20.152.1 includeRanges: - 10.20.152.16-10.20.152.255 --- apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-3-k8s-pods namespace: mosk-namespace-name labels: kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-3-k8s-pods: true spec: cidr: 10.20.153.0/24 gateway: 10.20.153.1 includeRanges: - 10.20.153.16-10.20.153.255 --- # Add more Subnet object templates as required using the above example # (one subnet per rack)
Configuration rules:
Use this subnet in the L2 template for all nodes in the cluster.
Use the
npTemplate.bridges.k8s-podsbridge name in the L2 template. This bridge name is reserved for the Kubernetes workloads network. When thek8s-podsbridge is defined in an L2 template, Calico CNI uses that network for routing the Pods traffic between nodes.
Optional. Add a subnet for the MOSK overlay network. this is the underlay network for VXLAN tunnels for the MOSK tenant traffic. If deployed with Tungsten Fabric, it is used for the MPLS over UDP+GRE traffic. Configuration examples:
Single-rack cluster
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: labels: kaas.mirantis.com/provider: baremetal name: neutron-tunnel-subnet namespace: <MOSKClusterNamespace> spec: cidr: 10.12.2.0/24 includeRanges: - 10.12.2.10-10.12.2.100
Multi-rack cluster
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-1-tenant-tunnel namespace: mosk-namespace-name labels: kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-1-tenant-tunnel: true spec: cidr: 10.20.161.0/24 gateway: 10.20.161.1 includeRanges: - 10.20.161.16-10.20.161.255 --- apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-2-tenant-tunnel namespace: mosk-namespace-name labels: kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-2-tenant-tunnel: true spec: cidr: 10.20.162.0/24 gateway: 10.20.162.1 includeRanges: - 10.20.162.16-10.20.162.255 --- apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-3-tenant-tunnel namespace: mosk-namespace-name labels: kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-3-tenant-tunnel: true spec: cidr: 10.20.163.0/24 gateway: 10.20.163.1 includeRanges: - 10.20.163.16-10.20.163.255 --- # Add more Subnet object templates as required using the above example # (one subnet per rack)
Configuration rules:
Use this subnet in the L2 template for the compute and gateway (controller) nodes in the MOSK cluster.
Assign this subnet to the interface connected to your MOSK overlay network.
This network is used to provide denied and secure tenant networks with the help of the tunneling mechanism (VLAN/GRE/VXLAN). If the VXLAN and GRE encapsulation takes place, the IP address assignment is required on interfaces at the node level. On the Tungsten Fabric deployments, this network is used for MPLS over UDP+GRE traffic.
Optional. Add a subnet for the MOSK live migration network. This subnet is used by the Compute service (OpenStack Nova) to transfer data during live migration. Depending on the cloud needs, you can place it on a dedicated physical network not to affect other networks during live migration. The IP address assignment is required on interfaces at the node level. Configuration examples:
Single-rack cluster
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: labels: kaas.mirantis.com/provider: baremetal name: live-migration-subnet namespace: <MOSKClusterNamespace> spec: cidr: 10.12.7.0/24 includeRanges: - 10.12.7.10-10.12.7.100
Multi-rack cluster
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-1-live-migration namespace: mosk-namespace-name labels: kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-1-live-migration: true spec: cidr: 10.20.171.0/24 gateway: 10.20.171.1 includeRanges: - 10.20.171.16-10.20.171.255 --- apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-2-live-migration namespace: mosk-namespace-name labels: kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-2-live-migration: true spec: cidr: 10.20.172.0/24 gateway: 10.20.172.1 includeRanges: - 10.20.172.16-10.20.172.255 --- apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: name: rack-3-live-migration namespace: mosk-namespace-name labels: kaas.mirantis.com/provider: baremetal cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack-3-live-migration: true spec: cidr: 10.20.173.0/24 gateway: 10.20.173.1 includeRanges: - 10.20.173.16-10.20.173.255 --- # Add more Subnet object templates as required using the above example # (one subnet per rack)
Configuration rules:
Use this subnet in the L2 template for compute nodes in the MOSK cluster.
Assign this subnet to the interface connected to your MOSK overlay network.
Verify that the subnet is successfully created:
kubectl get subnet kaas-mgmt -oyaml
In the system output, verify the
Subnetobject status.Status fields of the Subnet object¶ Parameter
Description
stateSince 23.1Contains a short state description and a more detailed one if applicable. The short status values are as follows:
OK- object is operational.ERR- object is non-operational. This status has a detailed description in themessageslist.TERM- object was deleted and is terminating.
messagesSince 23.1Contains error or warning messages if the object state is
ERR. For example,ERR: Wrong includeRange for CIDR….statusMessageDeprecated since MOSK 23.1 and will be removed in one of the following releases in favor of
stateandmessages. Since MOSK 23.2, this field is not set for the objects of newly created clusters.cidrReflects the actual CIDR, has the same meaning as
spec.cidr.gatewayReflects the actual gateway, has the same meaning as
spec.gateway.nameserversReflects the actual name servers, has same meaning as
spec.nameservers.rangesSpecifies the address ranges that are calculated using the fields from
spec: cidr, includeRanges, excludeRanges, gateway, useWholeCidr. These ranges are directly used for nodes IP allocation.allocatableIncludes the number of currently available IP addresses that can be allocated for nodes from the subnet.
allocatedIPsSpecifies the list of IPv4 addresses with the corresponding
IPaddrobject IDs that were already allocated from the subnet.capacityContains the total number of IP addresses being held by ranges that equals to a sum of the
allocatableandallocatedIPsparameters values.objCreatedDate, time, and IPAM version of the
SubnetCR creation.objStatusUpdatedDate, time, and IPAM version of the last update of the
statusfield in theSubnetCR.objUpdatedDate, time, and IPAM version of the last
SubnetCR update bykaas-ipam.Example of a successfully created subnet:
apiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: labels: ipam/UID: 6039758f-23ee-40ba-8c0f-61c01b0ac863 kaas.mirantis.com/provider: baremetal ipam/SVC-k8s-lcm: "1" name: kaas-mgmt namespace: default spec: cidr: 10.0.0.0/24 excludeRanges: - 10.0.0.100 - 10.0.0.101-10.0.0.120 gateway: 10.0.0.1 includeRanges: - 10.0.0.50-10.0.0.90 nameservers: - 172.18.176.6 status: allocatable: 38 allocatedIPs: - 10.0.0.50:0b50774f-ffed-11ea-84c7-0242c0a85b02 - 10.0.0.51:1422e651-ffed-11ea-84c7-0242c0a85b02 - 10.0.0.52:1d19912c-ffed-11ea-84c7-0242c0a85b02 capacity: 41 cidr: 10.0.0.0/24 gateway: 10.0.0.1 objCreated: 2021-10-21T19:09:32Z by v5.1.0-20210930-121522-f5b2af8 objStatusUpdated: 2021-10-21T19:14:18.748114886Z by v5.1.0-20210930-121522-f5b2af8 objUpdated: 2021-10-21T19:09:32.606968024Z by v5.1.0-20210930-121522-f5b2af8 nameservers: - 172.18.176.6 ranges: - 10.0.0.50-10.0.0.90
Proceed to creating L2 templates as described in Create L2 templates.
Create L2 templates¶
After you create subnets for the MOSK cluster as described in Create subnets, follow the procedure below to create L2 templates for different types of OpenStack nodes in the cluster.
See the following subsections for templates that implement the MOSK Reference Architecture: Networking. You may adjust the templates according to the requirements of your architecture using the last two subsections of this section. They explain mandatory parameters of the templates and supported configuration options.
After you create subnets for one or more MOSK clusters or projects as described in Create subnets, follow the procedure below to create L2 templates for a MOSK cluster.
L2 templates are used directly during provisioning. This way, a hardware node obtains and applies a complete network configuration during the first system boot.
Caution
Update any L2 template created before Container Cloud 2.9.0 (Cluster releases 6.14.0, 5.15.0, or earlier) to the new format as described in Container Cloud Release Notes: Switch L2 templates to the new format.
Caution
Create L2 templates before adding any machines to your new MOSK cluster.
Log in to a local machine where your management cluster
kubeconfigis located and wherekubectlis installed.Note
The management cluster
kubeconfigis created during the last stage of the management cluster bootstrap.Create a set of
L2TemplateYAML files specific to your deployment using exemplary templates provided in Create L2 templates.Note
You can create several L2 templates with different configurations to be applied to different nodes of the same cluster. See Assign L2 templates to machines for details.
Add or edit the mandatory labels and parameters in the new L2 template. For description of mandatory labels and parameters, see L2Template.
Optional. To designate an L2 template as default, assign the
ipam/DefaultForClusterlabel to it. Only one L2 template in a cluster can have this label. It will be used for machines that do not have an L2 template explicitly assigned to them.Note
You may skip this step and add the default label along with other custom labels using the Container Cloud web UI, as described below in this procedure.
To assign the default template to the cluster:
Use the mandatory
cluster.sigs.k8s.io/cluster-namelabel in the L2 templatemetadatasection.Use the
cluster.sigs.k8s.io/cluster-namelabel or theclusterRefparameter in the L2 templatespecsection. During cluster update to 2.25.0, this deprecated parameter is automatically migrated to thecluster.sigs.k8s.io/cluster-namelabel.Optional. Add custom labels to the L2 template. You can refer to these labels to assign the L2 template to machines.
Add the L2 template to your management cluster. Select one of the following options:
kubectl --kubeconfig <pathToManagementClusterKubeconfig> apply -f <pathToL2TemplateYamlFile>
Log in to the Container Cloud web UI with the
m:kaas:namespace@operatororm:kaas:namespace@writerpermissions.Switch to the required non-
defaultproject using the Switch Project action icon located on top of the main left-side navigation panel.Caution
Do not create a MOSK cluster in the
defaultproject (Kubernetes namespace), which is dedicated for the management cluster only. If no projects are defined, first create a newmoskproject as described in Create a project for MOSK clusters.In the left sidebar, navigate to Networks and click the L2 Templates tab.
Click Create L2 Template.
Fill out the Create L2 Template form as required:
- Name
L2 template name.
- Cluster
Cluster name that the L2 template is being added for. To set the L2 template as default for all machines, also select Set default for the cluster.
- Specification
L2 specification in the YAML format that you have previously created. Click Edit to edit the L2 template if required.
Note
Before Container Cloud 2.28.0 (Cluster releases 17.3.0 and 16.3.0), the field name is YAML file, and you can upload the required YAML file instead of inserting and editing it.
- Labels
Available since Container Cloud 2.28.0 (Cluster releases 17.3.0 and 16.3.0). Key-value pairs attached to the L2 template. For details, see L2Template metadata.
Optional. Further modify the template, if required. For description of parameters, see L2Template.
kubectl --kubeconfig <pathToManagementClusterKubeconfig> \ -n <ProjectNameForNewMOSKCluster> edit l2template <L2templateName>
Caution
Modification of L2 templates in use is only allowed with a mandatory validation step from the infrastructure operator to prevent accidental cluster failures due to unsafe changes. The list of risks posed by modifying L2 templates includes:
Services running on hosts cannot reconfigure automatically to switch to the new IP addresses and/or interfaces.
Connections between services are interrupted unexpectedly, which can cause data loss.
Incorrect configurations on hosts can lead to irrevocable loss of connectivity between services and unexpected cluster partition or disassembly.
For details, see Modify network configuration on an existing machine.
Proceed with Add a machine. The resulting L2 template will be used to render the netplan configuration for the MOSK cluster machines.
The
kaas-ipamservice uses the data fromBareMetalHost,L2Template, andSubnetobjects to generate the netplan configuration for every cluster machine.Note
Before update of the management cluster to Container Cloud 2.29.0 (Cluster release 16.4.0), instead of
BareMetalHostInventory, use theBareMetalHostobject. For details, see BareMetalHost resource.Caution
While the Cluster release of the management cluster is 16.4.0,
BareMetalHostInventoryoperations are allowed tom:kaas@management-adminonly. This limitation is lifted once the management cluster is updated to the Cluster release 16.4.1 or later.The generated netplan configuration is saved in the
status.netconfigFilessection of theIpamHostobject. If thestatus.netconfigFilesStatefield of theIpamHostobject isOK, the configuration was rendered in theIpamHostobject successfully. Otherwise, the status contains an error message.Caution
The following fields of the
ipamHoststatus are renamed since MOSK 23.1 in the scope of theL2TemplateandIpamHostobjects refactoring:netconfigV2tonetconfigCandidatenetconfigV2statetonetconfigCandidateStatenetconfigFilesStatetonetconfigFilesStates(per file)
No user actions are required after renaming.
The format of
netconfigFilesStatechanged after renaming. ThenetconfigFilesStatesfield contains a dictionary of statuses of network configuration files stored innetconfigFiles. The dictionary contains the keys that are file paths and values that have the same meaning for each file thatnetconfigFilesStatehad:For a successfully rendered configuration file:
OK: <timestamp> <sha256-hash-of-rendered-file>, where a timestamp is in the RFC 3339 format.For a failed rendering:
ERR: <error-message>.
The
baremetal-providerservice copies data fromstatus.netconfigFilesof theIpamHostobject to theSpec.StateItemsOverwrites[‘deploy’][‘bm_ipam_netconfigv2’]parameter ofLCMMachine.The
lcm-agentservice on every host synchronizes theLCMMachinedata to its host. Thelcm-agentservice runs a playbook to update the netplan configuration on the host during thepre-downloadanddeployphases.
For a multi-rack MOSK cluster, you need to create one L2 template for each type of server in each rack. This may result in a large number of L2 templates in your configuration.
For example, if you have a three-rack deployment of MOSK with 4 types of nodes evenly distributed across three racks, you have to create at least the following L2 templates:
rack-1-k8s-manager,rack-2-k8s-manager,rack-3-k8s-managerfor Kubernetes control plane nodes, unless you use the compact control plane option.rack-1-mosk-control,rack-2-mosk-control,rack-3-mosk-controlfor OpenStack controller nodes in each rack.rack-1-mosk-compute,rack-2-mosk-compute,rack-3-mosk-computefor OpenStack compute nodes in each rack.rack-1-mosk-storage,rack-2-mosk-storage,rack-3-mosk-storagefor OpenStack storage nodes in each rack.
In total, twelve L2 templates are required for this relatively simple cluster. In the following sections, the examples cover only one rack, but can be easily expanded to more racks.
Note
Three servers are required for Kubernetes control plane and for the OpenStack control plane. So, you might not need more L2 templates for these roles when expanding beyond three racks.
Now, proceed to creating L2 templates for your cluster, starting from Create an L2 template for a Kubernetes manager node.
Caution
Modification of L2 templates in use is only allowed with a mandatory validation step from the infrastructure operator to prevent accidental cluster failures due to unsafe changes. The list of risks posed by modifying L2 templates includes:
Services running on hosts cannot reconfigure automatically to switch to the new IP addresses and/or interfaces.
Connections between services are interrupted unexpectedly, which can cause data loss.
Incorrect configurations on hosts can lead to irrevocable loss of connectivity between services and unexpected cluster partition or disassembly.
For details, see Modify network configuration on an existing machine.
According to the reference architecture, the Kubernetes manager nodes in the MOSK cluster must be connected to the following networks:
PXE network
API/LCM network (if you configure ARP announcement of the load-balancer IP address for the MOSK cluster API)
LCM network (if you configure BGP announcement of the load-balancer IP address for the MOSK cluster API)
Kubernetes workloads network
Caution
If you plan to deploy MOSK cluster with the compact control plane option, skip this section entirely and proceed with Create an L2 template for a MOSK controller node.
To create L2 templates for Kubernetes manager nodes:
Create or open the
mosk-l2templates.ymlfile that contains the L2 templates you are preparing.Add L2 templates using the following example. Adjust the values of specific parameters according to the specifications of your environment, specifically the name of your project (namespace) and cluster, IP address ranges and networks, subnet names.
L2 template example for Kubernetes manager node¶apiVersion: ipam.mirantis.com/v1alpha1 kind: L2Template metadata: labels: kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack1-mosk-manager: "true" name: rack1-mosk-manager namespace: mosk-namespace-name spec: autoIfMappingPrio: - provision - eno - ens - enp l3Layout: - subnetName: api-lcm scope: namespace - subnetName: rack1-k8s-pods scope: namespace npTemplate: |- version: 2 ethernets: {{nic 0}}: dhcp4: false dhcp6: false match: macaddress: {{mac 0}} set-name: {{nic 0}} mtu: 9000 {{nic 1}}: dhcp4: false dhcp6: false match: macaddress: {{mac 1}} set-name: {{nic 1}} mtu: 9000 {{nic 2}} dhcp4: false dhcp6: false match: macaddress: {{mac 2}} set-name: {{nic 2}} mtu: 9000 {{nic 3}}: dhcp4: false dhcp6: false match: macaddress: {{mac 3}} set-name: {{nic 3}} mtu: 9000 bonds: bond0: mtu: 9000 parameters: mode: 802.3ad mii-monitor-interval: 100 interfaces: - {{nic 0}} - {{nic 1}} vlans: k8s-lcm-v: id: 403 link: bond0 mtu: 9000 k8s-pods-v: id: 408 link: bond0 mtu: 9000 bridges: k8s-lcm: interfaces: [k8s-lcm-v] addresses: - {{ ip "k8s-lcm:api-lcm" }} nameservers: addresses: {{nameservers_from_subnet "api-lcm"}} gateway4: {{ gateway_from_subnet "api-lcm" }} k8s-pods: interfaces: [k8s-pods-v] addresses: - {{ip "k8s-pods:rack1-k8s-pods"}} mtu: 9000 routes: - to: 10.199.0.0/22 # aggregated address space for Kubernetes workloads via: {{gateway_from_subnet "rack1-k8s-pods"}}
Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.Note
Before MOSK 23.3, an L2 template requires
clusterRef: <clusterName>in thespecsection. Since MOSK 23.3, this parameter is deprecated and automatically migrated to thecluster.sigs.k8s.io/cluster-name: <clusterName>label.To create L2 templates for other racks, change the rack identifier in the names and labels above.
Proceed with Create an L2 template for a MOSK controller node. The resulting L2 templates will be used to render the netplan configuration for the managed cluster machines.
Caution
Modification of L2 templates in use is only allowed with a mandatory validation step from the infrastructure operator to prevent accidental cluster failures due to unsafe changes. The list of risks posed by modifying L2 templates includes:
Services running on hosts cannot reconfigure automatically to switch to the new IP addresses and/or interfaces.
Connections between services are interrupted unexpectedly, which can cause data loss.
Incorrect configurations on hosts can lead to irrevocable loss of connectivity between services and unexpected cluster partition or disassembly.
For details, see Modify network configuration on an existing machine.
According to the reference architecture, MOSK controller nodes must be connected to the following networks:
PXE network
LCM network
Kubernetes workloads network
Storage access network (if deploying with Ceph as a backend for ephemeral storage)
Floating IP and provider networks. Not required for deployment with Tungsten Fabric.
Tenant underlay networks. If deploying with VXLAN networking or with Tungsten Fabric. In the latter case, the BGP service is configured over this network.
To create L2 templates for MOSK controller nodes:
Create or open the
mosk-l2template.ymlfile that contains the L2 templates.Add L2 templates using the following example. Adjust the values of specific parameters according to the specification of your environment.
Example of an L2 template for a MOSK controller node¶apiVersion: ipam.mirantis.com/v1alpha1 kind: L2Template metadata: labels: kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack1-mosk-controller: "true" name: rack1-mosk-controller namespace: mosk-namespace-name spec: autoIfMappingPrio: - provision - eno - ens - enp l3Layout: - subnetName: mgmt-lcm scope: global - subnetName: rack1-k8s-lcm scope: namespace - subnetName: k8s-external scope: namespace - subnetName: rack1-k8s-pods scope: namespace - subnetName: rack1-ceph-public scope: namespace - subnetName: rack1-tenant-tunnel scope: namespace npTemplate: |- version: 2 ethernets: {{nic 0}}: dhcp4: false dhcp6: false match: macaddress: {{mac 0}} set-name: {{nic 0}} mtu: 9000 {{nic 1}}: dhcp4: false dhcp6: false match: macaddress: {{mac 1}} set-name: {{nic 1}} mtu: 9000 {{nic 2}} dhcp4: false dhcp6: false match: macaddress: {{mac 2}} set-name: {{nic 2}} mtu: 9000 {{nic 3}}: dhcp4: false dhcp6: false match: macaddress: {{mac 3}} set-name: {{nic 3}} mtu: 9000 bonds: bond0: mtu: 9000 parameters: mode: 802.3ad mii-monitor-interval: 100 interfaces: - {{nic 0}} - {{nic 1}} bond1: mtu: 9000 parameters: mode: 802.3ad mii-monitor-interval: 100 interfaces: - {{nic 2}} - {{nic 3}} vlans: k8s-lcm-v: id: 403 link: bond0 mtu: 9000 k8s-ext-v: id: 409 link: bond0 mtu: 9000 k8s-pods-v: id: 408 link: bond0 mtu: 9000 pr-floating: id: 407 link: bond1 mtu: 9000 stor-frontend: id: 404 link: bond0 addresses: - {{ip "stor-frontend:rack1-ceph-public"}} mtu: 9000 routes: - to: 10.199.16.0/22 # aggregated address space for Ceph public network via: {{ gateway_from_subnet "rack1-ceph-public" }} tenant-tunnel: id: 406 link: bond1 addresses: - {{ip "tenant-tunnel:rack1-tenant-tunnel"}} mtu: 9000 routes: - to: 10.195.0.0/22 # aggregated address space for tenant networks via: {{ gateway_from_subnet "rack1-tenant-tunnel" }} bridges: k8s-lcm: interfaces: [k8s-lcm-v] addresses: - {{ ip "k8s-lcm:rack1-k8s-lcm" }} nameservers: addresses: {{nameservers_from_subnet "rack1-k8s-lcm"}} routes: - to: 10.197.0.0/21 # aggregated address space for LCM and API/LCM networks via: {{ gateway_from_subnet "rack1-k8s-lcm" }} - to: {{ cidr_from_subnet "mgmt-lcm" }} via: {{ gateway_from_subnet "rack1-k8s-lcm" }} k8s-ext: interfaces: [k8s-ext-v] addresses: - {{ip "k8s-ext:k8s-external"}} nameservers: addresses: {{nameservers_from_subnet "k8s-external"}} gateway4: {{ gateway_from_subnet "k8s-external" }} mtu: 9000 k8s-pods: interfaces: [k8s-pods-v] addresses: - {{ip "k8s-pods:rack1-k8s-pods"}} mtu: 9000 routes: - to: 10.199.0.0/22 # aggregated address space for Kubernetes workloads via: {{ gateway_from_subnet "rack1-k8s-pods" }}
Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.Note
Before MOSK 23.3, an L2 template requires
clusterRef: <clusterName>in thespecsection. Since MOSK 23.3, this parameter is deprecated and automatically migrated to thecluster.sigs.k8s.io/cluster-name: <clusterName>label.
Caution
If you plan to deploy a MOSK cluster with
the compact control plane option and configure ARP announcement of the
load-balancer IP address for the MOSK cluster API, the
API/LCM network will be used for MOSK controller nodes.
Therefore, change the rack1-k8s-lcm subnet to the api-lcm one in
the corresponding L2Template object:
spec:
...
l3Layout:
...
- subnetName: api-lcm
scope: namespace
...
npTemplate: |-
...
bridges:
k8s-lcm:
interfaces: [k8s-lcm-v]
addresses:
- {{ ip "k8s-lcm:api-lcm" }}
nameservers:
addresses: {{nameservers_from_subnet "api-lcm"}}
routes:
- to: 10.197.0.0/21 # aggregated address space for LCM and API/LCM networks
via: {{ gateway_from_subnet "api-lcm" }}
- to: {{ cidr_from_subnet "mgmt-lcm" }}
via: {{ gateway_from_subnet "api-lcm" }}
...
Proceed with Create an L2 template for a MOSK compute node.
Caution
Modification of L2 templates in use is only allowed with a mandatory validation step from the infrastructure operator to prevent accidental cluster failures due to unsafe changes. The list of risks posed by modifying L2 templates includes:
Services running on hosts cannot reconfigure automatically to switch to the new IP addresses and/or interfaces.
Connections between services are interrupted unexpectedly, which can cause data loss.
Incorrect configurations on hosts can lead to irrevocable loss of connectivity between services and unexpected cluster partition or disassembly.
For details, see Modify network configuration on an existing machine.
According to the reference architecture, MOSK compute nodes must be connected to the following networks:
PXE network
LCM network
Kubernetes workloads network
Storage access network (if deploying with Ceph as a backend for ephemeral storage)
Floating IP and provider networks (if deploying OpenStack with DVR)
Tenant underlay networks
To create L2 templates for MOSK compute nodes:
Add L2 templates to the
mosk-l2templates.ymlfile using the following example. Adjust the values of parameters according to the specification of your environment.Example of an L2 template for a MOSK compute node¶apiVersion: ipam.mirantis.com/v1alpha1 kind: L2Template metadata: labels: kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack1-mosk-compute: "true" name: rack1-mosk-compute namespace: mosk-namespace-name spec: autoIfMappingPrio: - provision - eno - ens - enp l3Layout: - subnetName: rack1-k8s-lcm scope: namespace - subnetName: rack1-k8s-pods scope: namespace - subnetName: rack1-ceph-public scope: namespace - subnetName: rack1-tenant-tunnel scope: namespace npTemplate: |- version: 2 ethernets: {{nic 0}}: dhcp4: false dhcp6: false match: macaddress: {{mac 0}} set-name: {{nic 0}} mtu: 9000 {{nic 1}}: dhcp4: false dhcp6: false match: macaddress: {{mac 1}} set-name: {{nic 1}} mtu: 9000 {{nic 2}} dhcp4: false dhcp6: false match: macaddress: {{mac 2}} set-name: {{nic 2}} mtu: 9000 {{nic 3}}: dhcp4: false dhcp6: false match: macaddress: {{mac 3}} set-name: {{nic 3}} mtu: 9000 bonds: bond0: mtu: 9000 parameters: mode: 802.3ad mii-monitor-interval: 100 interfaces: - {{nic 0}} - {{nic 1}} bond1: mtu: 9000 parameters: mode: 802.3ad mii-monitor-interval: 100 interfaces: - {{nic 2}} - {{nic 3}} vlans: k8s-lcm-v: id: 403 link: bond0 mtu: 9000 k8s-pods-v: id: 408 link: bond0 mtu: 9000 pr-floating: id: 407 link: bond1 mtu: 9000 stor-frontend: id: 404 link: bond0 addresses: - {{ip "stor-frontend:rack1-ceph-public"}} mtu: 9000 routes: - to: 10.199.16.0/22 # aggregated address space for Ceph public network via: {{ gateway_from_subnet "rack1-ceph-public" }} tenant-tunnel: id: 406 link: bond1 addresses: - {{ip "tenant-tunnel:rack1-tenant-tunnel"}} mtu: 9000 routes: - to: 10.195.0.0/22 # aggregated address space for tenant networks via: {{ gateway_from_subnet "rack1-tenant-tunnel" }} bridges: k8s-lcm: interfaces: [k8s-lcm-v] addresses: - {{ ip "k8s-lcm:rack1-k8s-lcm" }} nameservers: addresses: {{nameservers_from_subnet "rack1-k8s-lcm"}} gateway4: {{ gateway_from_subnet "rack1-k8s-lcm" }} k8s-pods: interfaces: [k8s-pods-v] addresses: - {{ip "k8s-pods:k8s-pods-subnet"}} mtu: 9000 routes: - to: 10.199.0.0/22 # aggregated address space for Kubernetes workloads via: {{gateway_from_subnet "rack1-k8s-pods"}}
Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.Note
Before MOSK 23.3, an L2 template requires
clusterRef: <clusterName>in thespecsection. Since MOSK 23.3, this parameter is deprecated and automatically migrated to thecluster.sigs.k8s.io/cluster-name: <clusterName>label.Proceed with Create an L2 template for a MOSK storage node.
Caution
Modification of L2 templates in use is only allowed with a mandatory validation step from the infrastructure operator to prevent accidental cluster failures due to unsafe changes. The list of risks posed by modifying L2 templates includes:
Services running on hosts cannot reconfigure automatically to switch to the new IP addresses and/or interfaces.
Connections between services are interrupted unexpectedly, which can cause data loss.
Incorrect configurations on hosts can lead to irrevocable loss of connectivity between services and unexpected cluster partition or disassembly.
For details, see Modify network configuration on an existing machine.
According to the reference architecture, MOSK storage nodes in the MOSK cluster must be connected to the following networks:
PXE network
LCM network
Kubernetes workloads network
Storage access network
Storage replication network
To create L2 templates for MOSK storage nodes:
Add L2 templates to the
mosk-l2templates.ymlfile using the following example. Adjust the values of parameters according to the specification of your environment.Example of an L2 template for a MOSK storage node¶apiVersion: ipam.mirantis.com/v1alpha1 kind: L2Template metadata: labels: kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one cluster.sigs.k8s.io/cluster-name: mosk-cluster-name rack1-mosk-storage: "true" name: rack1-mosk-storage namespace: mosk-namespace-name spec: autoIfMappingPrio: - provision - eno - ens - enp l3Layout: - subnetName: rack1-k8s-lcm scope: namespace - subnetName: rack1-k8s-pods scope: namespace - subnetName: rack1-ceph-public scope: namespace - subnetName: rack1-ceph-cluster scope: namespace npTemplate: |- version: 2 ethernets: {{nic 0}}: dhcp4: false dhcp6: false match: macaddress: {{mac 0}} set-name: {{nic 0}} mtu: 9000 {{nic 1}}: dhcp4: false dhcp6: false match: macaddress: {{mac 1}} set-name: {{nic 1}} mtu: 9000 {{nic 2}} dhcp4: false dhcp6: false match: macaddress: {{mac 2}} set-name: {{nic 2}} mtu: 9000 {{nic 3}}: dhcp4: false dhcp6: false match: macaddress: {{mac 3}} set-name: {{nic 3}} mtu: 9000 bonds: bond0: mtu: 9000 parameters: mode: 802.3ad mii-monitor-interval: 100 interfaces: - {{nic 0}} - {{nic 1}} bond1: mtu: 9000 parameters: mode: 802.3ad mii-monitor-interval: 100 interfaces: - {{nic 2}} - {{nic 3}} vlans: k8s-lcm-v: id: 403 link: bond0 mtu: 9000 k8s-pods-v: id: 408 link: bond0 mtu: 9000 stor-frontend: id: 404 link: bond0 addresses: - {{ip "stor-frontend:rack1-ceph-public"}} mtu: 9000 routes: - to: 10.199.16.0/22 # aggregated address space for Ceph public network via: {{ gateway_from_subnet "rack1-ceph-public" }} stor-backend: id: 405 link: bond1 addresses: - {{ip "stor-backend:rack1-ceph-cluster"}} mtu: 9000 routes: - to: 10.199.32.0/22 # aggregated address space for Ceph cluster network via: {{ gateway_from_subnet "rack1-ceph-cluster" }} bridges: k8s-lcm: interfaces: [k8s-lcm-v] addresses: - {{ ip "k8s-lcm:rack1-k8s-lcm" }} nameservers: addresses: {{nameservers_from_subnet "rack1-k8s-lcm"}} gateway4: {{ gateway_from_subnet "rack1-k8s-lcm" }} k8s-pods: interfaces: [k8s-pods-v] addresses: - {{ip "k8s-pods:k8s-pods-subnet"}} mtu: 9000 routes: - to: 10.199.0.0/22 # aggregated address space for Kubernetes workloads via: {{gateway_from_subnet "rack1-k8s-pods"}}
Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.Note
Before MOSK 23.3, an L2 template requires
clusterRef: <clusterName>in thespecsection. Since MOSK 23.3, this parameter is deprecated and automatically migrated to thecluster.sigs.k8s.io/cluster-name: <clusterName>label.Proceed with the L2 template configuration procedure described in Create an L2 template for a new cluster.
This section contains an exemplary L2 template that demonstrates how to set up bonds and bridges on hosts for your managed clusters.
Configure bonding options using the parameters field. The only
mandatory option is mode. See the example below for details.
Note
You can set any mode supported by netplan and your hardware.
Important
Bond monitoring is disabled in Ubuntu by default. However,
Mirantis highly recommends enabling it using the Media Independent Interface
(MII) monitoring by setting the mii-monitor-interval parameter to a
non-zero value. For details, see Linux documentation: bond monitoring.
The Kubernetes LCM network connects LCM Agents running on nodes to the LCM API
of the management cluster. It is also used for communication between
kubelet and Kubernetes API server inside a Kubernetes cluster.
The MKE components use this network for communication inside a swarm
cluster.
To configure each node with an IP address that will be used for LCM traffic,
use the npTemplate.bridges.k8s-lcm bridge in the L2 template, as
demonstrated in the example below.
Each node of every cluster must have one and only IP address in the LCM network that is allocated from one of the
Subnetobjects having theipam/SVC-k8s-lcmlabel defined. Therefore, allSubnetobjects used for LCM networks must have theipam/SVC-k8s-lcmlabel defined.You can use any interface name for the LCM network traffic. The
Subnetobjects for the LCM network must have theipam/SVC-k8s-lcmlabel. For details, see Service labels and their life cycle.
If you want to use a dedicated network for Kubernetes pods traffic,
configure each node with an IPv4
address that will be used to route the pods traffic between nodes.
To accomplish that, use the npTemplate.bridges.k8s-pods bridge
in the L2 template, as demonstrated in the example below.
As defined in Container Cloud Reference Architecture: Host networking,
this bridge name is reserved for the Kubernetes pods network. When the
k8s-pods bridge is defined in an L2 template, Calico CNI uses that network
for routing the pods traffic between nodes.
You can use a dedicated network for external connection to the Kubernetes
services exposed by the cluster.
If enabled, MetalLB will listen and respond on the dedicated virtual bridge.
To accomplish that, configure each node where metallb-speaker is deployed
with an IPv4 address. For details on selecting nodes for metallb-speaker,
see Configure node selectors for MetalLB speakers. Both the MetalLB IP address ranges and the
IP addresses configured on those nodes must fit in the same CIDR.
The default route on the MOSK nodes that are connected to the external network must be configured with the default gateway in the external network.
Caution
The IP address ranges of the corresponding subnet used in
L2Template for the dedicated virtual brigde must be excluded from the
MetalLB address ranges.
You can configure dedicated networks for the Ceph cluster access and
replication traffic. Set labels on the Subnet CRs for the corresponding
networks, as described in Create subnets.
Container Cloud automatically configures Ceph to use the addresses from these
subnets. Ensure that the addresses are assigned to the storage nodes.
The Subnet objects used to assign IP addresses to these networks
must have corresponding labels ipam/SVC-ceph-public for the
Ceph public (storage access) network and ipam/SVC-ceph-cluster for the
Ceph cluster (storage replication) network.
apiVersion: ipam.mirantis.com/v1alpha1
kind: L2Template
metadata:
name: test-managed
namespace: managed-ns
labels:
cluster.sigs.k8s.io/cluster-name: mosk-cluster-name
spec:
autoIfMappingPrio:
- provision
- eno
- ens
- enp
l3Layout:
- subnetName: mgmt-lcm
scope: global
- subnetName: demo-lcm
scope: namespace
- subnetName: demo-ext
scope: namespace
- subnetName: demo-pods
scope: namespace
- subnetName: demo-ceph-cluster
scope: namespace
- subnetName: demo-ceph-public
scope: namespace
npTemplate: |
version: 2
ethernets:
ten10gbe0s0:
dhcp4: false
dhcp6: false
match:
macaddress: {{mac 2}}
set-name: {{nic 2}}
ten10gbe0s1:
dhcp4: false
dhcp6: false
match:
macaddress: {{mac 3}}
set-name: {{nic 3}}
bonds:
bond0:
interfaces:
- ten10gbe0s0
- ten10gbe0s1
mtu: 9000
parameters:
mode: 802.3ad
mii-monitor-interval: 100
vlans:
k8s-lcm-vlan:
id: 1009
link: bond0
k8s-ext-vlan:
id: 1001
link: bond0
k8s-pods-vlan:
id: 1002
link: bond0
stor-frontend:
id: 1003
link: bond0
stor-backend:
id: 1004
link: bond0
bridges:
k8s-lcm:
interfaces: [k8s-lcm-vlan]
addresses:
- {{ip "k8s-lcm:demo-lcm"}}
routes:
- to: {{ cidr_from_subnet "mgmt-lcm" }}
via: {{ gateway_from_subnet "demo-lcm" }}
k8s-ext:
interfaces: [k8s-ext-vlan]
addresses:
- {{ip "k8s-ext:demo-ext"}}
nameservers:
addresses: {{nameservers_from_subnet "demo-ext"}}
gateway4: {{ gateway_from_subnet "demo-ext" }}
k8s-pods:
interfaces: [k8s-pods-vlan]
addresses:
- {{ip "k8s-pods:demo-pods"}}
ceph-cluster:
interfaces: [stor-backend]
addresses:
- {{ip "ceph-cluster:demo-ceph-cluster"}}
ceph-public:
interfaces: [stor-frontend]
addresses:
- {{ip "ceph-public:demo-ceph-public"}}
Note
Before MOSK 23.3, an L2 template requires
clusterRef: <clusterName> in the spec section. Since MOSK 23.3,
this parameter is deprecated and automatically migrated to the
cluster.sigs.k8s.io/cluster-name: <clusterName> label.
Unsupported since MCC 2.28.0 (17.3.0 and 16.3.0)
Warning
The SubnetPool object is unsupported since Container Cloud
2.28.0 (Cluster releases 17.3.0 and 16.3.0). For details, see
Deprecation Notes: SubnetPool resource management.
This section contains an exemplary L2 template for automatic multiple
subnet creation as described in Automate multiple subnet creation using SubnetPool. This template
also contains the L3Layout section that allows defining the Subnet
scopes and enables auto-creation of the Subnet objects from the
SubnetPool objects. For details about auto-creation of the Subnet
objects see Automate multiple subnet creation using SubnetPool.
For details on how to create L2 templates, see Create an L2 template for a new cluster.
Caution
Do not assign an IP address to the PXE nic 0 NIC explicitly
to prevent the IP duplication during updates. The IP address is
automatically assigned by the bootstrapping engine.
Example of an L2 template for multiple subnets:
apiVersion: ipam.mirantis.com/v1alpha1
kind: L2Template
metadata:
name: test-managed
namespace: managed-ns
labels:
kaas.mirantis.com/provider: baremetal
kaas.mirantis.com/region: region-one
cluster.sigs.k8s.io/cluster-name: my-cluster
spec:
autoIfMappingPrio:
- provision
- eno
- ens
- enp
l3Layout:
- subnetName: lcm-subnet
scope: namespace
- subnetName: subnet-1
subnetPool: kaas-mgmt
scope: namespace
- subnetName: subnet-2
subnetPool: kaas-mgmt
scope: cluster
npTemplate: |
version: 2
ethernets:
onboard1gbe0:
dhcp4: false
dhcp6: false
match:
macaddress: {{mac 0}}
set-name: {{nic 0}}
# IMPORTANT: do not assign an IP address here explicitly
# to prevent IP duplication issues. The IP will be assigned
# automatically by the bootstrapping engine.
# addresses: []
onboard1gbe1:
dhcp4: false
dhcp6: false
match:
macaddress: {{mac 1}}
set-name: {{nic 1}}
ten10gbe0s0:
dhcp4: false
dhcp6: false
match:
macaddress: {{mac 2}}
set-name: {{nic 2}}
addresses:
- {{ip "2:subnet-1"}}
ten10gbe0s1:
dhcp4: false
dhcp6: false
match:
macaddress: {{mac 3}}
set-name: {{nic 3}}
addresses:
- {{ip "3:subnet-2"}}
bridges:
k8s-lcm:
interfaces: [onboard1gbe0]
addresses:
- {{ip "k8s-lcm:lcm-subnet"}}
gateway4: {{gateway_from_subnet "lcm-subnet"}}
nameservers:
addresses: {{nameservers_from_subnet "lcm-subnet"}}
Note
The kaas.mirantis.com/region label is removed from all MOSK
objects in 24.1. Therefore, do not add the label starting with this release.
On existing clusters updated to this release, or if added manually, MOSK
ignores this label.
In the template above, the following networks are defined
in the l3Layout section:
lcm-subnet- the subnet name to use for the LCM network innpTemplate. This subnet is shared between the project clusters because it has thenamespacedscope.Since a subnet pool is not in use, create the corresponding
Subnetobject before machines are attached to cluster manually. For details, see Create subnets for a managed cluster using CLI.Mark this
Subnetwith theipam/SVC-k8s-lcmlabel. The L2 template must contain the definition of the virtual Linux bridge (k8s-lcmin the L2 template example) that is used to set up the LCM network interface. IP addresses for the defined bridge must be assigned from the LCM subnet, which is marked with theipam/SVC-k8s-lcmlabel.Each node of every cluster must have one and only IP address in the LCM network that is allocated from one of the
Subnetobjects having theipam/SVC-k8s-lcmlabel defined. Therefore, allSubnetobjects used for LCM networks must have theipam/SVC-k8s-lcmlabel defined.You can use any interface name for the LCM network traffic. The
Subnetobjects for the LCM network must have theipam/SVC-k8s-lcmlabel. For details, see Service labels and their life cycle.
subnet-1- unless already created, this subnet will be created from thekaas-mgmtsubnet pool. The subnet name must be unique within the project. This subnet is shared between the project clusters.subnet-2- will be created from thekaas-mgmtsubnet pool. This subnet has theclusterscope. Therefore, the real name of theSubnetCR object consists of the subnet name defined inl3Layoutand the cluster UID. But thenpTemplatesection of the L2 template must contain only the subnet name defined inl3Layout. The subnets of theclusterscope are not shared between clusters.
Caution
Using the l3Layout section, define all subnets that are used
in the npTemplate section.
Defining only part of subnets is not allowed.
If labelSelector is used in l3Layout, use any custom
label name that differs from system names. This allows for easier
cluster scaling in case of adding new subnets as described in
Expand IP addresses capacity in an existing cluster.
Mirantis recommends using a unique label prefix such as
user-defined/.
Example of a complete template configuration for cluster creation¶
The following example contains all required objects of an advanced network and host configuration for a managed cluster.
The procedure below contains:
Various
.yamlobjects to be applied with a managed clusterkubeconfigUseful comments inside the
.yamlexample filesExample hardware and configuration data, such as
network,disk,auth, that must be updated accordingly to fit your cluster configurationExample templates, such as
l2templateandbaremetalhostprofline, that illustrate how to implement a specific configuration
Caution
The exemplary configuration described below is not production ready and is provided for illustration purposes only.
For illustration purposes, all files provided in this exemplary procedure are named by the Kubernetes object types:
Note
Before update of the management cluster to Container Cloud 2.29.0
(Cluster release 16.4.0), instead of BareMetalHostInventory, use the
BareMetalHost object. For details, see BareMetalHost resource.
Caution
While the Cluster release of the management cluster is 16.4.0,
BareMetalHostInventory operations are allowed to
m:kaas@management-admin only. This limitation is lifted once the
management cluster is updated to the Cluster release 16.4.1 or later.
managed-ns_BareMetalHostInventory_cz7700-managed-cluster-control-noefi.yaml
managed-ns_BareMetalHostInventory_cz7741-managed-cluster-control-noefi.yaml
managed-ns_BareMetalHostInventory_cz7743-managed-cluster-control-noefi.yaml
managed-ns_BareMetalHostInventory_cz812-managed-cluster-storage-worker-noefi.yaml
managed-ns_BareMetalHostInventory_cz813-managed-cluster-storage-worker-noefi.yaml
managed-ns_BareMetalHostInventory_cz814-managed-cluster-storage-worker-noefi.yaml
managed-ns_BareMetalHostInventory_cz815-managed-cluster-worker-noefi.yaml
managed-ns_BareMetalHostProfile_bmhp-cluster-default.yaml
managed-ns_BareMetalHostProfile_worker-storage1.yaml
managed-ns_Cluster_managed-cluster.yaml
managed-ns_KaaSCephCluster_ceph-cluster-managed-cluster.yaml
managed-ns_L2Template_bm-1490-template-controls-netplan-cz7700-pxebond.yaml
managed-ns_L2Template_bm-1490-template-controls-netplan.yaml
managed-ns_L2Template_bm-1490-template-workers-netplan.yaml
managed-ns_Machine_cz7700-managed-cluster-control-noefi-.yaml
managed-ns_Machine_cz7741-managed-cluster-control-noefi-.yaml
managed-ns_Machine_cz7743-managed-cluster-control-noefi-.yaml
managed-ns_Machine_cz812-managed-cluster-storage-worker-noefi-.yaml
managed-ns_Machine_cz813-managed-cluster-storage-worker-noefi-.yaml
managed-ns_Machine_cz814-managed-cluster-storage-worker-noefi-.yaml
managed-ns_Machine_cz815-managed-cluster-worker-noefi-.yaml
managed-ns_PublicKey_managed-cluster-key.yaml
managed-ns_cz7700-cred.yaml
managed-ns_cz7741-cred.yaml
managed-ns_cz7743-cred.yaml
managed-ns_cz812-cred.yaml
managed-ns_cz813-cred.yaml
managed-ns_cz814-cred.yaml
managed-ns_cz815-cred.yaml
managed-ns_Subnet_lcm-nw.yaml
managed-ns_Subnet_metallb-public-for-managed.yaml (obsolete)
managed-ns_Subnet_metallb-public-for-extiface.yaml
managed-ns_MetalLBConfig-lb-managed.yaml
managed-ns_MetalLBConfigTemplate-lb-managed-template.yaml (obsolete)
managed-ns_Subnet_storage-backend.yaml
managed-ns_Subnet_storage-frontend.yaml
default_Namespace_managed-ns.yaml
Caution
The procedure below presumes that you apply each new .yaml
file using kubectl create -f <file_name.yaml>.
To create an example configuration for a managed cluster creation:
Verify that you have configured the following items:
All
bmhnodes for PXE boot as described in Add a bare metal host using CLIAll physical NICs of the
bmhnodesAll required physical subnets and routing
Create an empty
.yamlfile with thenamespaceobject:apiVersion: v1
Select from the following options:
Create the required number of
.yamlfiles with theBareMetalHostCredentialobjects for eachbmhnode with the uniquenameand authenticationdata. The following example contains oneBareMetalHostCredentialobject:Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.managed-ns_cz815-cred.yamlapiVersion: kaas.mirantis.com/v1alpha1 kind: BareMetalHostCredential metadata: name: cz815-cred namespace: managed-ns labels: kaas.mirantis.com/region: region-one spec: username: admin password: value: supersecret
Create the required number of
.yamlfiles with theSecretobjects for eachbmhnode with the uniquenameand authenticationdata. The following example contains oneSecretobject:managed-ns_cz815-cred.yamlapiVersion: v1 data: password: YWRtaW4= username: ZW5naW5lZXI= kind: Secret metadata: labels: kaas.mirantis.com/credentials: 'true' kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: cz815-cred namespace: managed-ns
Create a set of files with the
bmhnodes configuration:managed-ns_BareMetalHostInventory_cz7700-managed-cluster-control-noefi.yamlapiVersion: kaas.mirantis.com/v1alpha1 kind: BareMetalHostInventory metadata: annotations: inspect.metal3.io/hardwaredetails-storage-sort-term: hctl ASC, wwn ASC, by_id ASC, name ASC labels: cluster.sigs.k8s.io/cluster-name: managed-cluster # we will use those label, to link machine to exact bmh node kaas.mirantis.com/baremetalhost-id: cz7700 kaas.mirantis.com/provider: baremetal name: cz7700-managed-cluster-control-noefi namespace: managed-ns spec: bmc: address: 192.168.1.12 bmhCredentialsName: 'cz7740-cred' bootMACAddress: 0c:c4:7a:34:52:04 bootMode: legacy online: true
managed-ns_BareMetalHostInventory_cz7741-managed-cluster-control-noefi.yamlapiVersion: kaas.mirantis.com/v1alpha1 kind: BareMetalHostInventory metadata: annotations: inspect.metal3.io/hardwaredetails-storage-sort-term: hctl ASC, wwn ASC, by_id ASC, name ASC labels: cluster.sigs.k8s.io/cluster-name: managed-cluster kaas.mirantis.com/baremetalhost-id: cz7741 kaas.mirantis.com/provider: baremetal name: cz7741-managed-cluster-control-noefi namespace: managed-ns spec: bmc: address: 192.168.1.76 bmhCredentialsName: 'cz7741-cred' bootMACAddress: 0c:c4:7a:34:92:f4 bootMode: legacy online: true
managed-ns_BareMetalHostInventory_cz7743-managed-cluster-control-noefi.yamlapiVersion: kaas.mirantis.com/v1alpha1 kind: BareMetalHostInventory metadata: annotations: inspect.metal3.io/hardwaredetails-storage-sort-term: hctl ASC, wwn ASC, by_id ASC, name ASC labels: cluster.sigs.k8s.io/cluster-name: managed-cluster kaas.mirantis.com/baremetalhost-id: cz7743 kaas.mirantis.com/provider: baremetal name: cz7743-managed-cluster-control-noefi namespace: managed-ns spec: bmc: address: 192.168.1.78 bmhCredentialsName: 'cz7743-cred' bootMACAddress: 0c:c4:7a:34:66:fc bootMode: legacy online: true
managed-ns_BareMetalHostInventory_cz812-managed-cluster-storage-worker-noefi.yamlapiVersion: kaas.mirantis.com/v1alpha1 kind: BareMetalHostInventory metadata: annotations: inspect.metal3.io/hardwaredetails-storage-sort-term: hctl ASC, wwn ASC, by_id ASC, name ASC labels: cluster.sigs.k8s.io/cluster-name: managed-cluster kaas.mirantis.com/baremetalhost-id: cz812 kaas.mirantis.com/provider: baremetal name: cz812-managed-cluster-storage-worker-noefi namespace: managed-ns spec: bmc: address: 192.168.1.182 bmhCredentialsName: 'cz812-cred' bootMACAddress: 0c:c4:7a:bc:ff:2e bootMode: legacy online: true
managed-ns_BareMetalHostInventory_cz813-managed-cluster-storage-worker-noefi.yamlapiVersion: kaas.mirantis.com/v1alpha1 kind: BareMetalHostInventory metadata: annotations: inspect.metal3.io/hardwaredetails-storage-sort-term: hctl ASC, wwn ASC, by_id ASC, name ASC labels: cluster.sigs.k8s.io/cluster-name: managed-cluster kaas.mirantis.com/baremetalhost-id: cz813 kaas.mirantis.com/provider: baremetal name: cz813-managed-cluster-storage-worker-noefi namespace: managed-ns spec: bmc: address: 192.168.1.183 bmhCredentialsName: 'cz813-cred' bootMACAddress: 0c:c4:7a:bc:fe:36 bootMode: legacy online: true
managed-ns_BareMetalHostInventory_cz814-managed-cluster-storage-worker-noefi.yamlapiVersion: kaas.mirantis.com/v1alpha1 kind: BareMetalHostInventory metadata: annotations: inspect.metal3.io/hardwaredetails-storage-sort-term: hctl ASC, wwn ASC, by_id ASC, name ASC labels: cluster.sigs.k8s.io/cluster-name: managed-cluster kaas.mirantis.com/baremetalhost-id: cz814 kaas.mirantis.com/provider: baremetal name: cz814-managed-cluster-storage-worker-noefi namespace: managed-ns spec: bmc: address: 192.168.1.184 bmhCredentialsName: 'cz814-cred' bootMACAddress: 0c:c4:7a:bc:fb:20 bootMode: legacy online: true
managed-ns_BareMetalHostInventory_cz815-managed-cluster-worker-noefi.yamlapiVersion: kaas.mirantis.com/v1alpha1 kind: BareMetalHostInventory metadata: annotations: inspect.metal3.io/hardwaredetails-storage-sort-term: hctl ASC, wwn ASC, by_id ASC, name ASC labels: cluster.sigs.k8s.io/cluster-name: managed-cluster kaas.mirantis.com/baremetalhost-id: cz815 kaas.mirantis.com/provider: baremetal name: cz815-managed-cluster-worker-noefi namespace: managed-ns spec: bmc: address: 192.168.1.185 bmhCredentialsName: 'cz815-cred' bootMACAddress: 0c:c4:7a:bc:fc:3e bootMode: legacy online: true
managed-ns_BareMetalHost_cz7700-managed-cluster-control-noefi.yamlapiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/controlplane: controlplane # we will use those label, to link machine to exact bmh node kaas.mirantis.com/baremetalhost-id: cz7700 kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one annotations: kaas.mirantis.com/baremetalhost-credentials-name: cz7700-cred name: cz7700-managed-cluster-control-noefi namespace: managed-ns spec: bmc: address: 192.168.1.12 # credentialsName is updated automatically during cluster deployment credentialsName: '' bootMACAddress: 0c:c4:7a:34:52:04 bootMode: legacy online: true
managed-ns_BareMetalHost_cz7741-managed-cluster-control-noefi.yamlapiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/controlplane: controlplane kaas.mirantis.com/baremetalhost-id: cz7741 kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one annotations: kaas.mirantis.com/baremetalhost-credentials-name: cz7741-cred name: cz7741-managed-cluster-control-noefi namespace: managed-ns spec: bmc: address: 192.168.1.76 credentialsName: '' bootMACAddress: 0c:c4:7a:34:92:f4 bootMode: legacy online: true
managed-ns_BareMetalHost_cz7743-managed-cluster-control-noefi.yamlapiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/controlplane: controlplane kaas.mirantis.com/baremetalhost-id: cz7743 kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one annotations: kaas.mirantis.com/baremetalhost-credentials-name: cz7743-cred name: cz7743-managed-cluster-control-noefi namespace: managed-ns spec: bmc: address: 192.168.1.78 credentialsName: '' bootMACAddress: 0c:c4:7a:34:66:fc bootMode: legacy online: true
managed-ns_BareMetalHost_cz812-managed-cluster-storage-worker-noefi.yamlapiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/worker: worker kaas.mirantis.com/baremetalhost-id: cz812 kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one annotations: kaas.mirantis.com/baremetalhost-credentials-name: cz812-cred name: cz812-managed-cluster-storage-worker-noefi namespace: managed-ns spec: bmc: address: 192.168.1.182 credentialsName: '' bootMACAddress: 0c:c4:7a:bc:ff:2e bootMode: legacy online: true
managed-ns_BareMetalHost_cz813-managed-cluster-storage-worker-noefi.yamlapiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/worker: worker kaas.mirantis.com/baremetalhost-id: cz813 kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one annotations: kaas.mirantis.com/baremetalhost-credentials-name: cz813-cred name: cz813-managed-cluster-storage-worker-noefi namespace: managed-ns spec: bmc: address: 192.168.1.183 credentialsName: '' bootMACAddress: 0c:c4:7a:bc:fe:36 bootMode: legacy online: true
managed-ns_BareMetalHost_cz814-managed-cluster-storage-worker-noefi.yamlapiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/worker: worker kaas.mirantis.com/baremetalhost-id: cz814 kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one annotations: kaas.mirantis.com/baremetalhost-credentials-name: cz814-cred name: cz814-managed-cluster-storage-worker-noefi namespace: managed-ns spec: bmc: address: 192.168.1.184 credentialsName: '' bootMACAddress: 0c:c4:7a:bc:fb:20 bootMode: legacy online: true
managed-ns_BareMetalHost_cz815-managed-cluster-worker-noefi.yamlapiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/worker: worker kaas.mirantis.com/baremetalhost-id: cz815 kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one annotations: kaas.mirantis.com/baremetalhost-credentials-name: cz815-cred name: cz815-managed-cluster-worker-noefi namespace: managed-ns spec: bmc: address: 192.168.1.185 credentialsName: '' bootMACAddress: 0c:c4:7a:bc:fc:3e bootMode: legacy online: true
managed-ns_BareMetalHost_cz7700-managed-cluster-control-noefi.yamlapiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/controlplane: controlplane # we will use those label, to link machine to exact bmh node kaas.mirantis.com/baremetalhost-id: cz7700 kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: cz7700-managed-cluster-control-noefi namespace: managed-ns spec: bmc: address: 192.168.1.12 # The secret for credentials requires the username and password # keys in the Base64 encoding. credentialsName: cz7700-cred bootMACAddress: 0c:c4:7a:34:52:04 bootMode: legacy online: true
managed-ns_BareMetalHost_cz7741-managed-cluster-control-noefi.yamlapiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/controlplane: controlplane kaas.mirantis.com/baremetalhost-id: cz7741 kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: cz7741-managed-cluster-control-noefi namespace: managed-ns spec: bmc: address: 192.168.1.76 credentialsName: cz7741-cred bootMACAddress: 0c:c4:7a:34:92:f4 bootMode: legacy online: true
managed-ns_BareMetalHost_cz7743-managed-cluster-control-noefi.yamlapiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/controlplane: controlplane kaas.mirantis.com/baremetalhost-id: cz7743 kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: cz7743-managed-cluster-control-noefi namespace: managed-ns spec: bmc: address: 192.168.1.78 credentialsName: cz7743-cred bootMACAddress: 0c:c4:7a:34:66:fc bootMode: legacy online: true
managed-ns_BareMetalHost_cz812-managed-cluster-storage-worker-noefi.yamlapiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/worker: worker kaas.mirantis.com/baremetalhost-id: cz812 kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: cz812-managed-cluster-storage-worker-noefi namespace: managed-ns spec: bmc: address: 192.168.1.182 credentialsName: cz812-cred bootMACAddress: 0c:c4:7a:bc:ff:2e bootMode: legacy online: true
managed-ns_BareMetalHost_cz813-managed-cluster-storage-worker-noefi.yamlapiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/worker: worker kaas.mirantis.com/baremetalhost-id: cz813 kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: cz813-managed-cluster-storage-worker-noefi namespace: managed-ns spec: bmc: address: 192.168.1.183 credentialsName: cz813-cred bootMACAddress: 0c:c4:7a:bc:fe:36 bootMode: legacy online: true
managed-ns_BareMetalHost_cz814-managed-cluster-storage-worker-noefi.yamlapiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/worker: worker kaas.mirantis.com/baremetalhost-id: cz814 kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: cz814-managed-cluster-storage-worker-noefi namespace: managed-ns spec: bmc: address: 192.168.1.184 credentialsName: cz814-cre bootMACAddress: 0c:c4:7a:bc:fb:20 bootMode: legacy online: true
managed-ns_BareMetalHost_cz815-managed-cluster-worker-noefi.yamlapiVersion: metal3.io/v1alpha1 kind: BareMetalHost metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/worker: worker kaas.mirantis.com/baremetalhost-id: cz815 kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: cz815-managed-cluster-worker-noefi namespace: managed-ns spec: bmc: address: 192.168.1.185 credentialsName: cz815-cred bootMACAddress: 0c:c4:7a:bc:fc:3e bootMode: legacy online: true
Verify that the
inspectingphase has started:KUBECONFIG=kubeconfig kubectl -n managed-ns get bmh -o wide
Example of system response:
NAME STATUS STATE CONSUMER BMC BOOTMODE ONLINE ERROR REGION cz7700-managed-cluster-control-noefi OK inspecting 192.168.1.12 legacy true region-one cz7741-managed-cluster-control-noefi OK inspecting 192.168.1.76 legacy true region-one cz7743-managed-cluster-control-noefi OK inspecting 192.168.1.78 legacy true region-one cz812-managed-cluster-storage-worker-noefi OK inspecting 192.168.1.182 legacy true region-one
Wait for inspection to complete. Usually, it takes up to 15 minutes.
Collect the
bmhhardware information to create thel2templateandbmhobjects:KUBECONFIG=kubeconfig kubectl -n managed-ns get bmh -o wide
Example of system response:
NAME STATUS STATE CONSUMER BMC BOOTMODE ONLINE ERROR REGION cz7700-managed-cluster-control-noefi OK available 192.168.1.12 legacy true region-one cz7741-managed-cluster-control-noefi OK available 192.168.1.76 legacy true region-one cz7743-managed-cluster-control-noefi OK available 192.168.1.78 legacy true region-one cz812-managed-cluster-storage-worker-noefi OK available 192.168.1.182 legacy true region-one
KUBECONFIG=kubeconfig kubectl -n managed-ns get bmh cz7700-managed-cluster-control-noefi -o yaml | less
Example of system response:
.. nics: - ip: "" mac: 0c:c4:7a:1d:f4:a6 model: 0x8086 0x10fb # discovered interfaces name: ens4f0 pxe: false # temporary PXE address discovered from baremetal-mgmt - ip: 172.16.170.30 mac: 0c:c4:7a:34:52:04 model: 0x8086 0x1521 name: enp9s0f0 pxe: true # duplicates temporary PXE address discovered from baremetal-mgmt # since we have fallback-bond configured on host - ip: 172.16.170.33 mac: 0c:c4:7a:34:52:05 model: 0x8086 0x1521 # discovered interfaces name: enp9s0f1 pxe: false ... storage: - by_path: /dev/disk/by-path/pci-0000:00:1f.2-ata-1 model: Samsung SSD 850 name: /dev/sda rotational: false sizeBytes: 500107862016 - by_path: /dev/disk/by-path/pci-0000:00:1f.2-ata-2 model: Samsung SSD 850 name: /dev/sdb rotational: false sizeBytes: 500107862016 ....
Create bare metal host profiles:
managed-ns_BareMetalHostProfile_bmhp-cluster-default.yamlapiVersion: metal3.io/v1alpha1 kind: BareMetalHostProfile metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster # This label indicates that this profile will be default in # namespaces, so machines w\o exact profile selecting will use # this template kaas.mirantis.com/defaultBMHProfile: 'true' kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: bmhp-cluster-default namespace: managed-ns spec: devices: - device: byPath: /dev/disk/by-path/pci-0000:00:1f.2-ata-1 minSize: 120Gi wipe: true partitions: - name: bios_grub partflags: - bios_grub size: 4Mi wipe: true - name: uefi partflags: - esp size: 200Mi wipe: true - name: config-2 size: 64Mi wipe: true - name: lvm_dummy_part size: 1Gi wipe: true - name: lvm_root_part size: 0 wipe: true - device: byPath: /dev/disk/by-path/pci-0000:00:1f.2-ata-2 minSize: 30Gi wipe: true - device: byPath: /dev/disk/by-path/pci-0000:00:1f.2-ata-3 minSize: 30Gi wipe: true partitions: - name: lvm_lvp_part size: 0 wipe: true - device: byPath: /dev/disk/by-path/pci-0000:00:1f.2-ata-4 wipe: true fileSystems: - fileSystem: vfat partition: config-2 - fileSystem: vfat mountPoint: /boot/efi partition: uefi - fileSystem: ext4 logicalVolume: root mountPoint: / - fileSystem: ext4 logicalVolume: lvp mountPoint: /mnt/local-volumes/ grubConfig: defaultGrubOptions: - GRUB_DISABLE_RECOVERY="true" - GRUB_PRELOAD_MODULES=lvm - GRUB_TIMEOUT=30 kernelParameters: modules: - content: 'options kvm_intel nested=1' filename: kvm_intel.conf sysctl: # For the list of options prohibited to change, refer to # https://docs.mirantis.com/mke/3.7/install/predeployment/set-up-kernel-default-protections.html fs.aio-max-nr: '1048576' fs.file-max: '9223372036854775807' fs.inotify.max_user_instances: '4096' kernel.core_uses_pid: '1' kernel.dmesg_restrict: '1' net.ipv4.conf.all.rp_filter: '0' net.ipv4.conf.default.rp_filter: '0' net.ipv4.conf.k8s-ext.rp_filter: '0' net.ipv4.conf.k8s-ext.rp_filter: '0' net.ipv4.conf.m-pub.rp_filter: '0' vm.max_map_count: '262144' logicalVolumes: - name: root size: 0 vg: lvm_root - name: lvp size: 0 vg: lvm_lvp postDeployScript: | #!/bin/bash -ex # used for test-debug only! echo "root:r00tme" | sudo chpasswd echo 'ACTION=="add|change", KERNEL=="sd[a-z]", ATTR{queue/rotational}=="0", ATTR{queue/scheduler}="deadline"' > /etc/udev/rules.d/60-ssd-scheduler.rules echo $(date) 'post_deploy_script done' >> /root/post_deploy_done preDeployScript: | #!/bin/bash -ex echo "$(date) pre_deploy_script done" >> /root/pre_deploy_done volumeGroups: - devices: - partition: lvm_root_part name: lvm_root - devices: - partition: lvm_lvp_part name: lvm_lvp - devices: - partition: lvm_dummy_part # here we can create lvm, but do not mount or format it somewhere name: lvm_forawesomeapp
managed-ns_BareMetalHostProfile_worker-storage1.yamlapiVersion: metal3.io/v1alpha1 kind: BareMetalHostProfile metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: worker-storage1 namespace: managed-ns spec: devices: - device: minSize: 120Gi wipe: true partitions: - name: bios_grub partflags: - bios_grub size: 4Mi wipe: true - name: uefi partflags: - esp size: 200Mi wipe: true - name: config-2 size: 64Mi wipe: true # Create dummy partition w\o mounting - name: lvm_dummy_part size: 1Gi wipe: true - name: lvm_root_part size: 0 wipe: true - device: # Will be used for Ceph, so required to be wiped byPath: /dev/disk/by-path/pci-0000:00:1f.2-ata-1 minSize: 30Gi wipe: true - device: byPath: /dev/disk/by-path/pci-0000:00:1f.2-ata-2 minSize: 30Gi wipe: true partitions: - name: lvm_lvp_part size: 0 wipe: true - device: byPath: /dev/disk/by-path/pci-0000:00:1f.2-ata-3 wipe: true - device: byPath: /dev/disk/by-path/pci-0000:00:1f.2-ata-4 minSize: 30Gi wipe: true partitions: - name: lvm_lvp_part_sdf wipe: true size: 0 fileSystems: - fileSystem: vfat partition: config-2 - fileSystem: vfat mountPoint: /boot/efi partition: uefi - fileSystem: ext4 logicalVolume: root mountPoint: / - fileSystem: ext4 logicalVolume: lvp mountPoint: /mnt/local-volumes/ grubConfig: defaultGrubOptions: - GRUB_DISABLE_RECOVERY="true" - GRUB_PRELOAD_MODULES=lvm - GRUB_TIMEOUT=30 kernelParameters: modules: - content: 'options kvm_intel nested=1' filename: kvm_intel.conf sysctl: # For the list of options prohibited to change, refer to # https://docs.mirantis.com/mke/3.6/install/predeployment/set-up-kernel-default-protections.html fs.aio-max-nr: '1048576' fs.file-max: '9223372036854775807' fs.inotify.max_user_instances: '4096' kernel.core_uses_pid: '1' kernel.dmesg_restrict: '1' net.ipv4.conf.all.rp_filter: '0' net.ipv4.conf.default.rp_filter: '0' net.ipv4.conf.k8s-ext.rp_filter: '0' net.ipv4.conf.k8s-ext.rp_filter: '0' net.ipv4.conf.m-pub.rp_filter: '0' vm.max_map_count: '262144' logicalVolumes: - name: root size: 0 vg: lvm_root - name: lvp size: 0 vg: lvm_lvp postDeployScript: | #!/bin/bash -ex # used for test-debug only! That would allow operator to logic via TTY. echo "root:r00tme" | sudo chpasswd # Just an example for enforcing "ssd" disks to be switched to use "deadline" i\o scheduler. echo 'ACTION=="add|change", KERNEL=="sd[a-z]", ATTR{queue/rotational}=="0", ATTR{queue/scheduler}="deadline"' > /etc/udev/ rules.d/60-ssd-scheduler.rules echo $(date) 'post_deploy_script done' >> /root/post_deploy_done preDeployScript: | #!/bin/bash -ex echo "$(date) pre_deploy_script done" >> /root/pre_deploy_done volumeGroups: - devices: - partition: lvm_root_part name: lvm_root - devices: - partition: lvm_lvp_part - partition: lvm_lvp_part_sdf name: lvm_lvp - devices: - partition: lvm_dummy_part name: lvm_forawesomeapp
Note
If you mount the
/vardirectory, before configuringBareMetalHostProfile, review Mounting recommendations for the /var directory.Create the
L2Templateobjects:managed-ns_L2Template_bm-1490-template-controls-netplan.yamlapiVersion: ipam.mirantis.com/v1alpha1 kind: L2Template metadata: labels: bm-1490-template-controls-netplan: anymagicstring cluster.sigs.k8s.io/cluster-name: managed-cluster kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: bm-1490-template-controls-netplan namespace: managed-ns spec: ifMapping: - enp9s0f0 - enp9s0f1 - eno1 - ens3f1 l3Layout: - scope: namespace subnetName: lcm-nw - scope: namespace subnetName: storage-frontend - scope: namespace subnetName: storage-backend - scope: namespace subnetName: metallb-public-for-extiface npTemplate: |- version: 2 ethernets: {{nic 0}}: dhcp4: false dhcp6: false match: macaddress: {{mac 0}} set-name: {{nic 0}} mtu: 1500 {{nic 1}}: dhcp4: false dhcp6: false match: macaddress: {{mac 1}} set-name: {{nic 1}} mtu: 1500 {{nic 2}}: dhcp4: false dhcp6: false match: macaddress: {{mac 2}} set-name: {{nic 2}} mtu: 1500 {{nic 3}}: dhcp4: false dhcp6: false match: macaddress: {{mac 3}} set-name: {{nic 3}} mtu: 1500 bonds: bond0: parameters: mode: 802.3ad #transmit-hash-policy: layer3+4 #mii-monitor-interval: 100 interfaces: - {{ nic 0 }} - {{ nic 1 }} bond1: parameters: mode: 802.3ad #transmit-hash-policy: layer3+4 #mii-monitor-interval: 100 interfaces: - {{ nic 2 }} - {{ nic 3 }} vlans: stor-f: id: 1494 link: bond1 addresses: - {{ip "stor-f:storage-frontend"}} stor-b: id: 1489 link: bond1 addresses: - {{ip "stor-b:storage-backend"}} m-pub: id: 1491 link: bond0 bridges: k8s-ext: interfaces: [m-pub] addresses: - {{ ip "k8s-ext:metallb-public-for-extiface" }} k8s-lcm: dhcp4: false dhcp6: false gateway4: {{ gateway_from_subnet "lcm-nw" }} addresses: - {{ ip "k8s-lcm:lcm-nw" }} nameservers: addresses: [ 172.18.176.6 ] interfaces: - bond0
managed-ns_L2Template_bm-1490-template-workers-netplan.yamlapiVersion: ipam.mirantis.com/v1alpha1 kind: L2Template metadata: labels: bm-1490-template-workers-netplan: anymagicstring cluster.sigs.k8s.io/cluster-name: managed-cluster kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: bm-1490-template-workers-netplan namespace: managed-ns spec: ifMapping: - eno1 - eno2 - ens7f0 - ens7f1 l3Layout: - scope: namespace subnetName: lcm-nw - scope: namespace subnetName: storage-frontend - scope: namespace subnetName: storage-backend - scope: namespace subnetName: metallb-public-for-extiface npTemplate: |- version: 2 ethernets: {{nic 0}}: match: macaddress: {{mac 0}} set-name: {{nic 0}} mtu: 1500 {{nic 1}}: dhcp4: false dhcp6: false match: macaddress: {{mac 1}} set-name: {{nic 1}} mtu: 1500 {{nic 2}}: dhcp4: false dhcp6: false match: macaddress: {{mac 2}} set-name: {{nic 2}} mtu: 1500 {{nic 3}}: dhcp4: false dhcp6: false match: macaddress: {{mac 3}} set-name: {{nic 3}} mtu: 1500 bonds: bond0: interfaces: - {{ nic 1 }} bond1: parameters: mode: 802.3ad #transmit-hash-policy: layer3+4 #mii-monitor-interval: 100 interfaces: - {{ nic 2 }} - {{ nic 3 }} vlans: stor-f: id: 1494 link: bond1 addresses: - {{ip "stor-f:storage-frontend"}} stor-b: id: 1489 link: bond1 addresses: - {{ip "stor-b:storage-backend"}} m-pub: id: 1491 link: {{ nic 1 }} bridges: k8s-lcm: interfaces: - {{ nic 0 }} gateway4: {{ gateway_from_subnet "lcm-nw" }} addresses: - {{ ip "k8s-lcm:lcm-nw" }} nameservers: addresses: [ 172.18.176.6 ] k8s-ext: interfaces: [m-pub]
managed-ns_L2Template_bm-1490-template-controls-netplan-cz7700-pxebond.yamlapiVersion: ipam.mirantis.com/v1alpha1 kind: L2Template metadata: labels: bm-1490-template-controls-netplan-cz7700-pxebond: anymagicstring cluster.sigs.k8s.io/cluster-name: managed-cluster kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: bm-1490-template-controls-netplan-cz7700-pxebond namespace: managed-ns spec: ifMapping: - enp9s0f0 - enp9s0f1 - eno1 - ens3f1 l3Layout: - scope: namespace subnetName: lcm-nw - scope: namespace subnetName: storage-frontend - scope: namespace subnetName: storage-backend - scope: namespace subnetName: metallb-public-for-extiface npTemplate: |- version: 2 ethernets: {{nic 0}}: dhcp4: false dhcp6: false match: macaddress: {{mac 0}} set-name: {{nic 0}} mtu: 1500 {{nic 1}}: dhcp4: false dhcp6: false match: macaddress: {{mac 1}} set-name: {{nic 1}} mtu: 1500 {{nic 2}}: dhcp4: false dhcp6: false match: macaddress: {{mac 2}} set-name: {{nic 2}} mtu: 1500 {{nic 3}}: dhcp4: false dhcp6: false match: macaddress: {{mac 3}} set-name: {{nic 3}} mtu: 1500 bonds: bond0: parameters: mode: 802.3ad #transmit-hash-policy: layer3+4 #mii-monitor-interval: 100 interfaces: - {{ nic 0 }} - {{ nic 1 }} bond1: parameters: mode: 802.3ad #transmit-hash-policy: layer3+4 #mii-monitor-interval: 100 interfaces: - {{ nic 2 }} - {{ nic 3 }} vlans: stor-f: id: 1494 link: bond1 addresses: - {{ip "stor-f:storage-frontend"}} stor-b: id: 1489 link: bond1 addresses: - {{ip "stor-b:storage-backend"}} m-pub: id: 1491 link: bond0 bridges: k8s-ext: interfaces: [m-pub] addresses: - {{ ip "k8s-ext:metallb-public-for-extiface" }} k8s-lcm: dhcp4: false dhcp6: false gateway4: {{ gateway_from_subnet "lcm-nw" }} addresses: - {{ ip "k8s-lcm:lcm-nw" }} nameservers: addresses: [ 172.18.176.6 ] interfaces: - bond0
Create the
Subnetobjects:managed-ns_Subnet_lcm-nw.yamlapiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster ipam/SVC-k8s-lcm: '1' kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: lcm-nw namespace: managed-ns spec: cidr: 172.16.170.0/24 excludeRanges: - 172.16.170.150 gateway: 172.16.170.1 includeRanges: - 172.16.170.150-172.16.170.250
managed-ns_Subnet_metallb-public-for-extiface.yamlapiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: metallb-public-for-extiface namespace: managed-ns spec: cidr: 172.16.168.0/24 gateway: 172.16.168.1 includeRanges: - 172.16.168.10-172.16.168.30
managed-ns_Subnet_storage-backend.yamlapiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster ipam/SVC-ceph-cluster: '1' kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: storage-backend namespace: managed-ns spec: cidr: 10.12.0.0/24
managed-ns_Subnet_storage-frontend.yamlapiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster ipam/SVC-ceph-public: '1' kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: storage-frontend namespace: managed-ns spec: cidr: 10.12.1.0/24
Create MetalLB configuration objects:
managed-ns_MetalLBConfig-lb-managed.yamlapiVersion: kaas.mirantis.com/v1alpha1 kind: MetalLBConfig metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: lb-managed namespace: managed-ns spec: ipAddressPools: - name: services spec: addresses: - 172.16.168.31-172.16.168.50 autoAssign: true avoidBuggyIPs: false l2Advertisements: - name: services spec: interfaces: - k8s-ext ipAddressPools: - services
managed-ns_Subnet_metallb-public-for-managed.yamlapiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster ipam/SVC-MetalLB: '1' kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: metallb-public-for-managed namespace: managed-ns spec: cidr: 172.16.168.0/24 includeRanges: - 172.16.168.31-172.16.168.50
managed-ns_MetalLBConfig-lb-managed.yamlNote
Applies since Container Cloud 2.21.0 and 2.21.1 for MOSK as TechPreview and since 2.24.0 as GA for management clusters. For managed clusters, is generally available since Container Cloud 2.25.0.
apiVersion: kaas.mirantis.com/v1alpha1 kind: MetalLBConfig metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: lb-managed namespace: managed-ns spec: templateName: lb-managed-template
managed-ns_MetalLBConfigTemplate-lb-managed-template.yamlNote
The
MetalLBConfigTemplateobject is available as Technology Preview since Container Cloud 2.24.0 (Cluster release 14.0.0) and is generally available since Container Cloud 2.25.0 (Cluster releases 17.0.0 and 16.0.0).apiVersion: ipam.mirantis.com/v1alpha1 kind: MetalLBConfigTemplate metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: lb-managed-template namespace: managed-ns spec: templates: l2Advertisements: | - name: services spec: ipAddressPools: - services
managed-ns_Subnet_metallb-public-for-managed.yamlapiVersion: ipam.mirantis.com/v1alpha1 kind: Subnet metadata: labels: cluster.sigs.k8s.io/cluster-name: managed-cluster ipam/SVC-MetalLB: '1' kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: metallb-public-for-managed namespace: managed-ns spec: cidr: 172.16.168.0/24 includeRanges: - 172.16.168.31-172.16.168.50
Create the
PublicKeyobject for a managed cluster connection. For details, see PublicKey resource.managed-ns_PublicKey_managed-cluster-key.yamlapiVersion: kaas.mirantis.com/v1alpha1 kind: PublicKey metadata: name: managed-cluster-key namespace: managed-ns spec: publicKey: ssh-rsa AAEXAMPLEXXX
Create the
Clusterobject. For details, see Cluster resource.managed-ns_Cluster_managed-cluster.yamlapiVersion: cluster.k8s.io/v1alpha1 kind: Cluster metadata: annotations: kaas.mirantis.com/lcm: 'true' labels: kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one name: managed-cluster namespace: managed-ns spec: clusterNetwork: pods: cidrBlocks: - 192.169.0.0/16 serviceDomain: '' services: cidrBlocks: - 10.232.0.0/18 providerSpec: value: apiVersion: baremetal.k8s.io/v1alpha1 dedicatedControlPlane: false helmReleases: - name: ceph-controller - enabled: true name: stacklight values: alertmanagerSimpleConfig: email: enabled: false slack: enabled: false logging: persistentVolumeClaimSize: 30Gi highAvailabilityEnabled: false logging: enabled: false prometheusServer: customAlerts: [] persistentVolumeClaimSize: 16Gi retentionSize: 15GB retentionTime: 15d watchDogAlertEnabled: false - name: metallb values: {} kind: BaremetalClusterProviderSpec loadBalancerHost: 172.16.168.3 publicKeys: - name: managed-cluster-key region: region-one release: mke-5-16-0-3-3-6
Create the
Machineobjects linked to eachbmhnode. For details, see Machine resource.managed-ns_Machine_cz7700-managed-cluster-control-noefi-.yamlapiVersion: cluster.k8s.io/v1alpha1 kind: Machine metadata: generateName: cz7700-managed-cluster-control-noefi- labels: cluster.sigs.k8s.io/cluster-name: managed-cluster cluster.sigs.k8s.io/control-plane: controlplane hostlabel.bm.kaas.mirantis.com/controlplane: controlplane kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one namespace: managed-ns spec: providerSpec: value: apiVersion: baremetal.k8s.io/v1alpha1 hostSelector: matchLabels: kaas.mirantis.com/baremetalhost-id: cz7700 kind: BareMetalMachineProviderSpec l2TemplateSelector: label: bm-1490-template-controls-netplan-cz7700-pxebond publicKeys: - name: managed-cluster-key
managed-ns_Machine_cz7741-managed-cluster-control-noefi-.yamlapiVersion: cluster.k8s.io/v1alpha1 kind: Machine metadata: generateName: cz7741-managed-cluster-control-noefi- labels: cluster.sigs.k8s.io/cluster-name: managed-cluster cluster.sigs.k8s.io/control-plane: controlplane hostlabel.bm.kaas.mirantis.com/controlplane: controlplane kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one namespace: managed-ns spec: providerSpec: value: apiVersion: baremetal.k8s.io/v1alpha1 bareMetalHostProfile: name: bmhp-cluster-default namespace: managed-ns hostSelector: matchLabels: kaas.mirantis.com/baremetalhost-id: cz7741 kind: BareMetalMachineProviderSpec l2TemplateSelector: label: bm-1490-template-controls-netplan publicKeys: - name: managed-cluster-key
managed-ns_Machine_cz7743-managed-cluster-control-noefi-.yamlapiVersion: cluster.k8s.io/v1alpha1 kind: Machine metadata: generateName: cz7743-managed-cluster-control-noefi- labels: cluster.sigs.k8s.io/cluster-name: managed-cluster cluster.sigs.k8s.io/control-plane: controlplane hostlabel.bm.kaas.mirantis.com/controlplane: controlplane kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one namespace: managed-ns spec: providerSpec: value: apiVersion: baremetal.k8s.io/v1alpha1 bareMetalHostProfile: name: bmhp-cluster-default namespace: managed-ns hostSelector: matchLabels: kaas.mirantis.com/baremetalhost-id: cz7743 kind: BareMetalMachineProviderSpec l2TemplateSelector: label: bm-1490-template-controls-netplan publicKeys: - name: managed-cluster-key
managed-ns_Machine_cz812-managed-cluster-storage-worker-noefi-.yamlapiVersion: cluster.k8s.io/v1alpha1 kind: Machine metadata: generateName: cz812-managed-cluster-storage-worker-noefi- labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/storage: storage hostlabel.bm.kaas.mirantis.com/worker: worker kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one namespace: managed-ns spec: providerSpec: value: apiVersion: baremetal.k8s.io/v1alpha1 bareMetalHostProfile: name: worker-storage1 namespace: managed-ns hostSelector: matchLabels: kaas.mirantis.com/baremetalhost-id: cz812 kind: BareMetalMachineProviderSpec l2TemplateSelector: label: bm-1490-template-workers-netplan publicKeys: - name: managed-cluster-key
managed-ns_Machine_cz813-managed-cluster-storage-worker-noefi-.yamlapiVersion: cluster.k8s.io/v1alpha1 kind: Machine metadata: generateName: cz813-managed-cluster-storage-worker-noefi- labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/storage: storage hostlabel.bm.kaas.mirantis.com/worker: worker kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one namespace: managed-ns spec: providerSpec: value: apiVersion: baremetal.k8s.io/v1alpha1 bareMetalHostProfile: name: worker-storage1 namespace: managed-ns hostSelector: matchLabels: kaas.mirantis.com/baremetalhost-id: cz813 kind: BareMetalMachineProviderSpec l2TemplateSelector: label: bm-1490-template-workers-netplan publicKeys: - name: managed-cluster-key
managed-ns_Machine_cz814-managed-cluster-storage-worker-noefi-.yamlapiVersion: cluster.k8s.io/v1alpha1 kind: Machine metadata: generateName: cz814-managed-cluster-storage-worker-noefi- labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/storage: storage hostlabel.bm.kaas.mirantis.com/worker: worker kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one namespace: managed-ns spec: providerSpec: value: apiVersion: baremetal.k8s.io/v1alpha1 bareMetalHostProfile: name: worker-storage1 namespace: managed-ns hostSelector: matchLabels: kaas.mirantis.com/baremetalhost-id: cz814 kind: BareMetalMachineProviderSpec l2TemplateSelector: label: bm-1490-template-workers-netplan publicKeys: - name: managed-cluster-key
managed-ns_Machine_cz815-managed-cluster-worker-noefi-.yamlapiVersion: cluster.k8s.io/v1alpha1 kind: Machine metadata: generateName: cz815-managed-cluster-worker-noefi- labels: cluster.sigs.k8s.io/cluster-name: managed-cluster hostlabel.bm.kaas.mirantis.com/worker: worker kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one si-role/node-for-delete: 'true' namespace: managed-ns spec: providerSpec: value: apiVersion: baremetal.k8s.io/v1alpha1 bareMetalHostProfile: name: worker-storage1 namespace: managed-ns hostSelector: matchLabels: kaas.mirantis.com/baremetalhost-id: cz815 kind: BareMetalMachineProviderSpec l2TemplateSelector: label: bm-1490-template-workers-netplan publicKeys: - name: managed-cluster-key
Verify that the
bmhnodes are in theprovisioningstate:KUBECONFIG=kubectl kubectl -n managed-ns get bmh -o wide
Example of system response:
NAME STATUS STATE CONSUMER BMC BOOTMODE ONLINE ERROR REGION cz7700-managed-cluster-control-noefi OK provisioning cz7700-managed-cluster-control-noefi-8bkqw 192.168.1.12 legacy true region-one cz7741-managed-cluster-control-noefi OK provisioning cz7741-managed-cluster-control-noefi-42tp2 192.168.1.76 legacy true region-one cz7743-managed-cluster-control-noefi OK provisioning cz7743-managed-cluster-control-noefi-8cwpw 192.168.1.78 legacy true region-one ...
Wait until all
bmhnodes are in theprovisionedstate.Verify that the
lcmmachinephase has started:KUBECONFIG=kubeconfig kubectl -n managed-ns get lcmmachines -o wide
Example of system response:
NAME CLUSTERNAME TYPE STATE INTERNALIP HOSTNAME AGENTVERSION cz7700-managed-cluster-control-noefi-8bkqw managed-cluster control Deploy 172.16.170.153 kaas-node-803721b4-227c-4675-acc5-15ff9d3cfde2 v0.2.0-349-g4870b7f5 cz7741-managed-cluster-control-noefi-42tp2 managed-cluster control Prepare 172.16.170.152 kaas-node-6b8f0d51-4c5e-43c5-ac53-a95988b1a526 v0.2.0-349-g4870b7f5 cz7743-managed-cluster-control-noefi-8cwpw managed-cluster control Prepare 172.16.170.151 kaas-node-e9b7447d-5010-439b-8c95-3598518f8e0a v0.2.0-349-g4870b7f5 ...
Verify that the
lcmmachinephase is complete and the Kubernetes cluster is created:KUBECONFIG=kubeconfig kubectl -n managed-ns get lcmmachines -o wide
Example of system response:
NAME CLUSTERNAME TYPE STATE INTERNALIP HOSTNAME AGENTVERSION cz7700-managed-cluster-control-noefi-8bkqw managed-cluster control Ready 172.16.170.153 kaas-node-803721b4-227c-4675-acc5-15ff9d3cfde2 v0.2.0-349-g4870b7f5 cz7741-managed-cluster-control-noefi-42tp2 managed-cluster control Ready 172.16.170.152 kaas-node-6b8f0d51-4c5e-43c5-ac53-a95988b1a526 v0.2.0-349-g4870b7f5 cz7743-managed-cluster-control-noefi-8cwpw managed-cluster control Ready 172.16.170.151 kaas-node-e9b7447d-5010-439b-8c95-3598518f8e0a v0.2.0-349-g4870b7f5 ...
Create the
KaaSCephClusterobject:managed-ns_KaaSCephCluster_ceph-cluster-managed-cluster.yamlapiVersion: kaas.mirantis.com/v1alpha1 kind: KaaSCephCluster metadata: name: ceph-cluster-managed-cluster namespace: managed-ns spec: cephClusterSpec: nodes: # Add the exact ``nodes`` names. # Obtain the name from "get bmh -o wide" ``consumer`` field. cz812-managed-cluster-storage-worker-noefi-58spl: roles: - mgr - mon # All disk configuration must be reflected in ``baremetalhostprofile`` storageDevices: - config: deviceClass: ssd fullPath: /dev/disk/by-id/scsi-1ATA_WDC_WDS100T2B0A-00SM50_200231434939 cz813-managed-cluster-storage-worker-noefi-lr4k4: roles: - mgr - mon storageDevices: - config: deviceClass: ssd fullPath: /dev/disk/by-id/scsi-1ATA_WDC_WDS100T2B0A-00SM50_200231440912 cz814-managed-cluster-storage-worker-noefi-z2m67: roles: - mgr - mon storageDevices: - config: deviceClass: ssd fullPath: /dev/disk/by-id/scsi-1ATA_WDC_WDS100T2B0A-00SM50_200231443409 pools: - default: true deviceClass: ssd name: kubernetes replicated: size: 3 role: kubernetes k8sCluster: name: managed-cluster namespace: managed-ns
Note
The
storageDevices[].fullPathfield is available since Container Cloud 2.25.0 (Cluster releases 17.0.0 and 16.0.0). For the clusters running earlier product versions, define the/dev/disk/by-idsymlinks usingstorageDevices[].nameinstead.Obtain
kubeconfigof the newly created managed cluster:KUBECONFIG=kubeconfig kubectl -n managed-ns get secrets managed-cluster-kubeconfig -o jsonpath='{.data.admin\.conf}' | base64 -d | tee managed.kubeconfig
Verify the status of the Ceph cluster in your managed cluster:
KUBECONFIG=managed.kubeconfig kubectl -n rook-ceph exec -it $(KUBECONFIG=managed.kubeconfig kubectl -n rook-ceph get pod -l "app=rook-ceph-tools" -o jsonpath='{.items[0].metadata.name}') -- ceph -s
Example of system response:
cluster: id: e75c6abd-c5d5-4ae8-af17-4711354ff8ef health: HEALTH_OK services: mon: 3 daemons, quorum a,b,c (age 55m) mgr: a(active, since 55m) osd: 3 osds: 3 up (since 54m), 3 in (since 54m) data: pools: 1 pools, 32 pgs objects: 273 objects, 555 MiB usage: 4.0 GiB used, 1.6 TiB / 1.6 TiB avail pgs: 32 active+clean io: client: 51 KiB/s wr, 0 op/s rd, 4 op/s wr
Automate multiple subnet creation using SubnetPool¶
Unsupported since MCC 2.28.0 (17.3.0 and 16.3.0)
Warning
The SubnetPool object is unsupported since Container Cloud
2.28.0 (Cluster releases 17.3.0 and 16.3.0). For details, see
Deprecation Notes: SubnetPool resource management.
Operators of Mirantis Container Cloud for on-demand self-service Kubernetes deployments will want their users to create networks without extensive knowledge about network topology or IP addresses. For that purpose, the Operator can prepare L2 network templates in advance for users to assign these templates to machines in their clusters.
The Operator can ensure that the users’ clusters have separate
IP address spaces using the SubnetPool resource.
SubnetPool allows for automatic creation of Subnet objects
that will consume blocks from the parent SubnetPool CIDR IP address
range. The SubnetPool blockSize setting defines the IP address
block size to allocate to each child Subnet. SubnetPool has a global
scope, so any SubnetPool can be used to create the Subnet objects
for any namespace and for any cluster.
You can use the SubnetPool resource in the L2Template resources to
automatically allocate IP addresses from an appropriate IP range that
corresponds to a specific cluster, or create a Subnet resource
if it does not exist yet. This way, every cluster will use subnets
that do not overlap with other clusters.
To automate multiple subnet creation using SubnetPool:
Log in to a local machine where your management cluster
kubeconfigis located and wherekubectlis installed.Note
The management cluster
kubeconfigis created during the last stage of the management cluster bootstrap.Create the
subnetpool.yamlfile with a number of subnet pools:Note
You can define either or both subnets and subnet pools, depending on the use case. A single L2 template can use either or both subnets and subnet pools.
kubectl --kubeconfig <pathToManagementClusterKubeconfig> apply -f <SubnetFileName.yaml>
Note
In the command above and in the steps below, substitute the parameters enclosed in angle brackets with the corresponding values.
Example of a
subnetpool.yamlfile:apiVersion: ipam.mirantis.com/v1alpha1 kind: SubnetPool metadata: name: kaas-mgmt namespace: default labels: kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one spec: cidr: 10.10.0.0/16 blockSize: /25 nameservers: - 172.18.176.6 gatewayPolicy: first
For the specification fields description of the
SubnetPoolobject, see SubnetPool spec.Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.Verify that the subnet pool is successfully created:
kubectl get subnetpool kaas-mgmt -oyaml
In the system output, verify the
statusfields of thesubnetpool.yamlfile. For the status fields description of theSunbetPoolobject, see SubnetPool status.Proceed to creating an L2 template for one or multiple managed clusters as described in Create L2 templates. In this procedure, select the exemplary L2 template for multiple subnets.
Caution
Using the
l3Layoutsection, define all subnets that are used in thenpTemplatesection. Defining only part of subnets is not allowed.If
labelSelectoris used inl3Layout, use any custom label name that differs from system names. This allows for easier cluster scaling in case of adding new subnets as described in Expand IP addresses capacity in an existing cluster.Mirantis recommends using a unique label prefix such as
user-defined/.
Add a machine¶
This section describes how to add a machine to a managed MOSK cluster using CLI for advanced configuration.
Create a machine using CLI¶
This section describes adding machines to a new MOSK cluster using Mirantis Container Cloud CLI.
If you need to add more machines to an existing MOSK cluster, see Add a controller node and Add a compute node.
To add machine to the MOSK cluster:
Log in to the host where your management cluster
kubeconfigis located and where kubectl is installed.Create a new text file
mosk-cluster-machines.yamland create the YAML definitons of theMachineresources. Use this as an example, and see the descriptions of the fields below:apiVersion: cluster.k8s.io/v1alpha1 kind: Machine metadata: name: mosk-node-role-name namespace: mosk-project labels: kaas.mirantis.com/provider: baremetal kaas.mirantis.com/region: region-one cluster.sigs.k8s.io/cluster-name: mosk-cluster spec: providerSpec: value: apiVersion: baremetal.k8s.io/v1alpha1 kind: BareMetalMachineProviderSpec bareMetalHostProfile: name: mosk-k8s-mgr namespace: mosk-project l2TemplateSelector: name: mosk-k8s-mgr hostSelector: {} l2TemplateMappingOverride: []
Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.Add the top level fields:
apiVersionAPI version of the object that is
cluster.k8s.io/v1alpha1.
kindObject type that is
Machine.
metadataThis section will contain the metadata of the object.
specThis section will contain the configuration of the object.
Add mandatory fields to the
metadatasection of theMachineobject definition.nameThe name of the
Machineobject.
namespaceThe name of the Project where the
Machinewill be created.
labelsThis section contains additional metadata of the machine. Set the following mandatory labels for the
Machineobject.kaas.mirantis.com/providerSet to
"baremetal".
kaas.mirantis.com/regionRegion name that matches the region name in the
Clusterobject.
cluster.sigs.k8s.io/cluster-nameThe name of the cluster to add the machine to.
Note
The
kaas.mirantis.com/regionlabel is removed from all MOSK objects in 24.1. Therefore, do not add the label starting with this release. On existing clusters updated to this release, or if added manually, MOSK ignores this label.
Configure the mandatory parameters of the
Machineobject inspecfield. AddproviderSpecfield that contains parameters for deployment on bare metal in a form of Kubernetes subresource.In the
providerSpecsection, add the following mandatory configuration parameters:apiVersionAPI version of the subresource that is
baremetal.k8s.io/v1alpha1.
kindObject type that is
BareMetalMachineProviderSpec.
bareMetalHostProfileReference to a configuration profile of a bare metal host. It helps to pick bare metal host with suitable configuration for the machine. This section includes two parameters:
nameName of a bare metal host profile
namespaceProject in which the bare metal host profile is created.
l2TemplateSelectorIf specified, contains the
name(first priority) orlabelof the L2 template that will be applied during a machine creation. Note that changing this field afterMachineobject is created will not affect the host network configuration of the machine.You should assign one of the templates you defined in Create L2 templates to the machine. If there is no suitable templates, you should create one per Create L2 templates.
hostSelectorThis parameter defines matching criteria for picking a bare metal host for the machine by label.
Any custom label that is assigned to one or more bare metal hosts using API can be used as a host selector. If the bare metal host objects with the specified label are missing, the
Machineobject will not be deployed until at least one bare metal host with the specified label is available.See Deploy a machine to a specific bare metal host for details.
l2TemplateIfMappingOverrideThis parameter contains a list of names of network interfaces of the host. It allows to override the default naming and ordering of network interfaces defined in L2 template referenced by the
l2TemplateSelector. This ordering informs the L2 templates how to generate the host network configuration.See Override network interfaces naming and order for details.
Depending on the role of the machine in the MOSK cluster, add labels to the
nodeLabelsfield.This field contains the list of node labels to be attached to a node for the user to run certain components on separate cluster nodes. The list of allowed node labels is located in the
Clusterobject statusproviderStatus.releaseRef.current.allowedNodeLabelsfield.If the
valuefield is not defined inallowedNodeLabels, a label can have any value. For example:allowedNodeLabels: - displayName: Stacklight key: stacklight
Before or after a machine deployment, add the required label from the allowed node labels list with the corresponding value to
spec.providerSpec.value.nodeLabelsinmachine.yaml. For example:nodeLabels: - key: stacklight value: enabled
Adding of a node label that is not available in the list of allowed node labels is restricted.
If you are NOT deploying MOSK with the compact control plane, add 3 dedicated Kubernetes manager nodes.
Add 3
Machineobjects for Kubernetes manager nodes using the following label:metadata: labels: cluster.sigs.k8s.io/control-plane: "true"
Note
The value of the label might be any non-empty string. On a worker node, this label must be omitted entirely.
Add 3
Machineobjects for MOSK controller nodes using the following labels:spec: providerSpec: value: nodeLabels: openstack-control-plane: enabled openstack-gateway: enabled
If you are deploying MOSK with compact control plane, add
Machineobjects for 3 combined control plane nodes using the following labels and parameters to thenodeLabelsfield:metadata: labels: cluster.sigs.k8s.io/control-plane: true spec: providerSpec: value: nodeLabels: openstack-control-plane: enabled openstack-gateway: enabled openvswitch: enabled
Add
Machineobjects for as many compute nodes as you want to install using the following labels:spec: providerSpec: value: nodeLabels: openstack-compute-node: enabled openvswitch: enabled
Save the text file and repeat the process to create configuration for all machines in your MOSK cluster.
Create machines in the cluster using command:
kubectl create -f mosk-cluster-machines.yaml
Proceed to Add a Ceph cluster.
See also
Create a machine using web UI¶
After you add bare metal hosts and create a managed cluster as described in Create a MOSK cluster, proceed with associating Kubernetes machines of your cluster with the previously added bare metal hosts using the Container Cloud web UI.
To add a Kubernetes machine to a MOSK cluster:
Log in to the Container Cloud web UI with the
m:kaas:namespace@operatororm:kaas:namespace@writerpermissions.Switch to the required project using the Switch Project action icon located on top of the main left-side navigation panel.
In the Clusters tab, click the required cluster name. The cluster page with the Machines list opens.
Click Create Machine button.
Fill out the Create New Machine form as required:
- Name
New machine name. If empty, a name is automatically generated in the
<clusterName>-<machineType>-<uniqueSuffix>format.
- Type
Machine type. Select Manager or Worker to create a Kubernetes manager or worker node.
Caution
The required minimum number of machines:
3 manager nodes for HA
3 worker storage nodes for a minimal Ceph cluster
- L2 Template
From the drop-down list, select the previously created L2 template, if any. For details, see Create L2 templates. Otherwise, leave the default selection to use the default L2 template of the cluster.
Note
Before Container Cloud 2.26.0 (Cluster releases 17.1.0 and 16.1.0), if you leave the default selection in the drop-down list, a preinstalled L2 template is used. Preinstalled templates are removed in the above-mentioned releases.
- Distribution
Operating system to provision the machine. From the drop-down list, select Ubuntu 22.04 Jammy as the machine distribution.
Warning
Do not use obsolete Ubuntu 20.04 distribution on greenfield deployments but only on existing clusters based on Ubuntu 20.04, which reaches end-of-life in April 2025. MOSK 24.3 release series is the last one to support Ubuntu 20.04 as the host operating system.
Update of management or MOSK clusters running Ubuntu 20.04 to the following major product release, where Ubuntu 22.04 is the only supported version, is not possible.
- Upgrade Index
Optional. A positive numeral value that defines the order of machine upgrade during a cluster update.
Note
You can change the upgrade order later on an existing cluster. For details, see Change the upgrade order of a machine.
Consider the following upgrade index specifics:
The first machine to upgrade is always one of the control plane machines with the lowest
upgradeIndex. Other control plane machines are upgraded one by one according to their upgrade indexes.If the
ClusterspecdedicatedControlPlanefield isfalse, worker machines are upgraded only after the upgrade of all control plane machines finishes. Otherwise, they are upgraded after the first control plane machine, concurrently with other control plane machines.If several machines have the same upgrade index, they have the same priority during upgrade.
If the value is not set, the machine is automatically assigned a value of the upgrade index.
- Host Configuration
Configuration settings of the bare metal host to be used for the machine:
- Host
From the drop-down list, select the previously created custom bare metal host to be used for the new machine.
- Host Profile
From the drop-down list, select the previously created custom bare metal host profile, if any. For details, see Create a custom bare metal host profile. Otherwise, leave the default selection.
- Labels
Add the required node labels for the worker machine to run certain components on a specific node. For example, for the StackLight nodes that run OpenSearch and require more resources than a standard node, add the StackLight label. The list of available node labels is obtained from
allowedNodeLabelsof your currentClusterrelease.If the
valuefield is not defined inallowedNodeLabels, from the drop-down list, select the required label and define an appropriate custom value for this label to be set to the node. For example, thenode-typelabel can have thestorage-ssdvalue to meet the service scheduling logic on a particular machine.Note
Due to the known issue 23002 fixed in Container Cloud 2.21.0 (Cluster releases 7.11.0 and 11.5.0), a custom value for a predefined node label cannot be set using the Container Cloud web UI. For a workaround, refer to the issue description.
Caution
If you deploy StackLight in the HA mode (recommended):
Add the StackLight label to minimum three worker nodes. Otherwise, StackLight will not be deployed until the required number of worker nodes is configured with the StackLight label.
Removal of the StackLight label from worker nodes along with removal of worker nodes with StackLight label can cause the StackLight components to become inaccessible. It is important to correctly maintain the worker nodes where the StackLight local volumes were provisioned. For details, see Delete a cluster machine.
To obtain the list of nodes where StackLight is deployed, refer to Container Cloud Release Notes: Upgrade managed clusters with StackLight deployed in HA mode.
If you move the StackLight label to a new worker machine on an existing cluster, manually deschedule all StackLight components from the old worker machine, which you remove the StackLight label from. For details, see Deschedule StackLight Pods from a worker machine.
Note
To add node labels after deploying a worker machine, navigate to the Machines page, click the More action icon in the last column of the required machine field, and select Configure machine.
- Count
Specify the number of machines to create. If you create a machine pool, specify the replicas count of the pool.
- Manager
Select Manager or Worker to create a Kubernetes manager or worker node.
Caution
The required minimum number of machines:
3 manager nodes for HA
3 worker storage nodes for a minimal Ceph cluster
- BareMetal Host Label
Assign the role to the new machine(s) to link the machine to a previously created bare metal host with the corresponding label. You can assign one role type per machine. The supported labels include:
- Manager
This node hosts the manager services of a managed cluster. For the reliability reasons, Container Cloud does not permit running end user workloads on the manager nodes or use them as storage nodes.
- Worker
The default role for any node in a managed cluster. Only the
kubeletservice is running on the machines of this type.
- Upgrade Index
Optional. A positive numeral value that defines the order of machine upgrade during a cluster update.
Note
You can change the upgrade order later on an existing cluster. For details, see Change the upgrade order of a machine.
Consider the following upgrade index specifics:
The first machine to upgrade is always one of the control plane machines with the lowest
upgradeIndex. Other control plane machines are upgraded one by one according to their upgrade indexes.If the
ClusterspecdedicatedControlPlanefield isfalse, worker machines are upgraded only after the upgrade of all control plane machines finishes. Otherwise, they are upgraded after the first control plane machine, concurrently with other control plane machines.If several machines have the same upgrade index, they have the same priority during upgrade.
If the value is not set, the machine is automatically assigned a value of the upgrade index.
- Distribution
Operating system to provision the machine. From the drop-down list, select the required Ubuntu distribution.
- L2 Template
From the drop-down list, select the previously created L2 template, if any. For details, see Create L2 templates. Otherwise, leave the default selection to use the default L2 template of the cluster.
Note
Before Container Cloud 2.26.0 (Cluster releases 17.1.0 and 16.1.0), if you leave the default selection in the drop-down list, a preinstalled L2 template is used. Preinstalled templates are removed in the above-mentioned releases.
- BM Host Profile
From the drop-down list, select the previously created custom bare metal host profile, if any. For details, see Create a custom bare metal host profile. Otherwise, leave the default selection.
- Node Labels
Add the required node labels for the worker machine to run certain components on a specific node. For example, for the StackLight nodes that run OpenSearch and require more resources than a standard node, add the StackLight label. The list of available node labels is obtained from
allowedNodeLabelsof your currentClusterrelease.If the
valuefield is not defined inallowedNodeLabels, from the drop-down list, select the required label and define an appropriate custom value for this label to be set to the node. For example, thenode-typelabel can have thestorage-ssdvalue to meet the service scheduling logic on a particular machine.Note
Due to the known issue 23002 fixed in Container Cloud 2.21.0 (Cluster releases 7.11.0 and 11.5.0), a custom value for a predefined node label cannot be set using the Container Cloud web UI. For a workaround, refer to the issue description.
Caution
If you deploy StackLight in the HA mode (recommended):
Add the StackLight label to minimum three worker nodes. Otherwise, StackLight will not be deployed until the required number of worker nodes is configured with the StackLight label.
Removal of the StackLight label from worker nodes along with removal of worker nodes with StackLight label can cause the StackLight components to become inaccessible. It is important to correctly maintain the worker nodes where the StackLight local volumes were provisioned. For details, see Delete a cluster machine.
To obtain the list of nodes where StackLight is deployed, refer to Container Cloud Release Notes: Upgrade managed clusters with StackLight deployed in HA mode.
If you move the StackLight label to a new worker machine on an existing cluster, manually deschedule all StackLight components from the old worker machine, which you remove the StackLight label from. For details, see Deschedule StackLight Pods from a worker machine.
Note
To add node labels after deploying a worker machine, navigate to the Machines page, click the More action icon in the last column of the required machine field, and select Configure machine.
Click Create.
At this point, Container Cloud adds the new machine object to the specified cluster. And the Bare Metal Operator Controller creates the relation to bare metal host with the labels matching the roles.
Provisioning of the newly created machine starts when the machine object is created and includes the following stages:
Creation of partitions on the local disks as required by the operating system and the Container Cloud architecture.
Configuration of the network interfaces on the host as required by the operating system and the Container Cloud architecture.
Installation and configuration of the Container Cloud LCM Agent.
Repeat the steps above for the remaining machines.
Now, proceed to Add a Ceph cluster.
Assign L2 templates to machines¶
To install MOSK on bare metal with Container Cloud, you must create L2 templates for each node type in the MOSK cluster. Additionally, you may have to create separate templates for nodes of the same type when they have different configuration.
To assign specific L2 templates to machines in a cluster:
Select from the following options to assign the templates to the cluster:
Since MOSK 23.3, use the
cluster.sigs.k8s.io/cluster-namelabel in thelabelssection.Before MOSK 23.3, use the
clusterRefparameter in thespecsection.
Add a unique identifier label to every L2 template. Typically, that would be the name of the MOSK node role, for example
l2template-compute, orl2template-compute-5nics.Assign an L2 template to a machine. Set the
l2TemplateSelectorfield in the machine spec to the name of the label added in the previous step. IPAM Controller uses this field to use a specific L2 template for the corresponding machine.Alternatively, you may set the
l2TemplateSelectorfield to the name of the L2 template.
Consider the following examples of an L2 template assignment to a machine.
apiVersion: ipam.mirantis.com/v1alpha1
kind: L2Template
metadata:
name: example-node-netconfig
namespace: my-project
labels:
kaas.mirantis.com/provider: baremetal
kaas.mirantis.com/region: region-one
l2template-example-node-netconfig: "1"
cluster.sigs.k8s.io/cluster-name: my-cluster
...
Note
The kaas.mirantis.com/region label is removed from all MOSK
objects in 24.1. Therefore, do not add the label starting with this release.
On existing clusters updated to this release, or if added manually, MOSK
ignores this label.
Note
Before MOSK 23.3, an L2 template requires
clusterRef: <clusterName> in the spec section. Since MOSK 23.3,
this parameter is deprecated and automatically migrated to the
cluster.sigs.k8s.io/cluster-name: <clusterName> label.
apiVersion: cluster.k8s.io/v1alpha1
kind: Machine
metadata:
name: machine1
namespace: my-project
labels:
cluster.sigs.k8s.io/cluster-name: my-cluster
...
...
spec:
providerSpec:
value:
l2TemplateSelector:
label: l2template-example-node-netconfig
...
apiVersion: cluster.k8s.io/v1alpha1
kind: Machine
metadata:
name: machine1
namespace: my-project
labels:
cluster.sigs.k8s.io/cluster-name: my-cluster
...
...
spec:
providerSpec:
value:
l2TemplateSelector:
name: example-node-netconfig
...
Now, proceed to Deploy a machine to a specific bare metal host.
Deploy a machine to a specific bare metal host¶
A machine in a MOSK cluster requires dedicated bare metal
host for deployment. In the Mirantis Container Cloud management API, bare metal
hosts are represented by the BareMetalHost objects that are automatically
generated by the related BareMetalHostInventory objects.
Note
The BareMetalHostInventory resource is available since the update
of the management cluster to the Cluster release 16.4.0 (Container Cloud
2.29.0). Before this release, the BareMetalHost object is used.
Since the above mentioned release, BareMetalHost is only used for
internal purposes of the Container Cloud private API. All configuration
changes must be applied using the BareMetalHostInventory objects.
For any existing BareMetalHost object, a BareMetalHostInventory
object is created automatically during cluster update.
m:kaas@management-adminonly. This limitation is lifted once the management cluster is updated to the Cluster release 16.4.1 or later.
All BareMetalHostInventory objects must be labeled upon creation with a
label that allows identifying the host and assigning it to a machine.
The labels may be unique, or applied to a group of hosts, based on similarities in their capacity, capabilities and hardware configuration, on their location, suitable role, or a combination of thereof.
In some cases, you may need to deploy a machine to a specific bare metal host. This is especially useful when some of your bare metal hosts have different hardware configuration than the rest.
To deploy a machine to a specific bare metal host:
Log in to the host where your management cluster
kubeconfigis located and where kubectl is installed.Identify the bare metal host that you want to associate with the specific machine. For example, host
host-1.kubectl get baremetalhostinventory host-1 -o yaml
kubectl get baremetalhost host-1 -o yaml
Add a label that will uniquely identify this host, for example, by the name of the host and machine that you want to deploy on it.
kubectl edit baremetalhostinventory host-1
Note
For details about labels, see BareMetalHostInventory resource.
kubectl edit baremetalhost host-1
Note
For details about labels, see BareMetalHost resource.
Configuration example:
kind: BareMetalHostInventory metadata: name: host-1 namespace: myProjectName labels: kaas.mirantis.com/baremetalhost-id: host-1-worker-HW11-cad5 ...
kind: BareMetalHost metadata: name: host-1 namespace: myProjectName labels: kaas.mirantis.com/baremetalhost-id: host-1-worker-HW11-cad5 ...
Open the text file with the YAML definition of the
Machineobject, created in Create a machine using CLI.Add a host selector that matches the label you have added to the
BareMetalHostobject in the previous step. For example:kind: Machine metadata: name: worker-HW11-cad5 namespace: myProjectName spec: ... providerSpec: value: apiVersion: baremetal.k8s.io/v1alpha1 kind: BareMetalMachineProviderSpec ... hostSelector: matchLabels: kaas.mirantis.com/baremetalhost-id: host-1-worker-HW11-cad5 ...
Once created, this machine will be associated with the specified bare metal host, and you can return to Create a machine using CLI.
Caution
The required minimum number of machines:
3 manager nodes for HA
3 worker storage nodes for a minimal Ceph cluster
Override network interfaces naming and order¶
An L2 template contains the ifMapping field that allows you to
identify Ethernet interfaces for the template. The Machine object
API enables the Operator to override the mapping from the L2 template
by enforcing a specific order of names of the interfaces when applied
to the template.
The field l2TemplateIfMappingOverride in the spec of the Machine
object contains a list of interfaces names. The order of the interfaces
names in the list is important because the L2Template object will
be rendered with NICs ordered as per this list.
Note
Changes in the l2TemplateIfMappingOverride field will apply
only once when the Machine and corresponding IpamHost objects
are created. Further changes to l2TemplateIfMappingOverride
will not reset the interfaces assignment and configuration.
Caution
The l2TemplateIfMappingOverride field must contain the names
of all interfaces of the bare metal host.
The following example illustrates how to include the override field to the
Machine object. In this example, we configure the interface eno1,
which is the second on-board interface of the server, to precede the first
on-board interface eno0.
apiVersion: cluster.k8s.io/v1alpha1
kind: Machine
metadata:
finalizers:
- foregroundDeletion
- machine.cluster.sigs.k8s.io
labels:
cluster.sigs.k8s.io/cluster-name: kaas-mgmt
cluster.sigs.k8s.io/control-plane: "true"
kaas.mirantis.com/provider: baremetal
kaas.mirantis.com/region: region-one
spec:
providerSpec:
value:
apiVersion: baremetal.k8s.io/v1alpha1
hostSelector:
matchLabels:
kaas.mirantis.com/baremetalhost-id: hw-master-0
image: {}
kind: BareMetalMachineProviderSpec
l2TemplateIfMappingOverride:
- eno1
- eno0
- enp0s1
- enp0s2
Note
The kaas.mirantis.com/region label is removed from all MOSK
objects in 24.1. Therefore, do not add the label starting with this release.
On existing clusters updated to this release, or if added manually, MOSK
ignores this label.
As a result of the configuration above, when used with the example
L2 template for bonds and bridges described in Create L2 templates,
the enp0s1 and enp0s2 interfaces will be in predictable
ordered state. This state will be used to create subinterfaces for
Kubernetes networks (k8s-pods) and for Kubernetes external
network (k8s-ext).
Also, you can use the non-case-sensitive list of NIC MAC addresses instead of the list of NIC names. For example:
apiVersion: cluster.k8s.io/v1alpha1
kind: Machine
...
spec:
providerSpec:
value:
...
kind: BareMetalMachineProviderSpec
l2TemplateIfMappingOverride:
- b4:96:91:6f:2e:10
- b4:96:91:6f:2e:11
- b5:a6:c1:6f:ee:02
- b5:a6:c1:6f:ee:02
Manually allocate IP addresses for bare metal hosts¶
Available since MCC 2.26.0 (16.1.0 and 17.1.0)
You can force the DHCP server to assign a particular IP address for a bare
metal host during PXE provisioning by adding the
host.dnsmasqs.metal3.io/address annotation with the desired IP address
value to the required bare metal host.
If you have a limited amount of free and unused IP addresses for a server provisioning, you can manually create bare metal hosts one by one and provision servers in small, manually managed batches.
For batching in small chunks, you can use the
host.dnsmasqs.metal3.io/address annotation to manually allocate IP
addresses along with the baremetalhost.metal3.io/detached annotation to
pause automatic host management by the bare metal Operator.
To pause bare metal hosts for a manual IP allocation during provisioning:
Set the
baremetalhost.metal3.io/detachedannotation for all bare metal hosts that pauses host management.Note
If the host provisioning has already started or completed, addition of this annotation deletes the information about the host from Ironic without triggering deprovisioning. The bare metal Operator recreates the host in Ironic once you remove the annotation. For details, see Metal3 documentation.
Add the
host.dnsmasqs.metal3.io/addressannotation with corresponding IP address values to a batch of bare metal hosts.Remove the
baremetalhost.metal3.io/detachedannotation from the batch used in the previous step.Repeat the steps 2 and 3 until all hosts are provisioned.
Add a Ceph cluster¶
After you add machines to your new bare metal managed cluster as described in Add a machine, create a Ceph cluster on top of this managed cluster.
For an advanced configuration through the KaaSCephCluster CR, see
Ceph advanced configuration. To configure Ceph Controller through Kubernetes
templates to manage Ceph node resources, see Enable Ceph tolerations and resources management.
The procedure below enables you to create a Ceph cluster with minimum three Ceph nodes that provides persistent volumes to the Kubernetes workloads in the managed cluster.
Create a Ceph cluster using the CLI¶
Verify that the overall status of the managed cluster is ready with all conditions in the
Readystate:kubectl -n <managedClusterProject> get cluster <clusterName> -o yaml
Substitute
<managedClusterProject>and<clusterName>with the corresponding managed cluster namespace and name.Example of system response:
status: providerStatus: ready: true conditions: - message: Helm charts are successfully installed(upgraded). ready: true type: Helm - message: Kubernetes objects are fully up. ready: true type: Kubernetes - message: All requested nodes are ready. ready: true type: Nodes - message: Maintenance state of the cluster is false ready: true type: Maintenance - message: TLS configuration settings are applied ready: true type: TLS - message: Kubelet is Ready on all nodes belonging to the cluster ready: true type: Kubelet - message: Swarm is Ready on all nodes belonging to the cluster ready: true type: Swarm - message: All provider instances of the cluster are Ready ready: true type: ProviderInstance - message: LCM agents have the latest version ready: true type: LCMAgent - message: StackLight is fully up. ready: true type: StackLight - message: OIDC configuration has been applied. ready: true type: OIDC - message: Load balancer 10.100.91.150 for kubernetes API has status HEALTHY ready: true type: LoadBalancer
Create a YAML file with the Ceph cluster specification:
apiVersion: kaas.mirantis.com/v1alpha1 kind: KaaSCephCluster metadata: name: <cephClusterName> namespace: <managedClusterProject> spec: k8sCluster: name: <clusterName> namespace: <managedClusterProject>
Substitute
<cephClusterName>with the required name of the Ceph cluster. This name will be used in the Ceph LCM operations.Select from the following options:
Add explicit network configuration of the Ceph cluster using the
networksection:spec: cephClusterSpec: network: publicNet: <publicNet> clusterNet: <clusterNet>
Substitute the following values:
<publicNet>is a CIDR definition or comma-separated list of CIDR definitions (if the managed cluster uses multiple networks) of public network for the Ceph data. The values should match the corresponding values of the clusterSubnetobject.<clusterNet>is a CIDR definition or comma-separated list of CIDR definitions (if the managed cluster uses multiple networks) of replication network for the Ceph data. The values should match the corresponding values of the clusterSubnetobject.
Configure
Subnetobjects for the Storage access network by settingipam/SVC-ceph-public: "1"andipam/SVC-ceph-cluster: "1"labels to the correspondingSubnetobjects. For more details, refer to Create subnets for a managed cluster using CLI, Step 5.
Configure Ceph Manager and Ceph Monitor roles to select nodes that must place Ceph Monitor and Ceph Manager daemons:
Obtain the names of machines to place Ceph Monitor and Ceph Manager daemons at:
kubectl -n <managedClusterProject> get machine
Add the
nodessection withmonandmgrroles defined:spec: cephClusterSpec: nodes: <mgr-node-1>: roles: - <role-1> - <role-2> ... <mgr-node-2>: roles: - <role-1> - <role-2> ...
Substitute
<mgr-node-X>with the correspondingMachineobject names and<role-X>with the corresponding roles of daemon placement, for example,monormgr.For other optional node parameters, see Ceph advanced configuration.
Configure Ceph OSD daemons for Ceph cluster data storage:
Note
This step involves the deployment of Ceph Monitor and Ceph Manager daemons on nodes that are different from the ones hosting Ceph cluster OSDs. However, you can also colocate Ceph OSDs, Ceph Monitor, and Ceph Manager daemons on the same nodes by configuring the
rolesandstorageDevicessections accordingly. This kind of configuration flexibility is particularly useful in scenarios such as hyper-converged clusters.Warning
The minimal production cluster requires at least three nodes for Ceph Monitor daemons and three nodes for Ceph OSDs.
Obtain the names of machines with disks intended for storing Ceph data:
kubectl -n <managedClusterProject> get machine
For each machine, use
status.providerStatus.hardware.storageto obtain information about node disks:kubectl -n <managedClusterProject> get machine <machineName> -o yaml
Example of system response with machine hardware details
status: providerStatus: hardware: storage: - byID: /dev/disk/by-id/wwn-0x05ad99618d66a21f byIDs: - /dev/disk/by-id/scsi-0QEMU_QEMU_HARDDISK_05ad99618d66a21f - /dev/disk/by-id/scsi-305ad99618d66a21f - /dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_05ad99618d66a21f - /dev/disk/by-id/wwn-0x05ad99618d66a21f byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:0 byPaths: - /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:0 name: /dev/sda serialNumber: 05ad99618d66a21f size: 61 type: hdd - byID: /dev/disk/by-id/wwn-0x26d546263bd312b8 byIDs: - /dev/disk/by-id/scsi-0QEMU_QEMU_HARDDISK_26d546263bd312b8 - /dev/disk/by-id/scsi-326d546263bd312b8 - /dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_26d546263bd312b8 - /dev/disk/by-id/wwn-0x26d546263bd312b8 byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:2 byPaths: - /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:2 name: /dev/sdb serialNumber: 26d546263bd312b8 size: 32 type: hdd - byID: /dev/disk/by-id/wwn-0x2e52abb48862dbdc byIDs: - /dev/disk/by-id/lvm-pv-uuid-MncrcO-6cel-0QsB-IKaY-e8UK-6gDy-k2hOtf - /dev/disk/by-id/scsi-0QEMU_QEMU_HARDDISK_2e52abb48862dbdc - /dev/disk/by-id/scsi-32e52abb48862dbdc - /dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_2e52abb48862dbdc - /dev/disk/by-id/wwn-0x2e52abb48862dbdc byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:1 byPaths: - /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:1 name: /dev/sdc serialNumber: 2e52abb48862dbdc size: 61 type: hdd
Select
by-idsymlinks on the disks to be used in the Ceph cluster. The symlinks must meet the following requirements:A
by-idsymlink must containstatus.providerStatus.hardware.storage.serialNumberA
by-idsymlink must not containwwn
For the example above, to use the
sdcdisk to store Ceph data on it, select the/dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_2e52abb48862dbdcsymlink. It is persistent and will not be affected by node reboot.Note
For details about storage device formats, see Mirantis Container Cloud Reference Architecture: Addressing storage devices.
Sepcify
by-idsymlinks:Specify the selected
by-idsymlinks in thespec.cephClusterSpec.nodes.storageDevices.fullPathfield along with thespec.cephClusterSpec.nodes.storageDevices.config.deviceClassfield:spec: cephClusterSpec: nodes: <storage-node-1>: storageDevices: - fullPath: <byIDSymlink-1> config: deviceClass: <deviceClass-1> - fullPath: <byIDSymlink-2> config: deviceClass: <deviceClass-1> - fullPath: <byIDSymlink-3> config: deviceClass: <deviceClass-2> ... <storage-node-2>: storageDevices: - fullPath: <byIDSymlink-4> config: deviceClass: <deviceClass-1> - fullPath: <byIDSymlink-5> config: deviceClass: <deviceClass-1> - fullPath: <byIDSymlink-6> config: deviceClass: <deviceClass-2> <storage-node-3>: storageDevices: - fullPath: <byIDSymlink-7> config: deviceClass: <deviceClass-1> - fullPath: <byIDSymlink-8> config: deviceClass: <deviceClass-1> - fullPath: <byIDSymlink-9> config: deviceClass: <deviceClass-2>
Specify the selected
by-idsymlinks in thespec.cephClusterSpec.nodes.storageDevices.namefield along with thespec.cephClusterSpec.nodes.storageDevices.config.deviceClassfield:spec: cephClusterSpec: nodes: <storage-node-1>: storageDevices: - name: <byIDSymlink-1> config: deviceClass: <deviceClass-1> - name: <byIDSymlink-2> config: deviceClass: <deviceClass-1> - name: <byIDSymlink-3> config: deviceClass: <deviceClass-2> ... <storage-node-2>: storageDevices: - name: <byIDSymlink-4> config: deviceClass: <deviceClass-1> - name: <byIDSymlink-5> config: deviceClass: <deviceClass-1> - name: <byIDSymlink-6> config: deviceClass: <deviceClass-2> <storage-node-3>: storageDevices: - name: <byIDSymlink-7> config: deviceClass: <deviceClass-1> - name: <byIDSymlink-8> config: deviceClass: <deviceClass-1> - name: <byIDSymlink-9> config: deviceClass: <deviceClass-2>
Substitute the following values:
<storage-node-X>with the correspondingMachineobject names<byIDSymlink-X>with theby-idsymlinks obtained fromstatus.providerStatus.hardware.storage.byIDs<deviceClass-X>with the disk types obtained fromstatus.providerStatus.hardware.storage.type
Configure the pools for Image, Block Storage, and Compute services:
Note
Ceph validates the specified pools. Therefore, do not omit any of the following pools.
spec: pools: - default: true deviceClass: hdd name: kubernetes replicated: size: 3 role: kubernetes - default: false deviceClass: hdd name: volumes replicated: size: 3 role: volumes - default: false deviceClass: hdd name: vms replicated: size: 3 role: vms - default: false deviceClass: hdd name: backup replicated: size: 3 role: backup - default: false deviceClass: hdd name: images replicated: size: 3 role: images
Each Ceph pool, depending on its role, has the default
targetSizeRatiovalue that defines the expected consumption of the total Ceph cluster capacity. The default ratio values for MOSK pools are as follows:20.0% for a Ceph pool with the role
volumes40.0% for a Ceph pool with the role
vms10.0% for a Ceph pool with the role
images10.0% for a Ceph pool with the role
backup
Optional. Configure Ceph Block Pools to use RBD. For the detailed configuration, refer to Pool parameters.
Example configuration:
spec: cephClusterSpec: pools: - name: kubernetes role: kubernetes deviceClass: hdd replicated: size: 3 targetSizeRatio: 10.0 default: true
Configure Ceph Object Storage to use OpenStack Swift Object Storage. For details, see RADOS Gateway parameters. Example configuration:
spec: cephClusterSpec: objectStorage: rgw: dataPool: deviceClass: hdd erasureCoded: codingChunks: 1 dataChunks: 2 failureDomain: host gateway: instances: 3 port: 80 securePort: 8443 metadataPool: deviceClass: hdd failureDomain: host replicated: size: 3 name: object-store preservePoolsOnDelete: false
Optional. Configure Ceph Shared Filesystem to use CephFS. For the detailed configuration, refer to Configure Ceph Shared File System (CephFS). Example configuration:
spec: cephClusterSpec: sharedFilesystem: cephFS: - name: cephfs-store dataPools: - name: cephfs-pool-1 deviceClass: hdd replicated: size: 3 failureDomain: host metadataPool: deviceClass: nvme replicated: size: 3 failureDomain: host metadataServer: activeCount: 1 activeStandby: false
When the Ceph cluster specification is complete, apply the built YAML file on the management cluster:
kubectl apply -f <kcc-template>.yaml
Substitue
<kcc-template>with the name of the file containing theKaaSCephClusterspecification.The resulting example of the
KaaSCephClustertemplateapiVersion: kaas.mirantis.com/v1alpha1 kind: KaaSCephCluster metadata: name: kaas-ceph namespace: mosk-namespace spec: k8sCluster: name: mosk-cluster namespace: mosk-namespace cephClusterSpec: network: publicNet: 10.10.0.0/24 clusterNet: 10.11.0.0/24 nodes: master-1: roles: - mon - mgr master-2: roles: - mon - mgr master-3: roles: - mon - mgr worker-1: storageDevices: - fullPath: /dev/disk/by-id/scsi-1ATA_WDC_WDS100T2B0A-00SM50_200231443409 config: deviceClass: ssd worker-2: storageDevices: - fullPath: /dev/disk/by-id/scsi-1ATA_WDC_WDS100T2B0A-00SM50_200231440912 config: deviceClass: ssd worker-3: storageDevices: - fullPath: /dev/disk/by-id/scsi-1ATA_WDC_WDS100T2B0A-00SM50_200231434939 config: deviceClass: ssd pools: - default: true deviceClass: hdd name: kubernetes replicated: size: 3 role: kubernetes - default: false deviceClass: hdd name: volumes replicated: size: 3 role: volumes - default: false deviceClass: hdd name: vms replicated: size: 3 role: vms - default: false deviceClass: hdd name: backup replicated: size: 3 role: backup - default: false deviceClass: hdd name: images replicated: size: 3 role: images objectStorage: rgw: dataPool: deviceClass: ssd erasureCoded: codingChunks: 1 dataChunks: 2 failureDomain: host gateway: instances: 3 port: 80 securePort: 8443 metadataPool: deviceClass: ssd failureDomain: host replicated: size: 3 name: object-store preservePoolsOnDelete: false
Wait for the
KaaSCephClusterstatus and forstatus.shortClusterInfo.stateto becomeReady:kubectl -n <managedClusterProject> get kcc -o yaml
Verify your Ceph cluster as described in Verify Ceph.
Once all pools are created, verify that an appropriate secret required for a successful deployment of the OpenStack services that rely on Ceph is created in the
openstack-ceph-sharednamespace:kubectl -n openstack-ceph-shared get secrets openstack-ceph-keys
Example of a positive system response:
NAME TYPE DATA AGE openstack-ceph-keys Opaque 7 36m
Create a Ceph cluster using the web UI¶
Warning
Mirantis highly recommends adding a Ceph cluster using the CLI instead of the web UI.
The web UI capabilities for adding a Ceph cluster are limited and lack flexibility in defining Ceph cluster specifications. For example, if an error occurs while adding a Ceph cluster using the web UI, usually you can address it only through the CLI.
The web UI functionality for managing Ceph cluster is going to be deprecated in one of the following releases.
Log in to the Container Cloud web UI with the
m:kaas:namespace@operatororm:kaas:namespace@writerpermissions.Switch to the required project using the Switch Project action icon located on top of the main left-side navigation panel.
In the Clusters tab, click the required cluster name. The Cluster page with the Machines and Ceph clusters lists opens.
In the Ceph Clusters block, click Create Cluster.
Configure the Ceph cluster in the Create New Ceph Cluster wizard that opens:
Create new Ceph cluster¶ Section
Parameter name
Description
General settings
Name
The Ceph cluster name.
Cluster Network
Replication network for Ceph OSDs. Must contain the CIDR definition and match the corresponding values of the cluster
L2Templateobject or the environment network values.Public Network
Public network for Ceph data. Must contain the CIDR definition and match the corresponding values of the cluster
L2Templateobject or the environment network values.Enable OSDs LCM
Select to enable LCM for Ceph OSDs.
Machines / Machine #1-3
Select machine
Select the name of the Kubernetes machine that will host the corresponding Ceph node in the Ceph cluster.
Manager, Monitor
Select the required Ceph services to install on the Ceph node.
Devices
Select the disk that Ceph will use.
Warning
Do not select the device for system services, for example,
sda.Warning
A Ceph cluster does not support removable devices that are hosts with hotplug functionality enabled. To use devices as Ceph OSD data devices, make them non-removable or disable the hotplug functionality in the BIOS settings for disks that are configured to be used as Ceph OSD data devices.
Enable Object Storage
Select to enable the single-instance RGW Object Storage.
To add more Ceph nodes to the new Ceph cluster, click + next to any Ceph Machine title in the Machines tab. Configure a Ceph node as required.
Warning
Do not add more than 3
Managerand/orMonitorservices to the Ceph cluster.After you add and configure all nodes in your Ceph cluster, click Create.
Open the
KaaSCephClusterCR for editing as described in Ceph advanced configuration.Verify that the following snippet is present in the
KaaSCephClusterconfiguration:network: clusterNet: 10.10.10.0/24 publicNet: 10.10.11.0/24
Configure the pools for Image, Block Storage, and Compute services.
Note
Ceph validates the specified pools. Therefore, do not omit any of the following pools.
spec: pools: - default: true deviceClass: hdd name: kubernetes replicated: size: 3 role: kubernetes - default: false deviceClass: hdd name: volumes replicated: size: 3 role: volumes - default: false deviceClass: hdd name: vms replicated: size: 3 role: vms - default: false deviceClass: hdd name: backup replicated: size: 3 role: backup - default: false deviceClass: hdd name: images replicated: size: 3 role: images
Each Ceph pool, depending on its role, has a default
targetSizeRatiovalue that defines the expected consumption of the total Ceph cluster capacity. The default ratio values for MOSK pools are as follows:20.0% for a Ceph pool with role
volumes40.0% for a Ceph pool with role
vms10.0% for a Ceph pool with role
images10.0% for a Ceph pool with role
backup
Once all pools are created, verify that an appropriate secret required for a successful deployment of the OpenStack services that rely on Ceph is created in the
openstack-ceph-sharednamespace:kubectl -n openstack-ceph-shared get secrets openstack-ceph-keys
Example of a positive system response:
NAME TYPE DATA AGE openstack-ceph-keys Opaque 7 36m
Verify your Ceph cluster as described in Verify Ceph.
Deploy OpenStack¶
This section instructs you on how to deploy OpenStack on top of Kubernetes as well as how to troubleshoot the deployment and access your OpenStack environment after deployment.
Deploy an OpenStack cluster¶
This section instructs you on how to deploy OpenStack on top of Kubernetes
using the OpenStack Controller (Rockoon) and
openstackdeployments.lcm.mirantis.com (OsDpl) CR.
To deploy an OpenStack cluster:
Verify that you have pre-configured the networking according to Networking.
Verify that the TLS certificates that will be required for the OpenStack cluster deployment have been pre-generated.
Note
The Transport Layer Security (TLS) protocol is mandatory on public endpoints.
Caution
To avoid certificates renewal with subsequent OpenStack updates during which additional services with new public endpoints may appear, we recommend using wildcard SSL certificates for public endpoints. For example,
*.it.just.works, whereit.just.worksis a cluster public domain.The sample code block below illustrates how to generate a self-signed certificate for the
it.just.worksdomain. The procedure presumes the cfssl and cfssljson tools are installed on the machine.mkdir cert && cd cert tee ca-config.json << EOF { "signing": { "default": { "expiry": "8760h" }, "profiles": { "kubernetes": { "usages": [ "signing", "key encipherment", "server auth", "client auth" ], "expiry": "8760h" } } } } EOF tee ca-csr.json << EOF { "CN": "kubernetes", "key": { "algo": "rsa", "size": 2048 }, "names":[{ "C": "<country>", "ST": "<state>", "L": "<city>", "O": "<organization>", "OU": "<organization unit>" }] } EOF cfssl gencert -initca ca-csr.json | cfssljson -bare ca tee server-csr.json << EOF { "CN": "*.it.just.works", "hosts": [ "*.it.just.works" ], "key": { "algo": "rsa", "size": 2048 }, "names": [ { "C": "US", "L": "CA", "ST": "San Francisco" }] } EOF cfssl gencert -ca=ca.pem -ca-key=ca-key.pem --config=ca-config.json -profile=kubernetes server-csr.json | cfssljson -bare server
Create the
openstackdeployment.yamlfile that will include the OpenStack cluster deployment configuration. For the configuration details, refer to OpenStack configuration and OpenStack Operator resources.Note
The resource of kind
OpenStackDeployment(OsDpl) is a custom resource defined by a resource of kindCustomResourceDefinition. The resource is validated with the help of the OpenAPI v3 schema.Configure the OsDpl resource depending on the needs of your deployment. For the configuration details, refer to OpenStack configuration.
Example of an OpenStackDeployment CR of minimum configuration¶apiVersion: lcm.mirantis.com/v1alpha1 kind: OpenStackDeployment metadata: name: openstack-cluster namespace: openstack spec: openstack_version: victoria preset: compute size: tiny internal_domain_name: cluster.local public_domain_name: it.just.works features: neutron: tunnel_interface: ens3 external_networks: - physnet: physnet1 interface: veth-phy bridge: br-ex network_types: - flat vlan_ranges: null mtu: null floating_network: enabled: False nova: live_migration_interface: ens3 images: backend: local
If required, enable DPDK, huge pages, and other supported Telco features as described in Advanced OpenStack configuration (optional).
To the
openstackdeploymentobject, add information about the TLS certificates:ssl:public_endpoints:ca_cert- CA certificate content (ca.pem)ssl:public_endpoints:api_cert- server certificate content (server.pem)ssl:public_endpoints:api_key- server private key (server-key.pem)
Verify that the Load Balancer network does not overlap your corporate or internal Kubernetes networks, for example, Calico IP pools. Also, verify that the pool of Load Balancer network is big enough to provide IP addresses for all Amphora VMs (
loadbalancers).If required, reconfigure the Octavia network settings using the following sample structure:
spec: services: load-balancer: octavia: values: octavia: settings: lbmgmt_cidr: "10.255.0.0/16" lbmgmt_subnet_start: "10.255.1.0" lbmgmt_subnet_end: "10.255.255.254"
If you are using the default backend to store OpenStack database backups, which is Ceph, you may want to increase the default size of the allocated storage since there is no automatic way to resize the backup volume once the cloud is deployed.
For the default sizes and configuration details, refer to Size of a backup storage.
Trigger the OpenStack deployment:
kubectl apply -f openstackdeployment.yaml
Monitor the status of your OpenStack deployment:
kubectl -n openstack get pods kubectl -n openstack describe osdpl osh-dev
Assess the current status of the OpenStack deployment using the
statussection output in the OsDpl resource:Get the OsDpl YAML file:
kubectl -n openstack get osdpl osh-dev -o yaml
Analyze the
statusoutput using the detailed description in OpenStack configuration.
Verify that the OpenStack cluster has been deployed:
clinet_pod_name=$(kubectl -n openstack get pods -l application=keystone,component=client | grep keystone-client | head -1 | awk '{print $1}') kubectl -n openstack exec -it $clinet_pod_name -- openstack service list
Example of a positive system response:
+----------------------------------+---------------+----------------+ | ID | Name | Type | +----------------------------------+---------------+----------------+ | 159f5c7e59784179b589f933bf9fc6b0 | cinderv3 | volumev3 | | 6ad762f04eb64a31a9567c1c3e5a53b4 | keystone | identity | | 7e265e0f37e34971959ce2dd9eafb5dc | heat | orchestration | | 8bc263babe9944cdb51e3b5981a0096b | nova | compute | | 9571a49d1fdd4a9f9e33972751125f3f | placement | placement | | a3f9b25b7447436b85158946ca1c15e2 | neutron | network | | af20129d67a14cadbe8d33ebe4b147a8 | heat-cfn | cloudformation | | b00b5ad18c324ac9b1c83d7eb58c76f5 | radosgw-swift | object-store | | b28217da1116498fa70e5b8d1b1457e5 | cinderv2 | volumev2 | | e601c0749ce5425c8efb789278656dd4 | glance | image | +----------------------------------+---------------+----------------+
Register a record on the customer DNS:
Caution
The DNS component is mandatory to access OpenStack public endpoints.
Obtain the full list of endpoints:
kubectl -n openstack get ingress -ocustom-columns=NAME:.metadata.name,HOSTS:spec.rules[*].host | awk '/namespace-fqdn/ {print $2}'
Example of system response:
barbican.<spec:public_domain_name> cinder.<spec:public_domain_name> cloudformation.<spec:public_domain_name> designate.<spec:public_domain_name> glance.<spec:public_domain_name> heat.<spec:public_domain_name> horizon.<spec:public_domain_name> keystone.<spec:public_domain_name> metadata.<spec:public_domain_name> neutron.<spec:public_domain_name> nova.<spec:public_domain_name> novncproxy.<spec:public_domain_name> octavia.<spec:public_domain_name> placement.<spec:public_domain_name>
Obtain the public endpoint IP:
kubectl -n openstack get services ingress
In the system response, capture
EXTERNAL-IP.Example of system response:
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE ingress LoadBalancer 10.96.32.97 10.172.1.101 80:34234/TCP,443:34927/TCP,10246:33658/TCP 4h56m
Ask the customer to create records for public endpoints, obtained earlier in this procedure, to
EXTERNAL-IPfrom the Ingress service.
See also
Advanced OpenStack configuration (optional)¶
This section includes configuration information for available advanced Mirantis OpenStack for Kubernetes features that include DPDK with the Neutron OVS backend, huge pages, CPU pinning, and other Enhanced Platform Awareness (EPA) capabilities.
Enable LVM ephemeral storage¶
TechPreview
Note
Consider this section as part of Deploy an OpenStack cluster.
This section instructs you on how to configure LVM as backend for the VM disks and ephemeral storage.
You can use flexible size units throughout bare metal host profiles.
For example, you can now use either sizeGiB: 0.1 or size: 100Mi
when specifying a device size.
Mirantis recommends using only one parameter name type and units throughout
the configuration files. If both sizeGiB and size are used,
sizeGiB is ignored during deployment and the suffix is adjusted
accordingly. For example, 1.5Gi will be serialized as 1536Mi. The size
without units is counted in bytes. For example, size: 120 means 120 bytes.
Warning
All data will be wiped during cluster deployment on devices
defined directly or indirectly in the fileSystems list of
BareMetalHostProfile. For example:
A raw device partition with a file system on it
A device partition in a volume group with a logical volume that has a file system on it
An mdadm RAID device with a file system on it
An LVM RAID device with a file system on it
The wipe field is always considered true for these devices.
The false value is ignored.
Therefore, to prevent data loss, move the necessary data from these file systems to another server beforehand, if required.
Warning
Usage of more than one nonvolatile memory express (NVMe) drive per node may cause update issues and is therefore not supported.
To enable LVM ephemeral storage:
In
BareMetalHostProfilein thespec:volumeGroupssection, add the following configuration for the OpenStack compute nodes:spec: devices: - device: byName: /dev/nvme0n1 minSize: 30Gi wipe: true partitions: - name: lvm_nova_vol size: 0 wipe: true volumeGroups: - devices: - partition: lvm_nova_vol name: nova-vol
For details about
BareMetalHostProfile, see Operations Guide: Create a custom host profile.Configure the
OpenStackDeploymentCR to deploy OpenStack with LVM ephemeral storage. For example:spec: features: nova: images: backend: lvm lvm: volume_group: "nova-vol"
Optional. Enable encryption for the LVM ephemeral storage by adding the following metadata in the
OpenStackDeploymentCR:spec: features: nova: images: encryption: enabled: true cipher: "aes-xts-plain64" key_size: 256
Caution
Both live and cold migrations are not supported for such instances.
Enable LVM block storage¶
TechPreview
Note
Consider this section as part of Deploy an OpenStack cluster.
This section instructs you on how to configure LVM as a backend for the OpenStack Block Storage service.
You can use flexible size units throughout bare metal host profiles.
For example, you can now use either sizeGiB: 0.1 or size: 100Mi
when specifying a device size.
Mirantis recommends using only one parameter name type and units throughout
the configuration files. If both sizeGiB and size are used,
sizeGiB is ignored during deployment and the suffix is adjusted
accordingly. For example, 1.5Gi will be serialized as 1536Mi. The size
without units is counted in bytes. For example, size: 120 means 120 bytes.
Warning
All data will be wiped during cluster deployment on devices
defined directly or indirectly in the fileSystems list of
BareMetalHostProfile. For example:
A raw device partition with a file system on it
A device partition in a volume group with a logical volume that has a file system on it
An mdadm RAID device with a file system on it
An LVM RAID device with a file system on it
The wipe field is always considered true for these devices.
The false value is ignored.
Therefore, to prevent data loss, move the necessary data from these file systems to another server beforehand, if required.
To enable LVM block storage:
Open
BareMetalHostProfilefor editing.In the
spec:volumeGroupssection, specify the following data for the OpenStack compute nodes. In the following example, we deploy a Cinder volume with LVM on compute nodes. However, you can use dedicated nodes for this purpose.spec: devices: - device: byName: /dev/nvme0n1 minSize: 30Gi wipe: true partitions: - name: lvm_cinder_vol size: 0 wipe: true volumeGroups: - devices: - partition: lvm_cinder_vol name: cinder-vol
Important
Since MOSK 23.1, the
open-iscsipackage does not install by default on bare metal hosts. Install it manually during cluster deployment inBareMetalHostProfilein thespec:postDeployScriptsection:spec: postDeployScript: | #!/bin/bash -ex apt-get update apt-get install --no-install-recommends -y open-iscsi echo $(date) 'post_deploy_script done' >> /root/post_deploy_done
For details about
BareMetalHostProfile, see Operations Guide: Create a custom host profile.Configure the
OpenStackDeploymentCR to deploy OpenStack with LVM block storage. For example:spec: nodes: rockoon-openstack-compute-node::enabled: features: cinder: volume: backends: lvm: lvm: volume_group: "cinder-vol"
Enable DPDK with OVS¶
TechPreview
Note
Consider this section as part of Deploy an OpenStack cluster.
This section instructs you on how to enable DPDK with the Neutron OVS back end.
Warning
Usage of third-party software, which is not part of Mirantis-supported configurations, for example, the use of custom DPDK modules, may block upgrade of an operating system distribution. Users are fully responsible for ensuring the compatibility of such custom components with the latest supported Ubuntu version.
To enable DPDK with OVS:
Verify that your deployment meets the following requirements:
The required drivers have been installed on the host operating system.
Different Poll Mode Driver (PMD) types may require different kernel drivers to properly work with NIC. For more information about the DPDK drivers, read DPDK official documentation: Linux Drivers and Overview of Networking Drivers.
The DPDK NICs are not used on the host operating system.
The huge pages feature is enabled on the host. See Enable huge pages for OpenStack for details.
Enable DPDK in the OsDpl custom resource through the
node specific overridessettings. For example:spec: nodes: <NODE-LABEL>::<NODE-LABEL-VALUE>: features: neutron: dpdk: bridges: - ip_address: 10.12.2.80/24 name: br-phy driver: igb_uio enabled: true nics: - bridge: br-phy name: nic01 pci_id: "0000:05:00.0" tunnel_interface: br-phy
Enable SR-IOV with OVS¶
Note
Consider this section as part of Deploy an OpenStack cluster.
This section instructs you on how to enable SR-IOV with the Neutron OVS back end.
To enable SR-IOV with OVS:
Verify that your deployment meets the following requirements:
NICs with the SR-IOV support are installed
SR-IOV and VT-d are enabled in BIOS
Enable IOMMU in the kernel by configuring
intel_iommu=onin the GRUB configuration file. Specify the parameter for compute nodes inBareMetalHostProfilein thegrubConfigsection:spec: grubConfig: defaultGrubOptions: - 'GRUB_CMDLINE_LINUX="$GRUB_CMDLINE_LINUX intel_iommu=on"'
Configure the
OpenStackDeploymentCR to deploy OpenStack with the VLAN tenant network encapsulation.Caution
To ensure correct appliance of the configuration changes, configure VLAN segmentation during the initial OpenStack deployment.
Configuration example:
spec: features: neutron: external_networks: - bridge: br-ex interface: pr-floating mtu: null network_types: - flat physnet: physnet1 vlan_ranges: null - bridge: br-tenant interface: bond0 network_types: - vlan physnet: tenant vlan_ranges: 490:499,1420:1459 tenant_network_types: - vlan
Enable SR-IOV in the
OpenStackDeploymentCR through the node-specific overrides settings. For example:spec: nodes: <NODE-LABEL>::<NODE-LABEL-VALUE>: features: neutron: sriov: enabled: true nics: - device: enp10s0f1 num_vfs: 7 physnet: tenant
Enable BGP VPN¶
TechPreview
Note
Consider this section as part of Deploy an OpenStack cluster.
The BGP VPN service is an extra OpenStack Neutron plugin that enables connection of OpenStack Virtual Private Networks with external VPN sites through either BGP/MPLS IP VPNs or E-VPN.
To enable the BGP VPN service:
Enable BGP VPN in the OsDpl custom resource through the
node specific overrides settings. For example:
spec:
features:
neutron:
bgpvpn:
enabled: true
route_reflector:
# Enable deploygin FRR route reflector
enabled: true
# Local AS number
as_number: 64512
# List of subnets we allow to connect to
# router reflector BGP
neighbor_subnets:
- 10.0.0.0/8
- 172.16.0.0/16
nodes:
rockoon-openstack-compute-node::enabled:
features:
neutron:
bgpvpn:
enabled: true
When the service is enabled, a route reflector is scheduled on nodes with
the openstack-frrouting: enabled label. Mirantis recommends collocating
the route reflector nodes with the OpenStack controller nodes. By default, two
replicas are deployed.
Encrypt the east-west traffic¶
TechPreview
Note
Consider this section as part of Deploy an OpenStack cluster.
MOSK allows configuring Internet Protocol Security (IPSec) encryption for the east-west tenant traffic between the OpenStack compute nodes and gateways. The feature uses the strongSwan open source IPSec solution. Authentication is accomplished through a pre-shared key (PSK). However, other authentication methods are upcoming.
To encrypt the east-west tenant traffic, enable ipsec in the
spec:features:neutron settings of the OpenStackDeployment CR:
spec:
features:
neutron:
ipsec:
enabled: true
Caution
Enabling IPSec adds extra headers to the tenant traffic. The header size varies depending on IPSec configuration.
Therefore, Mirantis recommends decreasing network MTU for virtual networks and reserve 73 bytes overhead for the worst-case scenario as described in Cisco documentation: Configuring IPSec VPN Fragmentation and MTU.
Enable Cinder backend for Glance¶
TechPreview
Note
Consider this section as part of Deploy an OpenStack cluster.
This section instructs you on how to configure Cinder backend for the for images through the OpenStackDeployment CR.
Note
This feature depends heavily on Cinder multi-attach, which enables you to simultaneously attach volumes to multiple instances. Therefore, only the block storage backends that support multi-attach can be used.
To configure a Cinder backend for Glance, define the backend identity in the OpenStackDeployment CR. This identity will be used as a name for the backend section in the Glance configuration file.
When defining the backend:
Configure one of the backends as default.
Configure each backend to use specific Cinder volume type.
Note
You can use the
cinder_volume_typeparameter instead ofbackend_name. If so, you have to create this volume type beforehand and take into account that the bootstrap script does not manage such volume types.
The blockstore identity definition example:
spec:
features:
glance:
backends:
cinder:
blockstore:
default: true
backend_name: <volume_type:volume_name>
# e.g. backend_name: lvm:lvm_store
spec:
features:
glance:
backends:
cinder:
blockstore:
default: true
cinder_volume_type: netapp
Enable Cinder volume encryption¶
TechPreview
Note
Consider this section as part of Deploy an OpenStack cluster.
This section instructs you on how to enable Cinder volume encryption
through the OpenStackDeployment CR using Linux Unified Key Setup (LUKS)
and store the encryption keys in Barbican. For details, see
Volume encryption.
To enable Cinder volume encryption:
In the
OpenStackDeploymentCR, specify the LUKS volume type and configure the required encryption parameters for the storage system to encrypt or decrypt the volume.The
volume_typesdefinition example:spec: services: block-storage: cinder: values: bootstrap: volume_types: volumes-hdd-luks: arguments: encryption-cipher: aes-xts-plain64 encryption-control-location: front-end encryption-key-size: 256 encryption-provider: luks volume_backend_name: volumes-hdd
To create an encrypted volume as a non-admin user and store keys in the Barbican storage, assign the
creatorrole to the user since the default Barbican policy allows only theadminorcreatorrole:openstack role add --project <PROJECT-ID> --user <USER-ID> --creator <CREATOR-ID> creator
Optional. To define an encrypted volume as a default one, specify
volumes-hdd-luksindefault_volume_typein the Cinder configuration:spec: services: block-storage: cinder: values: conf: cinder: DEFAULT: default_volume_type: volumes-hdd-luks
Advanced configuration for OpenStack compute nodes¶
Note
Consider this section as part of Deploy an OpenStack cluster.
The section describes how to perform advanced configuration for the OpenStack compute nodes. Such configuration can be required in some specific use cases, such as DPDK, SR-IOV, or huge pages features usage.
This section contains recommendations for configuration of an
OpenStackDeployment custom resource for the compute nodes
of the following types:
Compute nodes with the default configuration, without local NVMe storage and SR-IOV network interface cards (NICs)
Compute nodes with the NVMe local storage
Compute nodes with the SR-IOV NICs
Compute nodes with both the NVMe local storage and SR-IOV NICs
Note
If the local NVMe storage is enabled, Mirantis recommends using it and enable SR-IOV if possible.
Caution
Before using the NVMe local storage and mount point, define them
in BareMetalHostProfile. For example:
apiVersion: metal3.io/v1alpha1
kind: BareMetalHostProfile
...
spec:
devices:
...
- device:
byName: /dev/nvme0n1
minSizeGiB: 30
wipe: true
partitions:
- name: local-volumes-partition
sizeGiB: 0
wipe: true
...
fileSystems:
...
- fileSystem: ext4
partition: local-volumes-partition
# mountpoint for Nova images
mountPoint: /var/lib/nova
Note
If you mount the /var directory, review Mounting recommendations for the /var directory
before configuring BareMetalHostProfile.
Caution
To control the storage type (local NVMe or Ceph) for virtual machines, place a node into the OpenStack aggregate. For details, see OpenStack documentation: Host aggregates.
As defined in Node-specific configuration, each node with a non-default configuration
must be configured separately. Each Machine object must have a
configuration-specific label. For example, for a compute node with the local
NVMe storage:
apiVersion: cluster.k8s.io/v1alpha1
kind: Machine
...
spec:
providerSpec:
value:
nodeLabels:
- key: node-type
value: compute-nvme
Caution
The <NODE-LABEL> value must match one of the allowed labels
defined in the nodeLabels section of the Cluster object:
nodeLabels:
- key: <NODE-LABEL>
value: <NODE-LABEL-VALUE>
Mirantis recommends using node-type as the <NODE-LABEL> key. To view
the full list of allowed node labels:
kubectl \
-n <CLUSTER-NAMESPACE> \
get <CLUSTER-NAME> \
-o json \
| jq .status.providerStatus.releaseRefs.current.allowedNodeLabels
The list of node labels is read-only and cannot be modified.
For compute nodes with the SR-IOV NIC, use compute-sriov as the
node-type value of nodeLabels:
apiVersion: cluster.k8s.io/v1alpha1
kind: Machine
...
spec:
providerSpec:
value:
nodeLabels:
- key: node-type
value: compute-sriov
For compute nodes with the local NVMe storage and SR-IOV NICs, use the
compute-nvme-sriov as the node-type value of nodeLabels:
apiVersion: cluster.k8s.io/v1alpha1
kind: Machine
...
spec:
providerSpec:
value:
nodeLabels:
- key: node-type
value: compute-nvme-sriov
In the examples above, compute-sriov, compute-nvme-sriov, and
compute-nvme are human-readable string identifiers. You can use any unique
string value for each type of compute node.
In the OpenStackDeployment object of each node group, define its own
section that starts with <NODE-LABEL>::<NODE-LABEL-VALUE>::
apiVersion: lcm.mirantis.com/v1alpha1
kind: OpenStackDeployment
...
spec:
...
nodes:
node-type::compute-nvme:
features:
nova:
images:
backend: local
node-type::compute-sriov:
features:
neutron:
sriov:
enabled: true
nics:
- device: enp10s0f1
num_vfs: 7
physnet: tenant
node-type::compute-nvme-sriov:
features:
nova:
images:
backend: local
neutron:
sriov:
enabled: true
nics:
- device: enp10s0f1
num_vfs: 7
physnet: tenant
Note
Consider this section as part of Deploy an OpenStack cluster.
Mirantis OpenStack for Kubernetes (MOSK) enables you to configure the
vCPU model for all instances managed by the OpenStack Compute service (Nova)
using the following osdpl definition:
spec:
features:
nova:
vcpu_type: host-model
For the supported values and configuration examples, see Virtual CPU.
Note
Consider this section as part of Deploy an OpenStack cluster.
Note
The instruction provided in this section applies to both OpenStack with OVS and OpenStack with Tungsten Fabric topologies.
The huge pages OpenStack feature provides essential performance improvements
for applications that are highly memory IO-bound. Huge pages should be enabled
on a per compute node basis. By default, NUMATopologyFilter is enabled.
To activate the feature, you need to enable huge pages on the dedicated bare metal host as described in Enable huge pages in a host profile during the predeployment bare metal configuration.
Note
The multi-size huge pages are not fully supported by Kubernetes versions before 1.19. Therefore, define only one size in kernel parameters.
TechPreview
Note
Consider this section as part of Deploy an OpenStack cluster.
Warning
The below procedure applies only to deployments based on deprecated Ubuntu 20.04. For Ubuntu 22.04 that supports cgroup v2, use the cpushield module. For the procedure details, see Host operating system configuration.
CPU isolation is a way to force the system scheduler to use only some logical
CPU cores for processes. For compute hosts, you should typically isolate system
processes and virtual guests on different cores through the cpusets
mechanism in Linux kernel.
The Linux kernel and cpuset provide a mechanism to run tasks by limiting the
resources defined by a cpuset. The tasks can be moved from one cpuset to
another to use the resources defined in other cpusets. The cset Python tool
is a command-line interface to work with cpusets.
To configure CPU isolation using cpusets:
Configure core isolation:
Note
You can also automate this step during deployment by using the
postDeployscript as described in Create MOSK host profiles.cat <<-"EOF" > /usr/bin/setup-cgroups.sh #!/bin/bash set -x UNSHIELDED_CPUS=${UNSHIELDED_CPUS:-"0-3"} UNSHIELDED_MEM_NUMAS=${UNSHIELDED_MEM_NUMAS:-0} SHIELD_CPUS=${SHIELD_CPUS:-"4-15"} SHIELD_MODE=${SHIELD_MODE:-"cpuset"} # One of isolcpu or cpuset DOCKER_CPUS=${DOCKER_CPUS:-$UNSHIELDED_CPUS} DOCKER_MEM_NUMAS=${DOCKER_MEM_NUMAS:-$UNSHIELDED_MEM_NUMAS} KUBERNETES_CPUS=${KUBERNETES_CPUS:-$UNSHIELDED_CPUS} KUBERNETES_MEM_NUMAS=${KUBERNETES_MEM_NUMAS:-$UNSHIELDED_MEM_NUMAS} CSET_CMD=${CSET_CMD:-"python3 /usr/bin/cset"} if [[ ${SHIELD_MODE} == "cpuset" ]]; then ${CSET_CMD} set -c ${UNSHIELDED_CPUS} -m ${UNSHIELDED_MEM_NUMAS} -s system ${CSET_CMD} proc -m -f root -t system ${CSET_CMD} proc -k -f root -t system fi ${CSET_CMD} set --cpu=${DOCKER_CPUS} --mem=${DOCKER_MEM_NUMAS} --set=docker ${CSET_CMD} set --cpu=${KUBERNETES_CPUS} --mem=${KUBERNETES_MEM_NUMAS} --set=kubepods ${CSET_CMD} set --cpu=${DOCKER_CPUS} --mem=${DOCKER_MEM_NUMAS} --set=com.docker.ucp EOF chmod +x /usr/bin/setup-cgroups.sh cat <<-"EOF" > /etc/systemd/system/shield-cpus.service [Unit] Description=Shield CPUs DefaultDependencies=no After=systemd-udev-settle.service Before=lvm2-activation-early.service Wants=systemd-udev-settle.service [Service] ExecStart=/usr/bin/setup-cgroups.sh RemainAfterExit=true Type=oneshot Restart=on-failure #Service should restart on failure RestartSec=5s #Restart each five seconds until success [Install] WantedBy=basic.target EOF systemctl enable shield-cpus reboot
As root user, verify that isolation has been applied:
cset set -l
Example of system response:
cset: Name CPUs-X MEMs-X Tasks Subs Path ------------ ---------- - ------- - ----- ---- ---------- root 0-15 y 0 y 165 4 / kubepods 0-3 n 0 n 0 2 /kubepods docker 0-3 n 0 n 0 0 /docker system 0-3 n 0 n 65 0 /system com.docker.ucp 0-3 n 0 n 0 0 /com.docker.ucp
Run the
cpustresscontainer:docker run -it --name cpustress --rm containerstack/cpustress --cpu 4 --timeout 30s --metrics-brief
Verify that isolated cores are not affected:
htop
Example of system response highlighting the load created on all available Docker cores:
Note
Consider this section as part of Deploy an OpenStack cluster.
The majority of CPU topologies features are activated by NUMATopologyFilter
that is enabled by default. Such features do not require any further service
configuration and can be used directly on a vanilla MOSK
deployment. The list of the CPU topologies features includes, for example:
NUMA placement policies
CPU pinning policies
CPU thread pinning policies
CPU topologies
To enable libvirt CPU pinning through the node-specific overrides in the
OpenStackDeployment custom resource, use the following sample
configuration structure:
spec:
nodes:
<NODE-LABEL>::<NODE-LABEL-VALUE>:
services:
compute:
nova:
nova_compute:
values:
conf:
nova:
compute:
cpu_dedicated_set: 2-17
cpu_shared_set: 18-47
Note
Consider this section as part of Deploy an OpenStack cluster.
The Peripheral Component Interconnect (PCI) passthrough feature in OpenStack allows full access and direct control over physical PCI devices in guests. The mechanism applies to any kind of PCI devices including a Network Interface Card (NIC), Graphics Processing Unit (GPU), and any other device that can be attached to a PCI bus. The only requirement for the guest to properly use the device is to correctly install the driver.
To enable PCI passthrough in a MOSK deployment:
For Linux X86 compute nodes, verify that the following features are enabled on the host:
VT-d in BIOS
IOMMU on the host operating system as described in Enable SR-IOV with OVS.
Configure the
nova-apiservice that is scheduled on OpenStack controllers nodes. To generate the alias for PCI innova.conf, add the alias configuration through theOpenStackDeploymentCR.Note
When configuring PCI with SR-IOV on the same host, the values specified in
aliastake precedence. Therefore, add the SR-IOV devices topassthrough_whitelistexplicitly.For example:
spec: services: compute: nova: values: conf: nova: pci: alias: '{ "vendor_id":"8086", "product_id":"154d", "device_type":"type-PF", "name":"a1" }'
Configure the
nova-computeservice that is scheduled on OpenStack compute nodes. To enable Nova to pass PCI devices to virtual machines, configure thepassthrough_whitelistsection innova.confthrough the node-specific overrides in theOpenStackDeploymentCR. For example:spec: nodes: <NODE-LABEL>::<NODE-LABEL-VALUE>: services: compute: nova: nova_compute: values: conf: nova: pci: alias: '{ "vendor_id":"8086", "product_id":"154d", "device_type":"type-PF", "name":"a1" }' passthrough_whitelist: | [{"devname":"enp216s0f0","physical_network":"sriovnet0"}, { "vendor_id": "8086", "product_id": "154d" }]
Available since MOSK 23.1
MOSK enables you to configure initial oversubscription
through the OpenStackDeployment custom resource. For configuration details
and oversubscription considerations, refer to
Configuring initial resource oversubscription.
By default, the following values are applied:
8.0for the number of CPUs1.6for the disk space1.0for the amount of RAMNote
In MOSK 22.5 and earlier, the effective default value of RAM allocation ratio is
1.1.
Changing oversubscription configuration after deployment will only affect the newly added compute nodes and will not change oversubscription for already existing compute nodes. You can change oversubscription for existing compute nodes through the placement API as described in Change oversubscription settings for existing compute nodes.
Note
Consider this section as part of Deploy an OpenStack cluster.
Hyperconverged architecture combines OpenStack compute nodes along with Ceph nodes. To avoid nodes overloading, which can cause Ceph performance degradation and outage, limit the hardware resources consumption by the OpenStack compute services.
You can reserve hardware resources for non-workload related consumption using
the following nova-compute parameters. For details, see
OpenStack documentation: Overcommitting CPU and RAM
and OpenStack documentation: Configuration Options.
cpu_allocation_ratio- in case of a hyperconverged architecture, the value depends on the number of vCPU used for non-workload related operations, total number of vCPUs of a hyperconverged node, and on workload vCPU consumption:cpu_allocation_ratio = (${vCPU_count_on_a_hyperconverged_node} - ${vCPU_used_for_non_OpenStack_related_tasks}) / ${vCPU_count_on_a_hyperconverged_node} / ${workload_vCPU_utilization}To define the vCPU count used for non-OpenStack related tasks, use the following formula, considering the storage data plane performance tests:
vCPU_used_for_non-OpenStack_related_tasks = 2 * SSDs_per_hyperconverged_node + 1 * Ceph_OSDs_per_hyperconverged_node + 0.8 * Ceph_OSDs_per_hyperconverged_node
Consider the following example with 5 SSD disks for Ceph OSDs per hyperconverged node and 2 Ceph OSDs per disk:
vCPU_used_for_non-OpenStack_related_tasks = 2 * 5 + 1 * 10 + 0.8 * 10 = 28
In this case, if there are 40 vCPUs per hyperconverged node, 28 vCPUs are required for non-workload related calculations, and a workload consumes 50% of the allocated CPU time:
cpu_allocation_ratio = (40-28) / 40 / 0.5 = 0.6.
reserved_host_memory_mb- a dedicated variable in the OpenStack Nova configuration, to reserve memory for non-OpenStack related VM activities:reserved_host_memory_mb = 13 GB * Ceph_OSDs_per_hyperconverged_node
For example for 10 Ceph OSDs per hyperconverged node:
reserved_host_memory_mb = 13 GB * 10 = 130 GB = 133120
ram_allocation_ratio- the allocation ratio of virtual RAM to physical RAM. To completely exclude the possibility of memory overcommitting, set to1.
To limit HW resources for hyperconverged OpenStack compute nodes:
In the OpenStackDeployment CR, specify the cpu_allocation_ratio,
ram_allocation_ratio, and reserved_host_memory_mb parameters as
required using the calculations described above.
For example:
apiVersion: lcm.mirantis.com/v1alpha1
kind: OpenStackDeployment
spec:
services:
compute:
nova:
values:
conf:
nova:
DEFAULT:
cpu_allocation_ratio: 0.6
ram_allocation_ratio: 1
reserved_host_memory_mb: 133120
Note
For an existing OpenStack deployment:
Obtain the name of your
OpenStackDeploymentCR:kubectl -n openstack get osdpl
Open the
OpenStackDeploymentCR for editing and specify the parameters as required.kubectl -n openstack edit osdpl <osdpl name>
Available since MOSK 24.1 TechPreview
This section delves into virtual GPU configuration. It is specifically tailored for NVIDIA physical GPUs, with a focus on the A100 40GB GPU and NVIDIA AIE 4.1 drivers.
While setup procedures may vary among different cards and vendors, MOSK can generally ensure compatibility between the MOSK Compute service (Nova) and vGPU functionality, as long as the drivers for the physical GPU expose an VFIO mdev-compatible interface to the Linux host.
For configuration specifics of other physical GPUs, refer to the official documentation provided by the vendor.
Visit NVIDIA AI Enterprise documentation for comprehensive guidance on how to download the required drivers.
Also, if you have access to the NVIDIA NGC Catalog, search for the latest Infra Release that provides NVIDIA vGPU Host Driver there.
NVIDIA licensing
To fully utilize the capabilities of NVIDIA GPU virtualization, you may need to set up and configure the NVIDIA licensing server.
To install the acquired drivers within your cluster, add a custom
postDeployScript script to the custom BareMetalHostProfile object
used for the compute nodes with GPUs.
Note
For the instruction on how to create a BareMetalHostProfile
object, refer to Operations Guide: Create a custom host profile.
This script must solve the following tasks:
Download and install the drivers, if needed
Configure physical GPU according to your cluster requirements
Configure a startup task to reconfigure the physical GPU after node reboots.
Example postDeployScript script:
#!/bin/bash -ex
# Create a one time script that will initialize physical GPU right now and self-destruct
cat << EOF > /root/test_postdeploy_job.sh
#!/bin/bash -ex
systemctl enable initialize-vgpu
systemctl start --no-block initialize-vgpu
crontab -l | grep -v test_postdeploy_job.sh | crontab -
rm /root/test_postdeploy_job.sh
EOF
mkdir -p /var/spool/cron/crontabs/ && echo "*/1 * * * * sudo /root/test_postdeploy_job.sh >> /var/log/test_postdeploy_job.log 2>&1" >> /var/spool/cron/crontabs/root
chmod +x /root/test_postdeploy_job.sh
# Create a systemd unit that will re-initialize physical GPU on restart
cat << EOF > /etc/systemd/system/initialize-vgpu.service
[Unit]
Description=Configure VGPU
After=systemd-modules-load.service
[Service]
Type=oneshot
ExecStart=/root/initialize_vgpu.sh
RemainAfterExit=true
StandardOutput=journal
[Install]
RequiredBy=multi-user.target
EOF
cat << EOF > /root/initialize_vgpu.sh
#!/bin/bash
set -ex
while ! docker inspect ucp-kubelet;
do echo "Waiting lcm-agent is finished.";
sleep 1;
done
# Download and install the driver, dependencies and tools
if [[ ! -f /root/nvidia-vgpu-ubuntu-aie-535_535.129.03_amd64.deb ]]; then
apt update
apt install -y dkms unzip gcc libc-dev make linux-headers-$(uname -r) pciutils lshw mdevctl
wget https://my.intra.net//root/gpu-driver-x-y-z.deb -O /root/gpu-driver-x-y-z.deb
apt install /root/gpu-driver-x-y-z.deb
systemctl enable initialize-vgpu
fi
systemctl restart nvidia-vgpud.service
# Enable SR-IOV mode for the pGPU
/usr/lib/nvidia/sriov-manage -e <PCI-ADDRESS-OF-NVIDIA-CARD>
# Enable MIG mode for pGPU
nvidia-smi -i 0 -mig 1
systemctl enable nvidia-vgpu-mgr.service
systemctl start nvidia-vgpu-mgr.service
EOF
chmod +x /root/initialize_vgpu.sh
Virtual GPU types are similar to compute flavors as they determine the resources allocated to each virtual GPU. This allows for efficient allocation and optimization of GPU resources in virtualized environments.
Each physical GPU has a maximum number of virtual GPUs of a specific type that can be created on it, with no possibility for overallocation. In the time-sliced vGPU configuration, each particular physical GPU can only instantiate vGPUs of the same selected type. In the Multi-Instance GPU (MIG), a single physical GPU may be partitioned into several differently sized virtual GPUs.
Either way, prior to accepting workloads, Mirantis recommends determining the virtual GPU types that each of your physical GPU will provide. Altering these settings afterward necessitates terminating every virtual machine currently running on the physical GPU intended for reconfiguration or repurposing for another virtual GPU type.
This section outlines the process for partitioning physical GPUs into Multi-Instance GPUs (MIG) using the nvidia-smi tool provided by the NVIDIA Host GPU driver.
To list available virtual GPU instance profiles:
nvidia-smi mig -lgip
Example system response:
+-----------------------------------------------------------------------------+
| GPU instance profiles: |
| GPU Name ID Instances Memory P2P SM DEC ENC |
| Free/Total GiB CE JPEG OFA |
|=============================================================================|
| 0 MIG 1g.5gb 19 7/7 4.75 No 14 0 0 |
| 1 0 0 |
+-----------------------------------------------------------------------------+
| 0 MIG 1g.5gb+me 20 1/1 4.75 No 14 1 0 |
| 1 1 1 |
+-----------------------------------------------------------------------------+
| 0 MIG 1g.10gb 15 4/4 9.75 No 14 1 0 |
| 1 0 0 |
+-----------------------------------------------------------------------------+
| 0 MIG 2g.10gb 14 3/3 9.75 No 28 1 0 |
| 2 0 0 |
+-----------------------------------------------------------------------------+
| 0 MIG 3g.20gb 9 2/2 19.62 No 42 2 0 |
| 3 0 0 |
+-----------------------------------------------------------------------------+
| 0 MIG 4g.20gb 5 1/1 19.62 No 56 2 0 |
| 4 0 0 |
+-----------------------------------------------------------------------------+
| 0 MIG 7g.40gb 0 1/1 39.50 No 98 5 0 |
| 7 1 1 |
+-----------------------------------------------------------------------------+
To create seven, which is a maximum possible amount of instances according to the system response above, MIG vGPUs of the smallest size:
nvidia-smi mig -cgi 19,19,19,19,19,19,19
To create three differently sized vGPUs of 4g.20gb, 2g.10gb, and
1g.5gb sizes:
nvidia-smi mig -cgi 5,14,19
Caution
Keep in mind that not all combinations of differently sized vGPU instances are supported. Additionally, the order in which you create vGPUs is important.
For example configurations, see NVIDIA documentation.
To correctly configure the MOSK Compute service, you need to correlate the following naming schemes related to virtual GPU types:
The GPU instance profile as reported by nvidia-smi mig. For example,
MIG 1g.5gb.The vGPU type as reported by the driver. For example,
GRID A100-1-5C.The mdev class that corresponds to the vGPU type. For example,
nvidia-474.
For the compatibility between GPU instance profiles and virtual GPU types, refer to NVIDIA documentation: Virtual GPU Types for Supported GPUs.
To determine the mdev class supported by a specific virtual GPU type listed by a PCI device address, verify the output of the mdevctl types command executed on the compute node that has a physical GPU available on it:
mdevctl types
Example system response for MIGs:
0000:42:00.4
nvidia-1053
Available instances: 0
Device API: vfio-pci
Name: GRID A100-1-10C
Description: num_heads=1, frl_config=60, framebuffer=10240M, max_resolution=4096x2400, max_instance=4
...
nvidia-474
Available instances: 1
Device API: vfio-pci
Name: GRID A100-1-5C
Description: num_heads=1, frl_config=60, framebuffer=5120M, max_resolution=4096x2400, max_instance=7
...
The Name field from the example system output above corresponds to the
supported virtual GPU type, linking the GPU instance profile with the mdev class
supported by your physical GPU.
In the example above, the MIG 1g.5gb GPU instance profile corresponds
to the GRID A100-1-5C vGPU type as per NVIDIA documentation, and according
to the mdevctl types` output, it corresponds to the nvidia-474
mdev class.
Note
Notice that Available instances is zero for vGPU types that are
not actually supported by this given card and configuration. For MIGs,
the Available instances will be non-zero only for the virtual GPU types
for which the MIG virtual GPU instances have already been created. See
Partition to Multi-Instance GPUs.
The parameters you need to define for the nova-compute service on each
compute node with physical GPUs you want to expose as virtual GPUs include:
[devices]enabled_mdev_typesRequired. List of the mdev classes, see the previous step for details.
[devices]cleanup_mdev_devicesOptional. By default, the Compute service does not delete created mdev devices but reuses them instead. While this speeds up processes, it may pose challenges when reconfiguring the
enabled_mdev_typesparameter. Setcleanup_mdev_devicestoTruefor the Compute service to auto-delete created mdev devices upon instance deletion.
If you plan to use only time-sliced vGPUs and provide a single virtal GPU type
across the entire cloud, you only need to configure the options mentioned
above once globally for all compute nodes through the spec.services section
of the OpenStackDeployment custom resource.
With the configuration below, the Compute service will auto-detect all PCI
devices that provide this mdev type and automatically create required resource
providers in the placement service with the resource class VGPU.
Example configuration for the nvidia-474 mdev type:
kind: OpenStackDeployment
spec:
services:
compute:
nova:
values:
conf:
nova:
devices:
enabled_mdev_types: nvidia-474
cleanup_mdev_devices: true
If you plan to provide multiple time-sliced vGPU types, simplify the
configuration by grouping the nodes based on a node label (not necessarily
aggregates). Ensure that each group exposes only one mdev type using the
Node-specific configuration settings. Additionally, use custom resource classes to
facilitate flavor creation, ensuring consistent use of the CUSTOM_ prefix
for custom mdev_class.
For example, if you want to provide the nvidia-474 and nvidia-475 mdev
types, label your nodes with the vgpu-type=nvidia-474 and
vgpu-type=nvidia-475 labels and use the following node-specific settings:
kind: OpenStackDeployment
spec:
nodes:
vgpu-type::nvidia-474:
services:
compute:
nova:
nova_compute:
values:
conf:
nova:
devices:
enabled_mdev_types: nvidia-474
cleanup_mdev_devices: true
mdev_nvidia-474:
mdev_class: CUSTOM_VGPU_A100_1_5C
vgpu-type::nvidia-475:
services:
compute:
nova:
nova_compute:
values:
conf:
nova:
devices:
enabled_mdev_types: nvidia-475
cleanup_mdev_devices: true
mdev_nvidia-475:
mdev_class: CUSTOM_VGPU_A100_2_10C
The configuration above creates corresponding resource providers in the
placement service that provide CUSTOM_VGPU_A100_1_5C or
CUSTOM_VGPU_A100_2_10C resources.
You can use these resources during the definition of flavors for instances
with corresponding vGPU types.
In some cases, you may need to provide different vGPU types from a single compute node, for example, if the compute node has 2 physical GPUs and you want to create two different types of vGPU on them. For such scenarios, you should provide explicit PCI device addresses of these physical GPUs in the settings. This makes such configuration verbose in heterogeneous hardware environments where physical GPUs have different PCI addresses on each node. For example, when targeting node-specific settings by node name:
kind: OpenStackDeployment
spec:
nodes:
kubernetes.io/hostname::kaas-node-7af9aab1-596d-4ba3-a717-846653aa441a:
services:
compute:
nova:
nova_compute:
values:
conf:
nova:
devices:
enabled_mdev_types: nvidia-474,nvidia-475
cleanup_mdev_devices: true
mdev_nvidia-474:
device_addresses: 0000:42:00.0
mdev_class: CUSTOM_VGPU_A100_1_5C
mdev_nvidia-475:
device_addresses: 0000:43:00.0
mdev_class: CUSTOM_VGPU_A100_2_10C
In the SR-IOV mode, the driver typically creates more virtual functions than the maximum capacity of the physical GPU, even for the smallest virtual GPU type. Each virtual function can hold only one single virtual GPU. This leads to resource over-reporting to the placement service.
Therefore, to ensure efficient resource allocation and utilization within a homogeneous hardware environment, assuming that each compute node in it has the same PCI address for the physical GPU and the physical GPU has been partitioned to the MIG GPU instances identically:
Identify the number of instances created of each MIG profile.
Select random but not overlapping sets of PCI addresses from the list of virtual functions of the physical GPU. The amount of addresses in each set must correspond to the number of instances created of each MIG profile.
Assign the mdev type to the selected devices.
For example, for the environment with the following configuration:
3 MIG instances of
MIG 1.5gband 2 MIG instances ofMIG 2.10gb16 virtual functions created for the physical GPU with the PCI address range from
0000:42:00.0to0000:42:01.7
Pick 3 and 2 random PCI addresses from that pool and assign them to
CUSTOM_VGPU_A100_1_5C and CUSTOM_VGPU_A100_2_10C mdev classes
respectively:
spec:
services:
compute:
nova:
values:
conf:
nova:
devices:
enabled_mdev_types: nvidia-474,nvidia-475
cleanup_mdev_devices: true
mdev_nvidia-474:
device_addresses: 0000:42:00.0,0000:42:00.1,0000:42:00.2
mdev_class: CUSTOM_VGPU_A100_1_5C
mdev_nvidia-475:
device_addresses: 0000:42:01.0,0000:42:01.1
mdev_class: CUSTOM_VGPU_A100_2_10C
In a heterogeneous hardware environment, use node-specific settings to group nodes with the same PCI addresses and intended vGPU configuration, or use explicit setting for each node targeting node-specific settings to every node, sequentially if needed.
This section provides guidelines for verifying that virtual GPUs are correctly accounted for in the OpenStack Placement service, ensuring proper scheduling of instances that utilize virtual GPUs.
Firstly, verify that resource providers have been created with accurate
inventories. For each PCI device associated with a virtual GPU, including
virtual instances in the case of MIG/SR-IOV, there should be a nested resource
provider under the resource provider of the corresponding compute node.
The name of this nested resource provider should follow the format
<node-name>_pci_<pci-address-with-underscores>:
openstack resource provider list --resource CUSTOM_VGPU_A100_1_5C=1 -f yaml
Example system response:
- generation: 1
name: kaas-node-9d18b7c8-7ea8-4b13-abe9-0e76ee8db596.kaas-kubernetes-294cbb1cbf084789b931ebc54d3f9b05_pci_0000_42_00_4
parent_provider_uuid: c922488b-69eb-42a8-afad-dc7d3d56b8fd
root_provider_uuid: c922488b-69eb-42a8-afad-dc7d3d56b8fd
uuid: 963bb3ce-3ed1-421f-a186-a808c3460c48
...
Also, examine the inventory of each resource provider. It should exclusively
consist of the VGPU resource or any custom resource name configured in
the Compute service settings. The total capacity of the resource should match
the capacity reported by the mdevctl types output, reflecting the
capabilities of the PCI device for the specified mdev class. In the case
of MIG, this total capacity will always be 1.
openstack resource provider inventory list 963bb3ce-3ed1-421f-a186-a808c3460c48 -f yaml
Example system response:
- allocation_ratio: 1.0
max_unit: 1
min_unit: 1
reserved: 0
resource_class: CUSTOM_VGPU_A100_1_5C
step_size: 1
total: 1
used: 0
This section provides instructions for creating a flavor that requests a specific virtual GPU resource, using the mdev classes configured in the Compute service and registered in the placement service.
To create the flavor, use the openstack flavor create command.
Ensure that the flavor properties match the configured mdev classes in the
Compute service. For example, to request one vGPU of type nvidia-474 using
the resource class from the previous examples:
openstack flavor create --ram 1024 --vcpus 2 --disk 5 --property resources:CUSTOM_VGPU_A100_1_5C=1
Replace the --property resources:CUSTOM_VGPU_A100_1_5C=1 parameter with the
appropriate property matching the desired virtual GPU type and quantity.
Once the flavor is created, you can start launching instances using the created flavor as usual.
Enable image signature verification¶
TechPreview
Note
Consider this section as part of Deploy an OpenStack cluster.
Mirantis OpenStack for Kubernetes (MOSK) enables you to perform image signature verification when booting an OpenStack instance, uploading a Glance image with signature metadata fields set, and creating a volume from an image.
To enable signature verification, use the following osdpl definition:
spec:
features:
glance:
signature:
enabled: true
When enabled during initial deployment, all internal images such as Amphora, Ironic, and test (CirrOS, Fedora, Ubuntu) images, will be signed by a self-signed certificate.
Configure LoadBalancer for PowerDNS¶
Note
Consider this section as part of Deploy an OpenStack cluster.
Mirantis OpenStack for Kubernetes (MOSK) allows configuring
LoadBalancer for the Designate PowerDNS backend. For example, you can expose a
TCP port for zone transferring using the following exemplary osdpl
definition:
spec:
designate:
backend:
external_ip: 10.172.1.101
protocol: udp
type: powerdns
For the supported values, see LoadBalancer type for PowerDNS.
Access OpenStack after deployment¶
This section contains the guidelines on how to access your MOSK OpenStack environment.
Configure DNS to access OpenStack¶
DNS is a mandatory component for MOSK deployment, all records must be created on the customer DNS server. The OpenStack services are exposed through the Ingress NGINX controller.
Warning
This document describes how to temporarily configure DNS. The workflow contains non-permanent changes that will be rolled back during a managed cluster update or reconciliation loop. Therefore, proceed at your own risk.
To configure DNS to access your OpenStack environment:
Obtain the external IP address of the Ingress service:
kubectl -n openstack get services ingress
Example of system response:
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE ingress LoadBalancer 10.96.32.97 10.172.1.101 80:34234/TCP,443:34927/TCP,10246:33658/TCP 4h56m
Select from the following options:
If you have a corporate DNS server, update your corporate DNS service and create appropriate DNS records for all OpenStack public endpoints.
To obtain the full list of public endpoints:
kubectl -n openstack get ingress -ocustom-columns=NAME:.metadata.name,HOSTS:spec.rules[*].host | awk '/namespace-fqdn/ {print $2}'
Example of system response:
barbican.it.just.works cinder.it.just.works cloudformation.it.just.works designate.it.just.works glance.it.just.works heat.it.just.works horizon.it.just.works keystone.it.just.works neutron.it.just.works nova.it.just.works novncproxy.it.just.works octavia.it.just.works placement.it.just.works
If you do not have a corporate DNS server, perform one of the following steps:
Add the appropriate records to
/etc/hostslocally. For example:10.172.1.101 barbican.it.just.works 10.172.1.101 cinder.it.just.works 10.172.1.101 cloudformation.it.just.works 10.172.1.101 designate.it.just.works 10.172.1.101 glance.it.just.works 10.172.1.101 heat.it.just.works 10.172.1.101 horizon.it.just.works 10.172.1.101 keystone.it.just.works 10.172.1.101 neutron.it.just.works 10.172.1.101 nova.it.just.works 10.172.1.101 novncproxy.it.just.works 10.172.1.101 octavia.it.just.works 10.172.1.101 placement.it.just.works
Deploy your DNS server on top of Kubernetes:
Deploy a standalone CoreDNS server by including the following configuration into
coredns.yaml:apiVersion: lcm.mirantis.com/v1alpha1 kind: HelmBundle metadata: name: coredns namespace: osh-system spec: repositories: - name: hub_stable url: https://charts.helm.sh/stable releases: - name: coredns chart: hub_stable/coredns version: 1.8.1 namespace: coredns values: image: repository: mirantis.azurecr.io/openstack/extra/coredns tag: "1.6.9" isClusterService: false servers: - zones: - zone: . scheme: dns:// use_tcp: false port: 53 plugins: - name: cache parameters: 30 - name: errors # Serves a /health endpoint on :8080, required for livenessProbe - name: health # Serves a /ready endpoint on :8181, required for readinessProbe - name: ready # Required to query kubernetes API for data - name: kubernetes parameters: cluster.local - name: loadbalance parameters: round_robin # Serves a /metrics endpoint on :9153, required for serviceMonitor - name: prometheus parameters: 0.0.0.0:9153 - name: forward parameters: . /etc/resolv.conf - name: file parameters: /etc/coredns/it.just.works.db it.just.works serviceType: LoadBalancer zoneFiles: - filename: it.just.works.db domain: it.just.works contents: | it.just.works. IN SOA sns.dns.icann.org. noc.dns.icann.org. 2015082541 7200 3600 1209600 3600 it.just.works. IN NS b.iana-servers.net. it.just.works. IN NS a.iana-servers.net. it.just.works. IN A 1.2.3.4 *.it.just.works. IN A 1.2.3.4
Update the public IP address of the Ingress service:
sed -i 's/1.2.3.4/10.172.1.101/' coredns.yaml kubectl apply -f coredns.yaml
Verify that the DNS resolution works properly:
Assign an external IP to the service:
kubectl -n coredns patch service coredns-coredns --type='json' -p='[{"op": "replace", "path": "/spec/ports", "value": [{"name": "udp-53", "port": 53, "protocol": "UDP", "targetPort": 53}]}]' kubectl -n coredns patch service coredns-coredns --type='json' -p='[{"op": "replace", "path": "/spec/type", "value":"LoadBalancer"}]'
Obtain the external IP address of CoreDNS:
kubectl -n coredns get service coredns-coredns
Example of system response:
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE coredns-coredns ClusterIP 10.96.178.21 10.172.1.102 53/UDP,53/TCP 25h
Point your machine to use the correct DNS. It is
10.172.1.102in the example system response above.If you plan to launch Tempest tests or use the OpenStack client from a
keystone-client-XXXpod, verify that the Kubernetes built-in DNS service is configured to resolve your public FQDN records by adding your public domain toCorefile. For example, to add theit.just.worksdomain:kubectl -n kube-system get configmap coredns -oyaml
Example of system response:
apiVersion: v1 data: Corefile: | .:53 { errors health ready kubernetes cluster.local in-addr.arpa ip6.arpa { pods insecure fallthrough in-addr.arpa ip6.arpa } prometheus :9153 forward . /etc/resolv.conf cache 30 loop reload loadbalance } it.just.works:53 { errors cache 30 forward . 10.96.178.21 }
Access your OpenStack environment¶
This section explains how to access your OpenStack environment as admin user.
Before you proceed, make sure that you can access the Kubernetes API and have
privileges to read secrets from the openstack-external namespace in
Kubernetes or you are able to exec to the pods in the openstack
namespace.
You can use the built-in admin CLI client and execute the openstack
commands from a dedicated pod deployed in the openstack namespace:
kubectl -n openstack exec \
$(kubectl -n openstack get pod -l application=keystone,component=client -ojsonpath='{.items[*].metadata.name}') \
-ti -- bash
This pod has python-openstackclient and all required plugins already
installed. The python-openstackclient command-line client is configured
to use the admin user credentials. You can view the detailed configuration
for the openstack command in /etc/openstack/clouds.yaml file in
the pod.
Configure the external DNS resolution for OpenStack services as described in Configure DNS to access OpenStack.
Obtain the admin user credentials from the
openstack-identity-credentialssecret in theopenstack-externalnamespace:kubectl -n openstack-external get secrets openstack-identity-credentials -o jsonpath='{.data.clouds\.yaml}' | base64 -d
Example of a system response:
clouds: admin: auth: auth_url: https://keystone.it.just.works/ password: <ADMIN_PWD> project_domain_name: <ADMIN_PROJECT_DOMAIN> project_name: <ADMIN_PROJECT> user_domain_name: <ADMIN_USER_DOMAIN> username: <ADMIN_USER_NAME> endpoint_type: public identity_api_version: 3 interface: public region_name: CustomRegion admin-system: auth: auth_url: https://keystone.it.just.works/ password: <ADMIN_PWD> system_scope: all user_domain_name: <ADMIN_USER_DOMAIN> username: <ADMIN_USER_NAME> endpoint_type: public identity_api_version: 3 interface: public region_name: CustomRegion
Access Horizon through your browser using its public service. For example,
https://horizon.it.just.works.To log in, specify the user name and domain name obtained in previous step from the
<ADMIN_USER_NAME>and<ADMIN_USER_DOMAIN>values.If the OpenStack Identity service has been deployed with the OpenID Connect integration:
From the Authenticate using drop-down menu, select OpenID Connect.
Click Connect. You will be redirected to your identity provider to proceed with the authentication.
Note
If OpenStack has been deployed with self-signed TLS certificates for public endpoints, you may get a warning about an untrusted certificate. To proceed, allow the connection.
To be able to access your OpenStack environment through the CLI,
you need to configure the openstack client environment
using either an openstackrc environment file or clouds.yaml file.
Log in to Horizon as described in Access an OpenStack environment through Horizon.
Download the
openstackrcfile from the web UI.On any shell from which you want to run OpenStack commands, source the environment file for the respective project.
Obtain
clouds.yaml:mkdir -p ~/.config/openstack kubectl -n openstack-external get secrets openstack-identity-credentials -o jsonpath='{.data.clouds\.yaml}' | base64 -d > ~/.config/openstack/clouds.yaml
The OpenStack client looks for
clouds.yamlin the following locations: current directory,~/.config/openstack, and/etc/openstack.Export the
OS_CLOUDenvironment variable:export OS_CLOUD=admin
Now, you can use the openstack CLI as usual. For example:
openstack user list
Example of an expected system response:
+----------------------------------+-----------------+
| ID | Name |
+----------------------------------+-----------------+
| dc23d2d5ee3a4b8fae322e1299f7b3e6 | internal_cinder |
| 8d11133d6ef54349bd014681e2b56c7b | admin |
+----------------------------------+-----------------+
Note
If OpenStack was deployed with self-signed TLS certificates for public endpoints, you may need to use the openstack command-line client with certificate validation disabled. For example:
openstack --insecure user list
Troubleshoot an OpenStack deployment¶
This section provides the general debugging instructions for your OpenStack on Kubernetes deployment. Start your troubleshooting with the determination of the failing component that can include the OpenStack Controller (Rockoon), Helm, a particular pod, or service.
Note
For Kubernetes cluster debugging and troubleshooting, refer to Kubernetes official documentation: Troubleshoot clusters and Docker Enterprise v3.0 documentation: Monitor and troubleshoot.
Debugging the Helm releases¶
OpenStack Controller (Rockoon)
Since MOSK 25.1, the OpenStack Controller has been open-sourced under the name Rockoon and is maintained as an independent open-source project going forward.
As part of this transition, all openstack-controller pods are named
rockoon pods across the MOSK documentation and deployments. This change
does not affect functionality, but this is the reminder for the users to
utilize the new naming for pods and other related artifacts accordingly.
Note
MOSK uses direct communication with Helm 3.
Log in to the
rockoonpod, where the Helm v3 client is installed, or download the Helm v3 binary locally:kubectl -n osh-system get pods |grep rockoon
Example of a system response:
rockoon-5c5965688b-knck5 9/9 Running 0 21h rockoon-admission-6795d594b4-rp7kf 1/1 Running 0 22h rockoon-exporter-6f66547b67-pcxhh 1/1 Running 0 22h
Verify the Helm releases statuses:
helm3 --namespace openstack list --all
Example of a system response:
NAME NAMESPACE REVISION UPDATED STATUS CHART APP VERSION etcd openstack 4 2021-07-09 11:06:25.377538008 +0000 UTC deployed etcd-0.1.0-mcp-2735 ingress-openstack openstack 4 2021-07-09 11:06:24.892822083 +0000 UTC deployed ingress-0.1.0-mcp-2735 openstack-barbican openstack 4 2021-07-09 11:06:25.733684392 +0000 UTC deployed barbican-0.1.0-mcp-3890 openstack-ceph-rgw openstack 4 2021-07-09 11:06:25.045759981 +0000 UTC deployed ceph-rgw-0.1.0-mcp-2735 openstack-cinder openstack 4 2021-07-09 11:06:42.702963544 +0000 UTC deployed cinder-0.1.0-mcp-3890 openstack-designate openstack 4 2021-07-09 11:06:24.400555027 +0000 UTC deployed designate-0.1.0-mcp-3890 openstack-glance openstack 4 2021-07-09 11:06:25.5916904 +0000 UTC deployed glance-0.1.0-mcp-3890 openstack-heat openstack 4 2021-07-09 11:06:25.3998706 +0000 UTC deployed heat-0.1.0-mcp-3890 openstack-horizon openstack 4 2021-07-09 11:06:23.27538297 +0000 UTC deployed horizon-0.1.0-mcp-3890 openstack-iscsi openstack 4 2021-07-09 11:06:37.891858343 +0000 UTC deployed iscsi-0.1.0-mcp-2735 v1.0.0 openstack-keystone openstack 4 2021-07-09 11:06:24.878052272 +0000 UTC deployed keystone-0.1.0-mcp-3890 openstack-libvirt openstack 4 2021-07-09 11:06:38.185312907 +0000 UTC deployed libvirt-0.1.0-mcp-2735 openstack-mariadb openstack 4 2021-07-09 11:06:24.912817378 +0000 UTC deployed mariadb-0.1.0-mcp-2735 openstack-memcached openstack 4 2021-07-09 11:06:24.852840635 +0000 UTC deployed memcached-0.1.0-mcp-2735 openstack-neutron openstack 4 2021-07-09 11:06:58.96398517 +0000 UTC deployed neutron-0.1.0-mcp-3890 openstack-neutron-rabbitmq openstack 4 2021-07-09 11:06:51.454918432 +0000 UTC deployed rabbitmq-0.1.0-mcp-2735 openstack-nova openstack 4 2021-07-09 11:06:44.277976646 +0000 UTC deployed nova-0.1.0-mcp-3890 openstack-octavia openstack 4 2021-07-09 11:06:24.775069513 +0000 UTC deployed octavia-0.1.0-mcp-3890 openstack-openvswitch openstack 4 2021-07-09 11:06:55.271711021 +0000 UTC deployed openvswitch-0.1.0-mcp-2735 openstack-placement openstack 4 2021-07-09 11:06:21.954550107 +0000 UTC deployed placement-0.1.0-mcp-3890 openstack-rabbitmq openstack 4 2021-07-09 11:06:25.431404853 +0000 UTC deployed rabbitmq-0.1.0-mcp-2735 openstack-tempest openstack 2 2021-07-09 11:06:21.330801212 +0000 UTC deployed tempest-0.1.0-mcp-3890
If a Helm release is not in the
DEPLOYEDstate, obtain the details from the output of the following command:helm3 --namespace openstack history <release-name>
To verify the status of a Helm release:
helm3 --namespace openstack status <release-name>
Example of a system response:
NAME: openstack-memcached
LAST DEPLOYED: Fri Jul 9 11:06:24 2021
NAMESPACE: openstack
STATUS: deployed
REVISION: 4
TEST SUITE: None
Debugging the OpenStack Controller¶
OpenStack Controller (Rockoon)
Since MOSK 25.1, the OpenStack Controller has been open-sourced under the name Rockoon and is maintained as an independent open-source project going forward.
As part of this transition, all openstack-controller pods are named
rockoon pods across the MOSK documentation and deployments. This change
does not affect functionality, but this is the reminder for the users to
utilize the new naming for pods and other related artifacts accordingly.
The OpenStack Controller (Rockoon) is running in several containers in the
rockoon-xxxx pod in the osh-system namespace. For the full list
of containers and their roles, refer to OpenStack Controller (Rockoon).
To verify the status of the OpenStack Controller, run:
kubectl -n osh-system get pods
Example of a system response:
NAME READY STATUS RESTARTS AGE
rockoon-5c5965688b-knck5 9/9 Running 0 21h
rockoon-admission-6795d594b4-rp7kf 1/1 Running 0 22h
rockoon-exporter-6f66547b67-pcxhh 1/1 Running 0 22h
To verify the logs for the osdpl container, run:
kubectl -n osh-system logs -f <rockoon-xxxx> -c osdpl
Debugging the OsDpl CR¶
This section includes the ways to mitigate the most common issues with the OsDpl CR. We assume that you have already debugged the Helm releases and OpenStack Controller to rule out possible failures with these components as described in Debugging the Helm releases and Debugging the OpenStack Controller.
osdpl has DEPLOYED=false¶Possible root cause: One or more Helm releases have not been deployed successfully.
To determine if you are affected:
Verify the status of the osdpl object:
kubectl -n openstack get osdpl osh-dev
Example of a system response:
NAME AGE DEPLOYED DRAFT
osh-dev 22h false false
To debug the issue:
Identify the failed release by assessing the
status:childrensection in the OsDpl resource:Get the OsDpl YAML file:
kubectl -n openstack get osdpl osh-dev -o yaml
Analyze the
statusoutput using the detailed description in OpenStackDeploymentStatus custom resource.
For further debugging, refer to Debugging the Helm releases.
Init¶Possible root cause: MOSK uses the Kubernetes entrypoint
init container to resolve dependencies between objects. If the pod is stuck
in Init:0/X, this pod may be waiting for its dependencies.
To debug the issue:
Verify the missing dependencies:
kubectl -n openstack logs -f placement-api-84669d79b5-49drw -c init
Example of a system response:
Entrypoint WARNING: 2020/04/21 11:52:50 entrypoint.go:72: Resolving dependency Job placement-ks-user in namespace openstack failed: Job Job placement-ks-user in namespace openstack is not completed yet .
Entrypoint WARNING: 2020/04/21 11:52:52 entrypoint.go:72: Resolving dependency Job placement-ks-endpoints in namespace openstack failed: Job Job placement-ks-endpoints in namespace openstack is not completed yet .
Possible root cause: some OpenStack services depend on Ceph. These services
include OpenStack Image, OpenStack Compute, and OpenStack Block Storage.
If the Helm releases for these services are not present, the
openstack-ceph-keys secret may be missing in the openstack-ceph-shared
namespace.
To debug the issue:
Verify that the Ceph Controller has created the openstack-ceph-keys
secret in the openstack-ceph-shared namespace:
kubectl -n openstack-ceph-shared get secrets openstack-ceph-keys
Example of a positive system response:
NAME TYPE DATA AGE
openstack-ceph-keys Opaque 7 23h
If the secret is not present, create one manually.
Support dump¶
TechPreview
OpenStack Controller (Rockoon)
Since MOSK 25.1, the OpenStack Controller has been open-sourced under the name Rockoon and is maintained as an independent open-source project going forward.
As part of this transition, all openstack-controller pods are named
rockoon pods across the MOSK documentation and deployments. This change
does not affect functionality, but this is the reminder for the users to
utilize the new naming for pods and other related artifacts accordingly.
Support dump described in this section specifically targets OpenStack components, providing valuable insights for troubleshooting OpenStack-related problems.
To generate a support dump for your MOSK environment,
use the osctl sos report tool present within the rockoon
image.
This section focuses only on the essential capabilities of the tool. For all available parameters, consult osctl sos report --help.
The support dump is modular. Each module is responsible for specific
functionality. To enable or disable specific modules during support
dump creation, use the --collector option. If not specified,
all collectors are used.
Collector |
Description |
|---|---|
|
Collects logs from StackLight by connecting to the OpenSearch API. |
|
Collects data about objects from Kubernetes. |
|
Collects metadata associated with the Compute service (OpenStack Nova) from the OpenStack nodes. This encompasses a wide range of data, including instance details, general libvirt information, and so on. |
|
Collects metadata associated with the Networking service (OpenStack Neutron) from the OpenStack nodes. This encompasses a wide range of data, including Open vSwitch statistics, list of namespaces, IP address statistics in namespaces, Open vSwitch flows, and so on. |
Given the substantial amount of information, you can manage the components
included in a support dump using the mutually exclusive --component
or --all-components options. Within the elastic collector component,
you can specify which loggers to gather logs for. For example,
the --component nova option restricts log collection to pods related
to Nova, which names start with nova-* and libvirt-*.
Another filtering criterion involves specifying the host for which you intend
to collect support information. This can be accomplished through the use of
mutually exclusive --host or --all-hosts options. This feature is
particularly valuable for limiting the volume of data included in
the support dump.
Support dump works in the following modes:
reportGeneric report is created, no specific information such as resource UUID is included. The tool collects as much information as possible.
traceProvides more sophisticated filtering criteria rather than the
reportmode. For example, you can search for specific message patterns in OpenSearch.
Since MOSK 23.2, you can execute the osctl sos
commands directly in the rockoon pod. For example:
kubectl -n osh-system exec -it deployment/rockoon bash
osctl sos --since 1d \
--all-hosts \
--component neutron \
--collector elastic \
--workspace /tmp/ report
To get trace for a specific resource UUID in Neutron for a specific host, use the following command as an example:
kubectl -n osh-system exec -it deployment/rockoon bash
osctl sos --since 1d \
--host kaas-node-fe0734de-20e8-4493-9f7d-52c4f8a8a98c \
--component neutron \
--workspace /workspace/ \
--collector elastic trace --message ".4a055675-89b0-45c2-a3b3-a10dffa07f31."
For older MOSK versions, to start generating support dumps, execute the osctl sos commands from a manually started Docker container on any node of your cluster. For example, to create a generic report for the Neutron component:
docker run -v /home/ubuntu/sosreport/:/workspace -v /root/.kube/config:/root/.kube/config -it mirantis.azurecr.io/openstack/rockoon:1.0.1 bash
osctl sos --since 1d \
--elastic-url http://172.16.37.11:9200 \
--all-hosts \
--component neutron \
--collector elastic \
--workspace /workspace/ report
Note
172.16.37.11 is the IP address of the opensearch-master
StackLight service. To obtain it, run:
kubectl -n stacklight get svc opensearch-master -o jsonpath='{.spec.clusterIP}')
Deploy Tungsten Fabric¶
This section describes how to deploy Tungsten Fabric as a backend for networking for your MOSK environment.
Caution
Before you proceed with the Tungsten Fabric deployment, read through Tungsten Fabric known limitations.
Tungsten Fabric deployment prerequisites¶
Before you proceed with the actual Tungsten Fabric (TF) deployment, verify that your deployment meets the following prerequisites:
Your MOSK OpenStack cluster is deployed as described in Deploy an OpenStack cluster with the Tungsten Fabric backend enabled for Neutron using the following structure:
spec: features: neutron: backend: tungstenfabric
Your MOSK OpenStack cluster uses the correct value of
features:neutron:tunnel_interfacein theopenstackdeploymentobject. The TF Operator will consume this value through the shared secret and use it as a network interface from the underlay network to create encapsulated tunnels with the tenant networks.Considerations for
tunnel_interfacePlan this interface as a dedicated physical interface for TF overlay networks. TF uses
features:neutron:tunnel_interfaceto create thevhost0virtual interface and transfers the IP configuration from thetunnel_interfaceto the virtual one.Do not use
bridgesfrom L2 templates astunnel_interface. Such usage might lead to networking performance degradation and data plane downtime.
The Kubernetes nodes are labeled according to the TF node roles:
Tungsten Fabric (TF) node roles¶ Node role
Description
Kubernetes labels
Minimal count
TF control plane
Hosts the TF control plane services such as
database,messaging,api,svc,config.tfconfig=enabledtfcontrol=enabledtfwebui=enabledtfconfigdb=enabled3
TF analytics
Hosts the TF analytics services.
tfanalytics=enabledtfanalyticsdb=enabled3
TF vRouter
Hosts the TF vRouter module and vRouter Agent.
tfvrouter=enabledVaries
TF vRouter DPDK Technical Preview
Hosts the TF vRouter Agent in DPDK mode.
tfvrouter-dpdk=enabledVaries
Note
TF supports only Kubernetes OpenStack workloads. Therefore, you should label OpenStack compute nodes with the
tfvrouter=enabledlabel.Note
Do not specify the
openvswitch=enabledlabel for the OpenStack deployments with TF as a networking backend.
Deploy Tungsten Fabric¶
Deployment of Tungsten Fabric is managed by the tungstenfabric-operator
Helm resource in a respective ClusterRelease.
To deploy Tungsten Fabric:
Optional. Configure the ASN and encapsulation settings if you need custom values for these parameters. For configuration details, see Autonomous System Number (ASN).
Verify that you have completed all prerequisite steps as described in Tungsten Fabric deployment prerequisites.
Create the
tungstenfabric.yamlfile with the Tungsten Fabric resource configuration. For example:apiVersion: tf.mirantis.com/v2 kind: TFOperator metadata: name: openstack-tf namespace: tf spec: dataStorageClass: tungstenfabric-operator-bind-mounts
Configure the
TFOperatorcustom resource according to the needs of your deployment. For the configuration details, refer to TFOperator custom resource and Tungsten Fabric Operator resources.Trigger the Tungsten Fabric deployment:
kubectl apply -f tungstenfabric.yaml
Verify that Tungsten Fabric has been successfully deployed:
kubectl get pods -n tf
The successfully deployed TF services should appear in the
Runningstatus in the system response.If you have enabled StackLight, enable Tungsten Fabric monitoring by setting
tungstenFabricMonitoring.enabledtotrueas described in StackLight configuration procedure.Since MOSK 23.1,
tungstenFabricMonitoring.enabledis enabled by default during the Tungsten Fabric deployment. Therefore, skip this step.
Advanced Tungsten Fabric configuration (optional)¶
This section includes configuration information for available advanced Mirantis OpenStack for Kubernetes features that include SR-IOV and DPDK with the Neutron Tungsten Fabric backend.
Enable huge pages for OpenStack with Tungsten Fabric¶
Note
The instruction provided in this section applies to both OpenStack with OVS and OpenStack with Tungsten Fabric topologies.
The huge pages OpenStack feature provides essential performance improvements
for applications that are highly memory IO-bound. Huge pages should be enabled
on a per compute node basis. By default, NUMATopologyFilter is enabled.
To activate the feature, you need to enable huge pages on the dedicated bare metal host as described in Enable huge pages in a host profile during the predeployment bare metal configuration.
Note
The multi-size huge pages are not fully supported by Kubernetes versions before 1.19. Therefore, define only one size in kernel parameters.
Enable DPDK for Tungsten Fabric¶
TechPreview
This section describes how to enable DPDK mode for the Tungsten Fabric (TF) vRouter.
To enable DPDK for TF, follow one of the procedures below depending on the API version used:
Install the
vfio-pci(recommended) oruio_pci_genericdriver on the host operating system. For more information about drivers, see Linux Drivers.An example of the
DPDK_UIO_DRIVERconfiguration:spec: services: vRouter: agentDPDK: enabled: true envSettings: dpdk: - name: DPDK_UIO_DRIVER value: vfio-pci
Open the TF Operator custom resource for editing:
kubectl -n tf edit tfoperators.tf.mirantis.com openstack-tf
Enable DPDK:
spec: services: vRouter: agentDPDK: enabled: true
Install the
vfio-pci(recommended) oruio_pci_genericdriver on the host operating system. For more information about drivers, see Linux Drivers.An example of the
DPDK_UIO_DRIVERconfiguration:spec: tf-vrouter: agent-dpdk: enabled: true containers: - name: dpdk env: - name: DPDK_UIO_DRIVER value: vfio-pci
Verify that DPDK NICs are not used on the host operating system.
Note
For use in the Linux user space, DPDK NICs will be bound to specific Linux drivers, required by PMDs. In such a way, bounded NICs are not available for usage by standard Linux network utilities. Therefore, allocate a dedicated NIC(s) for the vRouter deployment in DPDK mode.
Enable huge pages on the host as described in Enable huge pages in a host profile.
Mark the hosts for deployment with DPDK with the
tfvrouter-dpdk=enabledlabel.Open the TF Operator custom resource for editing:
kubectl -n tf edit tfoperators.operator.tf.mirantis.com openstack-tf
Enable DPDK:
spec: controllers: tf-vrouter: agent-dpdk: enabled: true
Enable SR-IOV for Tungsten Fabric¶
This section instructs you on how to enable SR-IOV with the Neutron Tungsten Fabric (TF) backend.
To enable SR-IOV for TF:
Verify that your deployment meets the following requirements:
NICs with the SR-IOV support are installed
SR-IOV and VT-d are enabled in BIOS
Enable IOMMU in the kernel by configuring
intel_iommu=onin the GRUB configuration file. Specify the parameter for compute nodes inBareMetalHostProfilein thegrubConfigsection:spec: grubConfig: defaultGrubOptions: - 'GRUB_CMDLINE_LINUX="$GRUB_CMDLINE_LINUX intel_iommu=on"'
Enable SR-IOV in the
OpenStackDeploymentCR through the node-specific overrides settings. For example:spec: nodes: <NODE-LABEL>::<NODE-LABEL-VALUE>: features: neutron: sriov: enabled: true nics: - device: enp10s0f1 num_vfs: 7 physnet: tenant
Warning
After the
OpenStackDeploymentCR modification, the TF Operator generates a separate vRouter DaemonSet with specified settings. Thetf-vrouter-agent-<XXXXX>pods will be automatically restarted on the affected nodes causing the network services interruption on virtual machines running on these hosts.Optional. To modify a vRouter DaemonSet according to the SR-IOV definition in the
OpenStackDevelopmentCR, add vRouter custom specs to the TF Operator CR with the node label specified in theOpenStackDeploymentCR. For example:spec: nodes: sriov: labels: name: <NODE-LABEL> value: <NODE-LABEL-VALUE> nodeVRouter: enabled: true envSettings: agent: - name: VROUTER_GATEWAY value: <VROUTER-GATEWAY-IP>
spec: controllers: tf-vrouter: agent: customSpecs: - name: sriov label: name: <NODE-LABEL> value: <NODE-LABEL-VALUE> containers: - name: agent env: - name: <VROUTER-GATEWAY> value: <VROUTER-GATEWAY-IP>
Configure multiple Contrail API workers¶
TechPreview
Tungsten Fabric MOSK deployments use six workers of the
contrail-api service by default. This section instructs you on how to
change the default configuration if needed.
To configure the number of Contrail API workers on a TF deployment:
Specify the required number of workers in the
TFOperatorcustom resource:spec: features: config: configApiWorkerCount: 7
spec: controllers: tf-config: api: containers: - env: - name: CONFIG_API_WORKER_COUNT value: "7" name: api
Wait until all
tf-config-*pods are restarted.Verify the number of workers inside the running API container:
kubectl -n tf exec -ti tf-config-rclzq -c api -- ps aux --width 500 kubectl -n tf exec -ti tf-config-rclzq -c api -- ls /etc/contrail/
Verify that the ps output lists one API process with PID
"1"and the number of workers set in theTFOperatorcustom resource.In
/etc/contrail/, verify that the number of configuration filescontrail-api-X.confmatches the number of workers set in theTFOperatorcustom resource.
Disable Tungsten Fabric analytics services¶
Available since MOSK 23.3 TechPreview
By default, analytics services are part of basic setups for Tungsten Fabric deployments. To obtain a more lightweight setup, you can disable these services through the custom resource of the Tungsten Fabric Operator.
Warning
Disabling of the Tungsten Fabric analytics services requires
restart of the data plane services for existing environments and must be
planned in advance. While calculating the maintenance window for this
operation, take into account the deletion of the analytics DaemonSets
and automatic restart of the tf-config, tf-control, and
tf-webui pods.
To disable Tungsten Fabric analytics services:
Open the
TFOperatorcustom resource for editing:kubectl -n tf edit tfoperators.tf.mirantis.com openstack-tf
kubectl -n tf edit tfoperators.operator.tf.mirantis.com openstack-tf
Disable Tungsten Fabric analytics services in the
TFOperatorcustom resource:spec: services: analytics: enabled: false
spec: settings: disableTFAnalytics: true
Clean up the Kubernetes resources. To free up the space that has been used by Cassandra, ZooKeeper, and Kafka analytics storage, manually delete the related PVC:
kubectl -n tf delete pvc -l app=cassandracluster,cassandracluster=tf-cassandra-analytics kubectl -n tf delete pvc -l app=tf-zookeeper-nal kubectl -n tf delete pvc -l app=tf-kafka
Remove the
tfanalytics=enabledandtfanalyticsdb=enabledlabels from nodes, as they are not required by the Tungsten Fabric Operator anymore.Manually restart the vRouter pods:
Note
To avoid network disruption, restart the vRouter pods in chunks.
kubectl -n tf get pod -l app=tf-vrouter-agent kubectl -n tf delete pod <POD_NAME>
Delete terminated nodes from the Tungsten Fabric configuration through the Tungsten Fabric web UI:
Caution
With disabled Tungsten Fabric analytics, the Tungsten Fabric web UI may not work properly.
Log in to the Tungsten Fabric web UI.
On Configure > Infrastructure > Nodes > Analytics Nodes, delete all terminated analytics nodes.
On Configure > Infrastructure > Nodes > Database Analytics Nodes, delete all terminated database analytics nodes.
Depending on the MOSK version, proceed accordingly:
Disable monitoring of the Tungsten Fabric analytics services in StackLight by setting the following parameter in StackLight values of the
Clusterobject tofalse:tungstenFabricMonitoring: analyticsEnabled: false
When done, the monitoring of the Tungsten Fabric analytics components will become disabled and Kafka alerts along with the Kafka dashboard will disappear from StackLight.
The
tungstenFabricMonitoring.analyticsEnabledsetting is automatically configured based on the state of the Tungsten Fabric analytics services, which are enabled or disabled.However, you can still override this setting. If set manually, the configuration overrides the default behavior and does not reflect the actual state of Tungsten Fabric analytics.
Now, with the Tungsten Fabric analytics services successfully disabled, you have optimized resource utilization and system performance. While these services are deactivated, related alerts may still be present in StackLight. However, do not consider such alerts as indicative of the actual status of the analytics services.
Troubleshoot the Tungsten Fabric deployment¶
This section provides the general debugging instructions for your Tungsten Fabric (TF) on Kubernetes deployment.
Enable debug logs for the Tungsten Fabric services¶
To enable debug logging for the Tungsten Fabric (TF) services:
Open the TF custom resource for modification:
kubectl -n tf edit tfoperators.tf.mirantis.com openstack-tf
Set the
logLevelvariable to theSYS_DEBUGvalue for the required TF service. For example, for theconfig-apiservice:spec: services: config: tf: logging: api: logLevel: SYS_DEBUG
Open the TF custom resource for modification:
kubectl -n tf edit tfoperators.operator.tf.mirantis.com openstack-tf
Set the
LOG_LEVELvariable to theSYS_DEBUGvalue for the required TF service. For example, for theconfig-apiservice:spec: controllers: tf-config: api: containers: - name: api env: - name: LOG_LEVEL value: SYS_DEBUG
Warning
After the TF custom resource modification, the pods
related to the affected services will be restarted. This rule does not
apply to the tf-vrouter-agent-<XXXXX> pods as their update strategy
differs. Therefore, if you enable the debug logging for the services in
a tf-vrouter-agent-<XXXXX> pod, restart this pod manually after you
modify the custom resource.
Troubleshoot access to the Tungsten Fabric web UI¶
If you cannot access the Tungsten Fabric (TF) web UI service, verify that the FQDN of the TF web UI is resolvable on your PC by running one of the following commands:
host tf-webui.it.just.works
# or
ping tf-webui.it.just.works
# or
dig host tf-webui.it.just.works
All commands above should resolve the web UI domain name to the IP address
that should match the EXTERNAL-IPs subnet dedicated to Kubernetes.
If the TF web UI domain name has not been resolved to the IP address, your PC
is using a different DNS or the DNS does not contain the record for the TF web
UI service. To resolve the issue, define the IP address of the Ingress service
from the openstack namespace of Kubernetes in the hosts file of your
machine. To obtain the Ingress IP address:
kubectl -n openstack get svc ingress -o custom-columns=HOSTS:.status.loadBalancer.ingress[*].ip
If the web UI domain name is resolvable but you still cannot access the service, verify the connectivity to the cluster.
Disable TX offloading on NICs used by vRouter¶
In the following cases, a TCP-based service may not work on VMs:
If the setup has nested VMs.
If VMs are running in the ESXi hypervisor.
If the Network Interface Cards (NICs) do not support the IP checksum calculation and generate an incorrect checksum. For example, the Broadcom Corporation NetXtreme BCM5719 Gigabit Ethernet PCIe NIC cards.
To resolve the issue, disable the transmit (TX) offloading on all OpenStack compute nodes for the affected NIC used by the vRouter as described below.
To identify the issue:
Verify whether ping is working between VMs on different hypervisor hosts and the TCP services are working.
Run the following command for the vRouter Agent and verify whether the output includes the number of
Checksum errors:kubectl -n tf exec tf-vrouter-agent-XXXXX -c agent -- dropstats
Run the following command and verify if the output includes the
cksum incorrectentries:kubectl -n tf exec tf-vrouter-agent-XXXXX -c agent -- tcpdump -i <tunnel interface> -v -nn | grep -i incorrect
Example of system response:
tcpdump: listening on <tunnel interface>, link-type EN10MB (Ethernet), capture size 262144 bytes <src ip.port> > <dst ip.port>: Flags [S.], cksum 0x43bf (incorrect -> 0xb8dc), \ seq 1901889431, ack 1081063811, win 28960, options [mss 1420,sackOK,\ TS val 456361578 ecr 41455995,nop,wscale 7], length 0 <src ip.port> > <dst ip.port>: Flags [S.], cksum 0x43bf (incorrect -> 0xb8dc), \ seq 1901889183, ack 1081063811, win 28960, options [mss 1420,sackOK,\ TS val 456361826 ecr 41455995,nop,wscale 7], length 0 <src ip.port> > <dst ip.port>: Flags [S.], cksum 0x43bf (incorrect -> 0xb8dc), \ seq 1901888933, ack 1081063811, win 28960, options [mss 1420,sackOK,\ TS val 456362076 ecr 41455995,nop,wscale 7], length 0
Run the following command for the vRouter Agent container and verify whether the output includes the information about a drop for an unknown reason:
kubectl -n tf exec tf-vrouter-agent-XXXXX -c agent -- flow -l
To disable the TX offloading on NICs used by vRouter:
Open the
TFOperatorcustom resource (CR) for editing:kubectl -n tf edit tfoperators.operator.tf.mirantis.com openstack-tf
Specify the
DISABLE_TX_OFFLOADvariable with the"YES"value for the vRouter Agent container:spec: features: vRouter: disableTXOffload: true
spec: controllers: tf-vrouter: agent: containers: - name: agent env: - name: DISABLE_TX_OFFLOAD value: "YES"
Warning
Once you modify the
TFOperatorCR, thetf-vrouter-agent-<XXXXX>pods will not restart automatically because they use the OnDelete update strategy. Restart such pods manually, considering that the vRouter pods restart causes network services interruption for the VMs hosted on the affected nodes.To disable TX offloading on a specific subset of nodes, use custom vRouter settings. For details, see Custom vRouter settings.
Warning
Once you add a new
CustomSpec, a new daemon set will be generated and thetf-vrouter-agent-<XXXXX>pods will be automatically restarted. The vRouter pods restart causes network services interruption for VMs hosted on the affected node. Therefore, plan this procedure accordingly.
See also
Operations Guide¶
This guide outlines the post-deployment Day-2 operations for a Mirantis OpenStack for Kubernetes environment. It describes how to configure and manage the MOSK components, perform different types of cloud verification, and enable additional features depending on your cloud needs. The guide also contains day-to-day maintenance procedures such as how to back up and restore, update and upgrade, or troubleshoot your MOSK cluster.
Cluster update¶
Updating a MOSK cluster ensures that the system remains secure, efficient, and up-to-date with the latest features and performance improvements, as well as receives fixes for the known CVEs. This section provides comprehensive details and step-by-step procedures to guide you through the process of updating your cluster.
Update to a major version¶
This section describes the workflow you as a cloud operator need to follow to correctly update your Mirantis OpenStack for Kubernetes (MOSK) cluster to a major release version.
Note
The hereby guide applies to the clusters running MOSK of version 23.1 and above. In case you have an older version and looking to update, please contact Mirantis support to get intructions valid for your cluster.
The instructions below are generic and apply to any MOSK cluster regardless of its configuration specifics. However, every major release may have its own update peculiarities. Therefore, to accurately plan and successfully perform an update, in addition to the hereby document, read the update-related section in the Release Notes of the target MOSK version.
Depending on the payload of a target release, the update mechanism can perform the changes on different levels of the stack, from the configuration of the host operating system to the code of OpenStack itself. The update mechanism is designed to avoid the impact on the workloads and cloud users as much as possible. The life-cycle management logic minimizes the downtime for the cloud API by means of smart management of the cluster components under the hood and only requests your involvement when a human decision is required to proceed.
Though the update mechanism may change the internal components of the cluster, it will always preserve the major versions of OpenStack, that is, the APIs that cloud users and workloads deal with. After the cluster is successfully updated, you can initiate a separate upgrade procedure to obtain the latest supported OpenStack version.
Before you begin¶
Before starting an update, we recommend that you closely peruse the Release Compatibility Matrix document and Release notes of the target release, as well as thoroughly plan maintenance windows for each update phase depending on the configurational of your cluster.
Read carefully Release Compatibility Matrix and Release Notes of the target MOSK version paying particular attention to the following:
Current Mirantis Container Cloud software version and the need to first update to the latest cluster release version
Update notes provided in the Release notes for the target MOSK version
New product features that will get enabled in your cloud by default
New product features that may have already been configured in your cloud as customizations and now need to be properly re-enabled to be eligible for further support
Any changes in the behavior of the product features enabled in your cloud
List of the addressed and known issues in the target MOSK version
Warning
If your cloud configuration is known to have any custom configuration that was not explicitly approved by Mirantis, make sure to bring this up with your dedicated Mirantis representative before proceeding with the update. Mirantis cannot guarantee the safe updating of a customized cloud.
Depending on the payload brought by a particular target release, a generic cluster update includes from three to six major phases.
The first three phases are present in any update. They focus on the containerized components of the software stack and have minimal impact on the cloud users and workloads.
The remaining phases are only present if any changes need to be made to the foundation layers: the underlay Kubernetes cluster and host operating system. For the changes to take effect, you may need to reboot the cluster nodes. This procedure imposes a severe impact on cloud workloads and, therefore, needs to be thoroughly planned across several sequential maintenance windows.
Important
To effectively plan a cluster update, keep in mind the architecture of your specific cloud. Depending on the selected design, the components of a MOSK cluster may have different distribution across the nodes (physical servers) comprising the underlay bare metal Kubernetes cluster. The more components are collocated on a single node, the harder is the impact on the functions of the cloud when the changes are applied.
The tables below will help you to plan your cluster update and include the following information for each mandatory and additional update phase:
- What happens during the phase
Includes the phase milestones. The nature of changes that are going to be applied is important to understand in order to estimate the exact impact the update is going to have on your cluster.
Consult the Update notes section of the target MOSK release for the detailed information about the changes it brings and the impact these changes are going to imply when getting applied to your cluster.
- Impact
Describes possible impact on cloud users and workloads.
The provided information about the impact represents the worst-case scenario in the cluster architectures that imply a combination of several roles on the same physical servers, such as hyper-converged compute nodes and clusters with a compact control plane.
The impact estimation presumes that your cluster uses one of the standard architectures provided by the product and follows Mirantis design guidelines.
- Time to complete
Provides a rough estimation of the time required to complete the phase.
The estimates for a phase timeline presume that your cluster uses one of the standard architectures provided by the product and follows Mirantis design guidelines.
Warning
During the update, try to prevent users from performing write operations on the cloud resources. Any intensive manipulations may lead to workload corruption.
Phase 1: Life-cycle management modules update
Important
This phase is mandatory. It is always present in the update flow regardless of the contents of the target release.
What happens during the phase |
New versions of OpenStack, Tungsten Fabric, and Ceph controllers downloaded and installed. OpenStack and Tungsten Fabric images precached. |
|---|---|
Impact |
None |
Time to complete |
Depending on the quality of the Internet connectivity, up to 45 minutes. |
Phase 2: OpenStack and Tungsten Fabric components update
Important
This phase is mandatory. It is always present in the update flow regardless of the contents of the target release.
What happens during the phase |
New versions of OpenStack and Tungsten Fabric container images downloaded, services restarted sequentially. |
|---|---|
Impact |
|
Time to complete |
|
Phase 3: Ceph cluster update and upgrade
Important
This phase is mandatory. It is always present in the update flow regardless of the contents of the target release.
What happens during the phase |
New versions of Ceph components downloaded, services restarted. If applicable, Ceph switched to the latest major version. |
|---|---|
Impact |
Workloads may experience IO performance degradation for the virtual storage devices backed by Ceph. |
Time to complete |
The update of a Ceph cluster with 30 storage nodes can take up to 35 minutes. Additionally, 15 minutes are required for the major Ceph version upgrade, if any. |
Phase 4a: Host operating system update on Kubernetes master nodes
Important
This phase is optional. The presense of this phase in the update flow depends on the contents of the target release.
What happens during the phase |
New system packages downloaded and installed on the host operating system, other major changes get applied. |
|---|---|
Impact |
None |
Time to complete |
The nodes are updated sequentially. Up to 15 minutes per node. |
Phase 4b: Kubernetes components update on Kubernetes master nodes
Important
This phase is optional. The presense of this phase in the update flow depends on the contents of the target release.
What happens during the phase |
New versions of Kubernetes control plane components downloaded and installed. |
|---|---|
Impact |
For clusters with the compact control plane, some of the running cloud operations may fail over the course of the phase due to minor unavailability of the cloud API. For the compact control plane with gateway nodes collocated (Open vSwitch networking backend), workloads can experience temporary loss of the North-South connectivity. The downtime depends on the type of virtual routers in use. |
Time to complete |
Up to 40 minutes total |
Phases 5a and 5b: Host operating system and Kubernetes cluster
update on Kubernetes worker nodes
Important
Both phases, 5a and 5b, are applied together, either node by node (default) or to several nodes in parallel. The parallel updating is available since 23.1.
Take this into consideration when estimating the impact and planning the maintenance window.
What happens during the phase |
During the host operating system update:
During the Kubernetes cluster update:
|
|---|---|
Impact |
|
Time to complete |
By default, the nodes are updated sequentially as follows:
For MOSK 23.1 to 23.2 and newer releases, you can reduce update time by enabling parallel node update. The procedure is described further in the Enable parallel update of Kubernetes worker nodes subsection. |
Phase 6: Cluster nodes reboot
Important
This phase is optional. The presense of this phase in the update flow depends on the contents of the target release.
Important
An update to a newer MOSK version may require reboot of the cluster nodes for changes to take effect. Although, you can decide when to restart each particular node, an update can not be considered complete until all of the nodes get restarted.
To determine whether the reboot is required, consult the Step 4. Reboot the nodes with optional instance migration section.
What happens during the phase |
|
|---|---|
Impact |
|
Time to complete |
|
Step 1. Verify that the Container Cloud management cluster is up-to-date¶
MOSK relies on Mirantis Container Cloud to manage the underlying software stack for a cluster, as well as to deliver updates for all the components.
Since every MOSK release is tightly coupled with a Container Cloud release, a MOSK cluster update becomes possible once the management cluster is known to run the latest Container Cloud version. The management cluster periodically verifies public Mirantis repositories and updates itself automatically when a newer version becomes available. Having any of the managed clusters, including MOSK, running outdated Container Cloud version will prevent the management cluster from automatic self-update.
To identify the current version of the Container Cloud software your management cluster is running, refer to the Container Cloud web UI. You can also verify your management cluster status using CLI as described in Verify the management cluster status before MOSK update.
Step 2. Initiate MOSK cluster update¶
During an update of a MOSK cluster, numerous alerts may be seen in StackLight. This is expected behavior. Therefore, ignore or temporarily mute the alerts as described in Silence alerts.
Caution
During update, the false positive CalicoDataplaneFailuresHigh
alert may be firing. Disregard this alert, which will disappear once update
succeeds.
The observed behavior is typical for calico-node during upgrades,
as workload changes occur frequently. Consequently, there is a possibility
of temporary desynchronization in the Calico dataplane. This can
occasionally result in throttling when applying workload changes to the
Calico dataplane.
If you update MOSK to 23.1, verify that the
KaaSCephCluster custom resource does not contain the following entries.
If they exist, remove them.
In the
spec.cephClusterSpecsection, theexternalsection.In the
spec.cephClusterSpec.rookConfigsection, thems_crc_dataorms crc dataconfiguration key. After you remove the key, wait forrook-ceph-monpods to restart on the MOSK cluster.
Optional. Starting from MOSK 23.1 to 23.2 update, you can enable and configure parallel node update to reduce update time and minimize downtime:
To enable parallel update of Kubernetes worker nodes, set the
spec.providerSpec.value.maxWorkerUpgradeCountconfiguration parameter in the Mirantis Container Cloud management cluster as described in conf-upd-count.Consider the specifics of handling of parallel node updates by OpenStack, Ceph, and Tungsten Fabric Controllers to properly plan the maintenance window. For handling details and possible configuration, refer to Parallelizing node update operations.
TechPreview
Optional. Starting from MOSK 24.3, you can enable automatic node reboot of an update group, which contains a set of controller or worker machines. This option applies when a Cluster release update requires node reboot, for example, when kernel version update is available in the target Cluster release. The option reduces manual intervention and overall downtime during cluster update.
To enable automatic node reboot in an update group, set
spec.rebootIfUpdateRequires in the required UpdateGroup object. For
details, see UpdateGroup resource.
Caution
During a distribution upgrade, machines are always rebooted,
overriding rebootIfUpdateRequires: false.
Log in to the Container Cloud web UI with the
m:kaas:namespace@operatororm:kaas:namespace@writerpermissions.Switch to the required project using the Switch Project action icon located on top of the main left-side navigation panel.
In the Clusters tab, find the managed MOSK cluster.
Click the More action icon to see whether a new release is available. If that is the case, click Update cluster.
In the Release Update window, select the required Cluster release to update your managed cluster to.
The Description section contains the list of components versions to be installed with a new Cluster release.
Click Update.
Before the cluster update starts, Container Cloud performs a backup of MKE and Docker Swarm. The backup directory is located under:
/srv/backup/swarmon every Container Cloud node for Docker Swarm/srv/backup/ucpon one of the controller nodes for MKE
To view the update status, verify the cluster status on the Clusters page. Once the orange blinking dot near the cluster name disappears, the update is complete.
Step 3. Watch the cluster update¶
To view the update status through the Container Cloud web UI, navigate to the Clusters page. Once the orange blinking dot next to the cluster name disappears, the cluster update is complete.
Also, you can see the general status of each node during the update on the Container Cloud cluster view page.
The whole update process is controlled by lcm-controller, which runs in
the kaas namespace of the Container Cloud management cluster. Follow
its logs to watch the progress of the update, discover, and debug any issues.
The lcmclusterstate and lcmmachines objects in the mos namespace
of the Container Cloud management cluster provide detailed information about
the current phase of the update process in the context of the managed cluster
overall as well as specific nodes.
The lcmmachine object being in the Ready state indicates that a node
has been successfully updated.
To display the detailed view of the cluster update state, run:
kubectl -n child-ns get lcmclusterstates -o wide
Example system response:
NAME CLUSTERNAME TYPE ARG VALUE ACTUALVALUE ATTEMPT MESSAGE
cd-cz7506-child-cl-storage-worker-noefi-rgxhk child-cl cordon-drain cz7506-child-cl-storage-worker-noefi-rgxhk true 0 Error: following NodeWorkloadLocks are still active - ceph: UpdatingController,openstack: InProgress
sd-cz7506-child-cl-storage-worker-noefi-rgxhk child-cl swarm-drain cz7506-child-cl-storage-worker-noefi-rgxhk true 0 Error: waiting for kubernetes node kaas-node-5222a92f-5523-457c-8c69-b7aa0ffc235c to be drained first
To display the detailed view of the nodes update state, run:
kubectl -n child-ns get lcmmachines
Example system response:
NAME CLUSTERNAME TYPE STATE
cz5018-child-cl-storage-worker-noefi-dzttw child-cl worker Prepare
cz5019-child-cl-storage-worker-noefi-vxcm9 child-cl worker Prepare
cz7500-child-cl-control-storage-worker-noefi-nkk9t child-cl control Ready
cz7501-child-cl-control-storage-worker-noefi-7pcft child-cl control Ready
cz7502-child-cl-control-storage-worker-noefi-c7k6f child-cl control Ready
cz7503-child-cl-storage-worker-noefi-5lvd7 child-cl worker Prepare
cz7505-child-cl-storage-worker-noefi-jh4mc child-cl worker Prepare
cz7506-child-cl-storage-worker-noefi-rgxhk child-cl worker Prepare
Step 4. Reboot the nodes with optional instance migration¶
Depending on the target release content, you may need to reboot the cluster nodes for the changes to take effect. Running a MOSK cluster in a semi-updated state for an extended period may result in unpredictable behavior of the cloud and impact users and workloads. Therefore, when it is required, you need to reboot the cluster nodes as soon as possible to avoid potential risks.
Note
If you enabled rebootIfUpdateRequires as described in
Enable automatic node reboot in update groups, nodes will be automatically rebooted in update
groups during a Cluster release update that requires a reboot, for example,
when kernel version update is available in the target Cluster release.
For a distribution upgrade, continue reading the following subsections.
Verify the YAML definitions of the LCMMachine and Machine objects.
The node must be rebooted if the rebootRequired flag is set to true.
In addition, objects explicitly specify the reason for rebooting. For example:
The
LCMMachineobject of the node that requires rebooting:... status: hostInfo: rebootRequired: true rebootReason: "linux-image-5.13.0-51-generic"
The
Machineobject of the node that does not require rebooting:... status: ... providerStatus: ... reboot: reason: "" required: false status: Ready
Since MOSK 23.1, you can also use the Mirantis Container Cloud web UI to identify the nodes requiring reboot:
In the Clusters tab, click the required cluster name. The page with Machines opens.
Hover over the status of every machine. A machine to reboot contains the Reboot > The machine requires a reboot notification in the Status tooltip.
Restarting the cluster causes downtime of the cloud services running on the nodes. While the MOSK control plane is built for high availability and can tolerate temporary loss of at least 1/3 of services without a significant impact on user experience, rebooting nodes that host the elements of cloud data plane, such as network gateway nodes and compute nodes, has a detrimental effect on the cloud workloads, if not performed gracefully.
To configure the instance migration policy:
Edit the target compute node resource. For example:
kubectl edit node kaas-node-03ab613d-cf79-4830-ac70-ed735453481a
To mitigate the potential impact on the cloud workloads, define the migration mode and the number of attempts the OpenStack Controller should make to migrate a single instance running on it:
Instance migration configuration for hosts¶ Node annotation
Default
Description
instance_migration_modeliveDefines the instance migration mode for the host. The list of available options include:
live: the OpenStack Controller live migrates instances automatically. The update mechanism tries to move the memory and local storage of all instances on the node to another node without interrupting before applying any changes to the node. By default, the update mechanism makes three attempts to migrate each instance before falling back to themanualmode. 0manual: the OpenStack Controller waits for the Operator to migrate instances from the host. When it is time to update the host, the update mechanism asks you to manually migrate the instances and proceeds only once you confirm the node is safe to update. 1skip: the OpenStack Controller skips the instance check on the node and reboots it.
instance_migration_attempts3Defines the number of times the OpenStack Controller attempts to live-migrate a single instance before falling back to the
manualmode.- 0
Success of live migration depends on many factors including the selected vCPU type and model, the amount of data that needs to be transferred, the intensity of the disk IO and memory writes, the type of the local storage, and others. Instances using the following product features are known to have issues with live migration:
LVM-based ephemeral storage with and without encryption
Encrypted block storage volumes
CPU and NUMA node pinning
- 1
For the clouds relying on the converged LVM with iSCSI block storage that offer persistent volumes in a remote edge sub-region, it is important to keep in mind that applying a major change to a compute node may impact not only the instances running on this node but also the instances attached to the LVM devices hosted there. Mirantis recommends that in such environments you perform the update procedure in the
manualmode with mitigation measures taken by the Operator for each compute node. Otherwise, all the instances that have LVM with iSCSI volumes attached would need reboot to restore connectivity.
Configuration example that sets the instance migration mode to
liveand the number of attempts to live-migrate to5:apiVersion: v1 kind: Node metadata: name: kaas-node-03ab613d-cf79-4830-ac70-ed735453481a selfLink: /api/v1/nodes/kaas-node-03ab613d-cf79-4830-ac70-ed735453481a uid: 54be5139-aba7-47e7-92bf-5575773a12a6 resourceVersion: '299734609' creationTimestamp: '2021-03-24T16:03:11Z' labels: ... openstack-compute-node: enabled openvswitch: enabled annotations: openstack.lcm.mirantis.com/instance_migration_mode: "live" openstack.lcm.mirantis.com/instance_migration_attempts: "5" ...
If needed, as a cloud user, mark the instances that require individual handling during instance migration using the
openstack.lcm.mirantis.com:maintenance_action=<ACTION-TAG>server tag. For details, refer to Configure per-instance migration mode.
See also
Since MOSK 23.1, you can reboot several cluster nodes in one go by using the Graceful reboot mechanism provided by Mirantis Container Cloud. The mechanism restarts the selected nodes one by one, honoring the instance migration policies.
For older versions of MOSK, you need to reboot each node manually as follows:
For each node in the cluster:
If
manualinstance migration policy is configured for the node, perform manual migration once the node is ready to reboot (see below).
When a node that has a manual instance migration policy is ready to be
restarted, the life-cycle management mechanism notifies you about that by
creating a NodeMaintenanceRequest object for the node and setting the
active status attribute for the corresponding NodeWorkloadLock object.
Note
Verify the status:errorMessage attribute before proceeding.
To view the NodeWorkloadLock objects details for a specific node, run:
kubectl get nodeworkloadlocks <NODE-NAME> -o yaml
Example system response:
apiVersion: lcm.mirantis.com/v1alpha1
kind: NodeWorkloadLock
metadata:
annotations:
inner_state: active
creationTimestamp: "2022-02-04T13:24:48Z"
generation: 1
name: openstack-kaas-node-b2a55089-5b03-4698-9879-8756e2e81df5
resourceVersion: "173934"
uid: 0cb4428f-dd0d-401d-9d5e-e9e97e077422
spec:
controllerName: openstack
nodeName: kaas-node-b2a55089-5b03-4698-9879-8756e2e81df5
status:
errorMessage: 2022-02-04 14:43:52.674125 Some servers ['0ab4dd8f-ef0d-401d-9d5e-e9e97e077422'] are still present on host kaas-node-b2a55089-5b03-4698-9879-8756e2e81df5.
Waiting unless all of them are migrated manually or instance_migration_mode is set to 'skip'
release: 8.5.0-rc+22.1
state: active
Note
For MOSK compute nodes, you need to manually shut down all instances running on it, or perform cold or live migration of the instances.
After the update¶
Once your MOSK cluster update is complete, proceed with the following:
Perform the post-update steps recommended in the update notes of the target release if any.
Use the standard configuration mechanisms to re-enable the new product features that could previously exist in your cloud as a custom configuration.
To ensure the cluster operability, execute a set of smoke tests as described in Run Tempest tests.
Optional. Proceed with the upgrade of OpenStack.
If necessary, expire alert silences in StackLight as described in Silence alerts.
What to do if the update hangs or fails¶
If an update phase takes significantly longer than expected according to the tables included in Plan the cluster update, you should consider the update process hung.
If you observe errors that are not described explicitly in the documentation, immediately contact Mirantis support.
To see any issues that might have occurred during the update, verify the logs
of the lcm-controller pods in the kaas namespace of the Container Cloud
management cluster.
To troubleshoot the update that involves the operating system upgrade with host reboot, refer to Troubleshoot an operating system upgrade with host restart.
Container Cloud and MOSK life-cycle management mechanism does not provide a way to perform a cluster-wide rollback of an update.
Update to a patch version¶
Patch releases aim to significantly shorten the cycle of CVE fixes delivery onto your MOSK deployments to help you avoid cyber threats and data breaches.
Your management bare-metal cluster obtains patch releases automatically the same way as major releases. A new patch MOSK release version becomes available through the Container Cloud web UI after the automatic upgrade of the management cluster.
It is not possible to update between the patch releases that belong to different release series in one go. For example, you can update from MOSK 23.1.1 to 23.1.2, but you cannot immediately update from MOSK 23.1.x to 23.2.x because you need to update to the major MOSK 23.2 release first.
Caution
If you delay the Container Cloud upgrade and schedule it at a later time as described in Schedule Mirantis Container Cloud updates, make sure to schedule a longer maintenance window as the upgrade queue can include several patch releases along with the major release upgrade.
Pre-update actions¶
Read the Update notes part of the target MOSK release notes to understand the changes it brings and the impact these changes are going to have on your cloud users and workloads.
The application of the patch releases may not require the cluster nodes reboot. Though, your cluster can contain nodes that require reboot after the last update to a major release, and this requirement will remain after update to any of the following patch releases. Therefore, Mirantis strongly recommends that you determine if there are such nodes in your cluster before you update to the next patch release and reboot them if any, as described in Step 4. Reboot the nodes with optional instance migration.
For some MOSK versions, applying a patch release may require restart of the containers that host the elements of the cloud data plane. In case of Open vSwitch-based clusters, this may result in up to 5 minute downtime of workload network connectivity for each compute node.
For MOSK prior to 24.1 series, you can determine whether applying a patch release is going to require the restart of the data plane by consulting the Release artifacts part of the release notes of the current and target MOSK releases. The data plane restart will only happen if there are new versions of the container images related to the data plane.
It is possible to avoid the downtime for the cloud data by explicitly pinning the image versions of the following components:
Open vSwitch
Kubernetes entrypoint
However, pinning these images will result in the cloud data plane not receiving any security or bugfixes during the update.
To pin the images:
Depending on the proxy configuration, the image base URL differs. To obtain the list of currently used images on the cluster, run:
kubectl -n openstack get ds openvswitch-openvswitch-vswitchd-default -o yaml |grep "image:" | sort -u
Example of system response:
image: mirantis.azurecr.io/general/openvswitch:2.13-focal-20230211095312 image: mirantis.azurecr.io/openstack/extra/kubernetes-entrypoint:v1.0.1-48d1e8a-20220919122849
Add the
openvswitchandkubernetes-entrypointimages used on your cluster:Create a ConfigMap in the
openstacknamespace with the following content, replacing<OPENSTACKDEPLOYMENT-NAME>with the name of yourOpenStackDeploymentcustom resource:apiVersion: v1 kind: ConfigMap metadata: labels: penstack.lcm.mirantis.com/watch: "true" name: <OPENSTACKDEPLOYMENT-NAME>-artifacts namespace: openstack data: caracal: | dep_check: <KUBERNETES-ENTRYPOINT-IMAGE-URL> openvswitch_db_server: <OPENVSWITCH-IMAGE-URL> openvswitch_vswitchd: <OPENVSWITCH-IMAGE-URL>
Edit the
OpenStackDeploymentcustom resoruce as follows:spec: services: networking: openvswitch: values: images: tags: dep_check: <KUBERNETES-ENTRYPOINT-IMAGE-URL> openvswitch_db_server: <OPENVSWITCH-IMAGE-URL> openvswitch_vswitchd: <OPENVSWITCH-IMAGE-URL>
For example:
spec: services: networking: openvswitch: values: images: tags: dep_check: mirantis.azurecr.io/openstack/extra/kubernetes-entrypoint:v1.0.1-48d1e8a-20220919122849 openvswitch_db_server: mirantis.azurecr.io/general/openvswitch:2.13-focal-20230211095312 openvswitch_vswitchd: mirantis.azurecr.io/general/openvswitch:2.13-focal-20230211095312
Update a patch Cluster release of a MOSK cluster¶
Log in to the Container Cloud web UI with the
m:kaas:namespace@operatororm:kaas:namespace@writerpermissions.Switch to the required project using the Switch Project action icon located on top of the main left-side navigation panel.
In the Clusters tab, click Upgrade next to the More action icon located in the last column for each cluster where available.
Note
If Upgrade is greyed out, the cluster is in maintenance mode that must be disabled before you can proceed with cluster update. For details, see Disable maintenance mode on a cluster and machine.
If Upgrade does not display, your cluster is up-to-date.
In the Release update window, select the required patch Cluster release to update your managed cluster to.
The release notes for patch Cluster releases are available in Container Cloud Release Notes: Patch releases.
Click Update. To monitor update readiness, refer to Verify cluster status.
Note
Since Container Cloud 2.26.1 (patch Cluster releases 17.1.1 and 16.1.1), the update of Ubuntu packages with kernel minor version update may apply in certain releases.
In this case, cordon-drain and reboot of machines does not apply automatically, and all machines have the Reboot is required notification after the cluster update. You can manually handle the reboot of machines during a convenient maintenance window as described in Perform a graceful reboot of a cluster.
Verify the management cluster status before MOSK update¶
Before you start updating your managed clusters, Mirantis recommends verifying that the associated management cluster is upgraded successfully.
To verify that the management cluster is upgraded successfully:
Using
kubeconfigof the management cluster, verify the Cluster release version of the management cluster machines:for i in $(kubectl get lcmmachines | awk '{print $1}' | sed '1d'); do echo $i; kubectl get lcmmachines $i -o yaml | grep release | tail -1; doneExample of system response:
master-0 release: 14.0.0+3.6.5 master-1 release: 14.0.0+3.6.5 master-2 release: 14.0.0+3.6.5
Obtain the name of the latest available Container Cloud release object:
kubectl get kaasrelease
Example of system response:
NAME AGE kaas-2-15-0 63m kaas-2-14-0 40d
Using the name of the latest Container Cloud release object, obtain the latest available Cluster release version:
kubectl get -o yaml clusterrelease $(kubectl get kaasrelease kaas-2-15-0 -o yaml | egrep "^ +clusterRelease:" | cut -d: -f2 | tr -d ' ') | egrep "^ version:"
Example of system response:
version: 14.0.0+3.6.4
Compare the output obtained in the previous step with the output from the first step. The Cluster releases must match. If this is not the case, contact Mirantis support for further details.
Proceed to Step 2. Initiate MOSK cluster update.
Change the upgrade order of a machine¶
You can define the upgrade sequence for existing machines to allow prioritized machines to be upgraded first during a cluster update.
Consider the following upgrade index specifics:
The first machine to upgrade is always one of the control plane machines with the lowest
upgradeIndex. Other control plane machines are upgraded one by one according to their upgrade indexes.If the
ClusterspecdedicatedControlPlanefield isfalse, worker machines are upgraded only after the upgrade of all control plane machines finishes. Otherwise, they are upgraded after the first control plane machine, concurrently with other control plane machines.If several machines have the same upgrade index, they have the same priority during upgrade.
If the value is not set, the machine is automatically assigned a value of the upgrade index.
To define the upgrade order of an existing machine:
Log in to the Container Cloud web UI with the
m:kaas:namespace@operatororm:kaas:namespace@writerpermissions.Switch to the required project using the Switch Project action icon located on top of the main left-side navigation panel.
In the Clusters tab, click the required cluster name. The cluster page with the Machines list opens.
In one of the Unassigned machines settings menu, select Change upgrade index.
In the Configure Upgrade Priority window that opens, use the Up and Down arrows in the Upgrade Index field to configure the upgrade sequence of a machine. Click Update to apply changes.
Using the Pool info or Machine info options in the machine settings menu, verify that the Upgrade Priority Index contains the updated value.
Configure the parallel update of worker nodes¶
Available since MCC 2.25.0 (17.0.0 and 16.0.0)
Note
You can start using the below procedure during cluster update from 23.1 to 23.2. For details, see Parallelizing node update operations.
By default, worker machines are upgraded sequentially, which includes node draining, software upgrade, services restart, and so on. Though, MOSK enables you to parallelize node upgrade operations, significantly improving the efficiency of your deployment, especially on large clusters.
For upgrade workflow of the control plane, see Change the upgrade order of a machine.
Configure the parallel update of worker nodes using web UI¶
Available since MCC 2.25.0 (17.0.0 and 16.0.0)
Log in to the Container Cloud web UI with the
m:kaas:namespace@operatororm:kaas:namespace@writerpermissions.Switch to the required project using the Switch Project action icon located on top of the main left-side navigation panel.
In the Clusters tab, click the required cluster name. The cluster page with the Machines list opens.
On the Clusters page, click the More action icon in the last column of the required cluster and select Configure cluster.
In General Settings of the Configure cluster window, define the following parameters:
- Parallel Upgrade Of Worker Machines
The maximum number of the worker nodes to update simultaneously. It serves as an upper limit on the number of machines that are drained at a given moment of time. Defaults to
1.You can configure this option after deployment before the cluster update.
- Parallel Preparation For Upgrade Of Worker Machines
The maximum number of worker nodes being prepared at a given moment of time, which includes downloading of new artifacts. It serves as a limit for the network load that can occur when downloading the files to the nodes. Defaults to
50.
Configure the parallel update of worker nodes using CLI¶
Available since MCC 2.24.0 (15.0.1 and 14.0.1)
Open the
Clusterobject for editing.Adjust the following parameters as required:
Configuration of the parallel node update¶ Parameter
Default
Description
spec.providerSpec.maxWorkerUpgradeCount1The maximum number of the worker nodes to update simultaneously. It serves as an upper limit on the number of machines that are drained at a given moment of time.
Caution
Since Container Cloud 2.27.0 (Cluster releases 17.2.0 and 16.2.0),
maxWorkerUpgradeCountis deprecated and will be removed in one of the following releases. Use theconcurrentUpdatesparameter in theUpdateGroupobject instead. For details, see Create update groups for worker machines.spec.providerSpec.maxWorkerPrepareCount50The maximum number of workers being prepared at a given moment of time, which includes downloading of new artifacts. It serves as a limit for the network load that can occur when downloading the files to the nodes.
Save the
Clusterobject to apply the change.
Create update groups for worker machines¶
Available since MCC 2.27.0 (17.2.0 and 16.2.0)
The use of update groups provides enhanced control over update of worker
machines by allowing granular concurrency settings for specific machine groups.
This feature uses the UpdateGroup object to decouple the concurrency
settings from the global cluster level, providing flexibility based on the
workload characteristics of different machine sets.
The UpdateGroup objects are processed sequentially based on their indexes.
Update groups with the same indexes are processed concurrently. The control
update group is always processed first.
Note
The update order of a machine within the same group is determined by the upgrade index of a specific machine. For details, see Change the upgrade order of a machine.
The maxWorkerUpgradeCount parameter of the Cluster object is inherited
by the default update group. Changing maxWorkerUpgradeCount leads to
changing the concurrentUpdates parameter of the default update group.
Note
The maxWorkerUpgradeCount parameter of the Cluster object is
deprecated and will be removed in one of the following Container Cloud
releases. You can still use this parameter to change the
concurrentUpdates value of the default update group. However, Mirantis
recommends changing this value directly in the UpdateGroup object.
Update group for controller nodes¶
Available since MCC 2.28.0 (17.3.0 and 16.3.0) TechPreview
The update group for controller nodes is automatically generated during initial cluster creation with the following settings:
name: <cluster-name>-controlindex: 1concurrentUpdates: 1rebootIfUpdateRequires: false
Caution
During a distribution upgrade, machines are always rebooted,
overriding rebootIfUpdateRequires: false.
All control plane machines are automatically assigned to the update group for controller nodes with no possibility to change it.
Note
On existing clusters created before Container Cloud 2.28.0 (Cluster releases 17.2.0, 16.2.0, or earlier), the update group for controller nodes is created after Container Cloud upgrade to 2.28.0 (Cluster release 16.3.0) on the management cluster.
Caution
The index and concurrentUpdates parameters of the update
group for controller nodes are hardcoded and cannot be changed.
Example of the update group for controller nodes:
apiVersion: kaas.mirantis.com/v1alpha1
kind: UpdateGroup
metadata:
name: example-cluster-control
namespace: example-ns
labels:
cluster.sigs.k8s.io/cluster-name: example-cluster
spec:
index: 1
concurrentUpdates: 1
rebootIfUpdateRequires: false
Default update group¶
The default update group is automatically created during initial cluster creation with the following settings:
name: <cluster-name>-defaultindex: 1rebootIfUpdateRequires: falseconcurrentUpdates: inherited from themaxWorkerUpgradeCountparameter set in theClusterobject
Caution
During a distribution upgrade, machines are always rebooted,
overriding rebootIfUpdateRequires: false.
Note
On existing clusters created before Container Cloud 2.27.0 (Cluster releases 17.1.0, 16.1.0, or earlier), the default update group is created after Container Cloud upgrade to 2.27.0 (Cluster release 16.2.0) on the management cluster.
Example of the default update group:
apiVersion: kaas.mirantis.com/v1alpha1
kind: UpdateGroup
metadata:
name: example-cluster-default
namespace: example-ns
labels:
cluster.sigs.k8s.io/cluster-name: example-cluster
spec:
index: 1
concurrentUpdates: 1
rebootIfUpdateRequires: false
If you require custom update settings for worker machines, create one or
several custom UpdateGroup objects as described below.
Assign a machine to an update group using CLI¶
Note
All worker machines that are not assigned to any update group are automatically assigned to the default update group.
Create an
UpdateGroupobject with the required specification. For description of the object fields, see UpdateGroup resource.Label the machines to associate them with the newly created
UpdateGroupobject:kubectl label machine <machineName> kaas.mirantis.com/update-group=<UpdateGroupObjectName>
To change the update group of a machine, update the
kaas.mirantis.com/update-grouplabel of the machine with the new update group name. Removing this label from a machine automatically assigns such machine to the default update group.
Note
After creation of a custom UpdateGroup object, if you plan to
add a new machine that requires a non-default update group, manually add
the corresponding label to the machine as described above. Otherwise, the
default update group is applied to such machine.
Note
Before removing the UpdateGroup object, reassign all machines to
another update group.
Granularly update a managed cluster using the ClusterUpdatePlan object¶
Available since MCC 2.27.0 (17.2.0) TechPreview
You can control the process of a managed cluster update by manually launching
update stages using the ClusterUpdatePlan custom resource. Between the
update stages, a cluster remains functional from the perspective of cloud
users and workloads.
A ClusterUpdatePlan object contains the following funtionality:
The object is automatically created by the bare metal provider when a new Cluster release becomes available for your cluster.
The object is created in the management cluster for the same namespace that the corresponding managed cluster refers to.
The object contains a list of self-descriptive update steps that are cluster-specific. These steps are defined in the
specsection of the object with information about their impact on the cluster.The object starts cluster update when the operator manually changes the
commencefield of the first update step totrue. All steps have thecommenceflag initially set tofalseso that the operator can decide when to pause or resume the update process.The object has the following naming convention:
<managedClusterName>-<targetClusterReleaseVersion>.Since Container Cloud 2.28.0 (Cluster release 17.3.0), the object contains several StackLight alerts to notify the operator about the update progress and potential update issues. For details, see StackLight alerts: Container Cloud.
Granularly update a managed cluster using CLI¶
Verify that the management cluster is upgraded successfully as described in Verify the management cluster status before MOSK update.
Optional. Available since Container Cloud 2.29.0 (Cluster release 17.4.0) as Technology Preview. Enable update auto-pause to be triggered by specific StackLight alerts. For details, see Configure update auto-pause.
Open the
ClusterUpdatePlanobject for editing.Start cluster update by changing the
spec:steps:commencefield of the first update step totrue.Once done, the following actions are applied to the cluster:
The Cluster release in the corresponding
Clusterspecis changed to the target Cluster version defined in theClusterUpdatePlanspec.The cluster update starts and pauses before the next update step with
commence: falseset in theClusterUpdatePlanspec.
Caution
Cancelling an already started update
stepis not supported.The following example illustrates the
ClusterUpdatePlanobject of a MOSK cluster update that has completed:Example of a completed
ClusterUpdatePlanobjectObject: apiVersion: kaas.mirantis.com/v1alpha1 kind: ClusterUpdatePlan metadata: creationTimestamp: "2025-02-06T16:53:51Z" generation: 11 name: mosk-17.4.0 namespace: child resourceVersion: "6072567" uid: 82c072be-1dc5-43dd-b8cf-bc643206d563 spec: cluster: mosk releaseNotes: https://docs.mirantis.com/mosk/latest/25.1-series.html source: mosk-17-3-0-24-3 steps: - commence: true description: - install new version of OpenStack and Tungsten Fabric life cycle management modules - OpenStack and Tungsten Fabric container images pre-cached - OpenStack and Tungsten Fabric control plane components restarted in parallel duration: estimated: 1h30m0s info: - 15 minutes to cache the images and update the life cycle management modules - 1h to restart the components granularity: cluster id: openstack impact: info: - some of the running cloud operations may fail due to restart of API services and schedulers - DNS might be affected users: minor workloads: minor name: Update OpenStack and Tungsten Fabric - commence: true description: - Ceph version update - restart Ceph monitor, manager, object gateway (radosgw), and metadata services - restart OSD services node-by-node, or rack-by-rack depending on the cluster configuration duration: estimated: 8m30s info: - 15 minutes for the Ceph version update - around 40 minutes to update Ceph cluster of 30 nodes granularity: cluster id: ceph impact: info: - 'minor unavailability of object storage APIs: S3/Swift' - workloads may experience IO performance degradation for the virtual storage devices backed by Ceph users: minor workloads: minor name: Update Ceph - commence: true description: - new host OS kernel and packages get installed - host OS configuration re-applied - container runtime version gets bumped - new versions of Kubernetes components installed duration: estimated: 1h40m0s info: - about 20 minutes to update host OS per a Kubernetes controller, nodes updated one-by-one - Kubernetes components update takes about 40 minutes, all nodes in parallel granularity: cluster id: k8s-controllers impact: users: none workloads: none name: Update host OS and Kubernetes components on master nodes - commence: true description: - new host OS kernel and packages get installed - host OS configuration re-applied - container runtime version gets bumped - new versions of Kubernetes components installed - data plane components (Open vSwitch and Neutron L3 agents, TF agents and vrouter) restarted on gateway and compute nodes - storage nodes put to “no-out” mode to prevent rebalancing - by default, nodes are updated one-by-one, a node group can be configured to update several nodes in parallel duration: estimated: 8h0m0s info: - host OS update - up to 15 minutes per node (not including host OS configuration modules) - Kubernetes components update - up to 15 minutes per node - OpenStack controllers and gateways updated one-by-one - nodes hosting Ceph OSD, monitor, manager, metadata, object gateway (radosgw) services updated one-by-one granularity: machine id: k8s-workers-vdrok-child-default impact: info: - 'OpenStack controller nodes: some running OpenStack operations might not complete due to restart of components' - 'OpenStack compute nodes: minor loss of the East-West connectivity with the Open vSwitch networking back end that causes approximately 5 min of downtime' - 'OpenStack gateway nodes: minor loss of the North-South connectivity with the Open vSwitch networking back end: a non-distributed HA virtual router needs up to 1 minute to fail over; a non-distributed and non-HA virtual router failover time depends on many factors and may take up to 10 minutes' users: major workloads: major name: Update host OS and Kubernetes components on worker nodes, group vdrok-child-default - commence: true description: - restart of StackLight, MetalLB services - restart of auxiliary controllers and charts duration: estimated: 1h30m0s granularity: cluster id: mcc-components impact: info: - minor cloud API downtime due restart of MetalLB components users: minor workloads: none name: Auxiliary components update target: mosk-17-4-0-25-1 status: completedAt: "2025-02-07T19:24:51Z" startedAt: "2025-02-07T17:07:02Z" status: Completed steps: - duration: 26m36.355605528s id: openstack message: Ready name: Update OpenStack and Tungsten Fabric startedAt: "2025-02-07T17:07:02Z" status: Completed - duration: 6m1.124356485s id: ceph message: Ready name: Update Ceph startedAt: "2025-02-07T17:33:38Z" status: Completed - duration: 24m3.151554465s id: k8s-controllers message: Ready name: Update host OS and Kubernetes components on master nodes startedAt: "2025-02-07T17:39:39Z" status: Completed - duration: 1h19m9.359184228s id: k8s-workers-vdrok-child-default message: Ready name: Update host OS and Kubernetes components on worker nodes, group vdrok-child-default startedAt: "2025-02-07T18:03:42Z" status: Completed - duration: 2m0.772243006s id: mcc-components message: Ready name: Auxiliary components update startedAt: "2025-02-07T19:22:51Z" status: Completed
Monitor the
messageandstatusfields of the first step. Themessagefield contains information about the progress of the current step. Thestatusfield can have the following values:NotStartedScheduledSince MCC 2.28.0 (17.3.0)InProgressAutoPausedTechPreview since MCC 2.29.0 (17.4.0)StuckCompleted
The
Scheduledstatus indicates that a step is already triggered but its execution has not started yet.The
AutoPausedstatus indicates that the update process is paused by a firing StackLight alert defined in theUpdateAutoPauseobject. For details, see Configure update auto-pause.The
Stuckstatus indicates an issue and that the step can not fit into the ETA defined in thedurationfield for this step. The ETA for each step is defined statically and does not change depending on the cluster.Caution
The status is not populated for the
ClusterUpdatePlanobjects that have not been started by adding thecommence: trueflag to the first object step. Therefore, always start updating the object from the first step.Optional. Available since Container Cloud 2.28.0 (Cluster releases 17.3.0 and 16.3.0). Add or remove update groups of worker nodes on the fly, unless the update of the group that is being removed has already been scheduled, or if a newly set group will have an index that is lower or equal to another group that is already scheduled. These changes are reflected in
ClusterUpdatePlan.You can also reassign a machine to a different update group while the cluster is being updated, but only if the new update group has an index higher than the index of the last scheduled worker update group. Disabled machines are considered as updated immediately.
Note
Depending on the number of update groups for worker nodes present in the cluster, the number of steps in
specdiffers. Each update group for worker nodes that has at least one machine will be represented by a step with the IDk8s-workers-<UpdateGroupName>.Proceed with changing the
commenceflag of the following update steps granularly depending on the cluster update requirements.Caution
Launch the update steps sequentially. A consecutive step is not started until the previous step is completed.
See also
Granularly update a managed cluster using the Container Cloud web UI¶
Available since MCC 2.29.0 (17.4.0 and 16.4.0)
Verify that the management cluster is upgraded successfully as described in Verify the management cluster status before MOSK update.
Optional. Available since Container Cloud 2.29.0 (Cluster release 17.4.0) as Technology Preview. Enable update auto-pause to be triggered by specific StackLight alerts. For details, see Configure update auto-pause.
Log in to the Container Cloud web UI with the
m:kaas:namespace@operatororm:kaas:namespace@writerpermissions.Switch to the required project using the Switch Project action icon located on top of the main left-side navigation panel.
On the Clusters page, in the Updates column of the required cluster, click the Available link. The Updates tab opens.
Note
If the Updates column is absent, it indicates that the cluster is up-to-date.
Note
For your convenience, the Cluster updates menu is also available in the right-side kebab menu of the cluster on the Clusters page.
On the Updates page, click the required version in the Target column to open update details, including the list of update steps, current and target cluster versions, and estimated update time.
In the Target version section of the Cluster update window, click Release notes and carefully read updates about target release, including the Update notes section that contains important pre-update and post-update steps.
Expand each step to verify information about update impact and other useful details.
Select one of the following options:
Enable Auto-commence all at the top-right of the first update step section and click Start Update to launch update and start each step automatically.
Click Start Update to only launch the first update step.
Note
This option allows you to auto-commence consecutive steps while the current step is in progress. Enable the Auto-commence toggle for required steps and click Save to launch the selected steps automatically. You will only be prompted to confirm the consecutive step, all remaining steps will be launched without a manual confirmation.
Before launching the update, you will be prompted to manually type in the target Cluster release name and confirm that you have read release notes about target release.
Caution
Cancelling an already started update step is not supported.
Monitor the status of each step by hovering over the In Progress icon at the top-right of the step window. While the step is in progress, its current status is updated every minute.
Once the required step is completed, the Waiting for input status at the top of the update window is displayed requiring you to confirm the next step.
The update history is retained in the Updates tab with the completion timestamp. The update plans that were not started and can no longer be used are cleaned up automatically.
Configure update auto-pause¶
Available since MOSK 25.1 TechPreview
Uinsg the UpdateAutoPause object, the operator can define specific
StackLight alerts that trigger auto-pause of an update phase execution in
a MOSK cluster. The feature enhances update management of
MOSK clusters by preventing harmful changes to be propagated
across the entire cloud.
Note
The feature is not available for management clusters.
When an update auto-pause is configured on a cluster, the following workflow applies:
During cluster updates, the system continuously monitors for the alerts defined in the
UpdateAutoPauseobjectIf any configured alert fires:
The update process automatically pauses
The
commencefield is removed from all steps that have not startedThe
commencefield is removed from the steps related toUpdate host OS and Kubernetes components on worker nodeseven if the step is in progress, and the step is pausedThe
ClusterUpdatePlanstatus changes toAutoPausedThe firing alerts are recorded in the
UpdateAutoPausestatusA condition is added to the
Clusterobject indicating the pause state
Configure auto-pausing of a MOSK cluster update¶
Verify that StackLight is enabled on the MOSK cluster.
Create an
UpdateAutoPauseobject with the name that matches your cluster name within the cluster namespace. For example:apiVersion: kaas.mirantis.com/v1alpha1 kind: UpdateAutoPause metadata: name: managed-cluster-example # Must match cluster name namespace: managed-cluster-ns # Must match cluster namespace spec: alerts: - AlertName1 - AlertName2
The list of alerts can include standard and custom StackLight alerts previously configured for the cluster.
For the object spec, see UpdateAutoPause resource.
Apply the configuration:
kubectl apply -f update-autopause.yaml
Resume paused updates¶
Select one of the following options:
Investigate and resolve the conditions that triggered the alerts, then wait for the alerts to clear automatically
Remove the problematic alert from the
UpdateAutoPauseconfiguration
Set the
commencefield totruefor the relevantUpdatePlansteps to resume the update.
Caution
Admission Controller blocks attempts to set commence: true
while alerts defined in the UpdateAutoPause object are still firing.
Monitor the status of an update auto-pause¶
You can monitor the status of an update auto-pause using the following resources:
The
UpdateAutoPauseobject status:kubectl get updateautopause <cluster-name> -n <namespace> -o yaml
The
ClusterUpdatePlanobject status that displays the following details:The
AutoPausedstatus when updates are paused.Messages indicating which alerts caused the pause and other relevant information.
StackLight alerts:
ClusterUpdateAutoPaused, which indicates that an update is currently paused.ClusterUpdateStepAutoPaused, which describes specific steps that are paused.
For alert details, see Container Cloud.
Calculate a maintenance window duration for update Deprecated¶
Deprecation notice
The maintenance window duration calculator is deprecated. Starting from
MOSK 25.1, cloud operators should use the
ClusterUpdatePlan API instead. For details, refer to
ClusterUpdatePlan resource.
This section provides an online calculator for quick calculation of the approximate time required to update your MOSK cluster that uses Open vSwitch as a networking backend.
Additionally, for a more accurate calculation, consider any cluster-specific factors that can have a large impact on the update time in some edge cases, such as number of routers, frequency of CPU, and so on.
Getting access¶
This section contains instructions on how to get access to different systems of a MOSK cluster.
To obtain endpoints of the MKE web UI and StackLight web UIs such as Prometheus, Alertmanager, Alerta, OpenSearch Dashboards, and Grafana, in the Clusters tab of the Container Cloud web UI, navigate to More > Cluster info.
Note
The Alertmanager web UI displays alerts received by all configured receivers, which can be mistaken for duplicates. To only display the alerts received by a particular receiver, use the Receivers filter.
Generate a kubeconfig for a MOSK cluster using API¶
This section describes how to generate a MOSK cluster
kubeconfig using the Container Cloud API. You can also download a
MOSK cluster kubeconfig using the
Download Kubeconfig option in the Container Cloud web UI. For
details, see Connect to a MOSK cluster.
To generate a MOSK cluster kubeconfig using API:
Obtain the following details:
Your
<username>with the corresponding password that were created after the management cluster bootstrap as described in Create initial users after a management cluster bootstrap.The
kubeconfigof your<username>that you can download through the Container Cloud web UI using Download Kubeconfig located under your<username>on the top-left of the page.
Obtain the
<cluster>object of the<cluster_name>MOSK cluster:kubectl get cluster <cluster_name> -n <project_name> -o yaml
Obtain the access token from Keycloak for the
<username>user:curl -d 'client_id=<cluster.status.providerStatus.oidc.clientId>' --data-urlencode 'username=<username>' --data-urlencode 'password=<password>' -d 'grant_type=password' -d 'response_type=id_token' -d 'scope=openid' <cluster.status.providerStatus.oidc.issuerURL>/protocol/openid-connect/token
Generate the MOSK cluster
kubeconfigusing the data from<cluster.status>and<token>obtained in the previous steps. Use the following template as an example:apiVersion: v1 clusters: - name: <cluster_name> cluster: certificate-authority-data: <cluster.status.providerStatus.apiServerCertificate> server: https://<cluster.status.providerStatus.loadBalancerHost>:443 contexts: - context: cluster: <cluster_name> user: <username> name: <username>@<cluster_name> current-context: <username>@<cluster_name> kind: Config preferences: {} users: - name: <username> user: auth-provider: config: client-id: <cluster.status.providerStatus.oidc.clientId> idp-certificate-authority-data: <cluster.status.providerStatus.oidc.certificate> idp-issuer-url: <cluster.status.providerStatus.oidc.issuerUrl> refresh-token: <token.refresh_token> id-token: <token.id_token> name: oidc
Connect to a MOSK cluster¶
Note
The Container Cloud web UI communicates with Keycloak to authenticate users. Keycloak is exposed using HTTPS with self-signed TLS certificates that are not trusted by web browsers.
To use your own TLS certificates for Keycloak, refer to Configure TLS certificates for cluster applications.
After you deploy a MOSK management or managed cluster, connect to the cluster to verify the availability and status of the nodes as described below.
To connect to a MOSK cluster:
Log in to the Container Cloud web UI with the
m:kaas:namespace@operatororm:kaas:namespace@writerpermissions.Switch to the required project using the Switch Project action icon located on top of the main left-side navigation panel.
In the Clusters tab, click the required cluster name. The cluster page with the Machines list opens.
Verify the status of the manager nodes. Once the first manager node is deployed and has the Ready status, the Download Kubeconfig option for the cluster being deployed becomes active.
Open the Clusters tab.
Click the More action icon in the last column of the required cluster and select Download Kubeconfig:
Enter your user password.
Not recommended. Select Offline Token to generate an offline IAM token. Otherwise, for security reasons, the
kubeconfigtoken expires every 30 minutes of the Container Cloud API idle time and you have to downloadkubeconfigagain with a newly generated token.Click Download.
Verify the availability of the managed cluster machines:
Export the
kubeconfigparameters to your local machine with access to kubectl. For example:export KUBECONFIG=~/Downloads/kubeconfig-test-cluster.yml
Obtain the list of available machines:
kubectl get nodes -o wide
The system response must contain the details of the nodes in the
READYstatus.
To connect to a management cluster:
Log in to a local machine where your management cluster
kubeconfigis located and wherekubectlis installed.Note
The management cluster
kubeconfigis created during the last stage of the management cluster bootstrap.Obtain the list of available management cluster machines:
kubectl get nodes -o wide
The system response must contain the details of the nodes in the
READYstatus.
Access the Keycloak Admin Console¶
Using the Keycloak Admin Console, you can create or delete a user as well as grant or revoke roles to or from a user. The Keycloak administrator is responsible for assigning roles to users depending on the level of access they need in a cluster.
Obtain access credentials using the CLI¶
Available since MCC 2.22.0 (Cluster release 11.6.0)
./container-cloud get keycloak-creds --mgmt-kubeconfig <pathToManagementClusterKubeconfig>
Optionally, use the --output key to save credentials in a YAML file.
Example of system response:
Keycloak admin credentials:
Address: https://<keycloak-ip-adress>/auth
Login: keycloak
Password: foobar
Obtain access credentials using kubectl¶
kubectl get cluster <mgmtClusterName> -o=jsonpath='{.status.providerStatus.helm.releases.iam.keycloak.url}'
The system response contains the URL to access the Keycloak Admin Console.
The user name is keycloak by default. The password is located in
passwords.yaml generated during bootstrap.
You can also obtain the password from the iam-api-secrets secret
in the kaas namespace of the management cluster and
decode the content of the keycloak_password key:
kubectl get secret iam-api-secrets -n kaas -o=jsonpath='{.data.keycloak_password}' | base64 -d
Access the Tungsten Fabric web UI¶
The Tungsten Fabric (TF) web UI allows for easy and fast TF resources
configuration, monitoring, and debugging. You can access
the TF web UI through either the Ingress service or the Kubernetes Service
directly. TLS termination for the https protocol is performed through the
Ingress service.
Note
Mirantis OpenStack for Kubernetes provides the TF web UI as is and does not include this service in the support Service Level Agreement.
To access the TF web UI through Ingress:
Log in to a local machine where kubectl is installed.
Obtain and export
kubeconfigof your managed cluster as described in Connect to a MOSK cluster.Obtain the password of the Admin user:
kubectl -n openstack get secret keystone-keystone-admin -ojsonpath='{.data.OS_PASSWORD}' | base64 -d
Obtain the external IP address of the Ingress service:
kubectl -n openstack get services ingress
Example of system response:
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE ingress LoadBalancer 10.96.32.97 10.172.1.101 80:34234/TCP,443:34927/TCP,10246:33658/TCP 4h56m
Note
Do not use the
EXTERNAL-IPvalue to directly access the TF web UI. Instead, use the FQDN from the list below.Obtain the FQDN of
tf-webui:Note
The command below outputs all host names assigned to the TF web UI service. Use one of them.
kubectl -n tf get ingress tf-webui -o custom-columns=HOSTS:.spec.rules[*].host
Configure DNS to access the TF web UI host as described in Configure DNS to access OpenStack.
Use your favorite browser to access the TF web UI at
https://<FQDN-WEBUI>.
Add a cluster to Lens¶
For quick and easy inspection and monitoring, you can add a MOSK cluster to Lens using the Container Cloud web UI. The following options are available in the More action icon menu of each cluster:
Add cluster to Lens
Open cluster in Lens
Before you can start monitoring your clusters in Lens, install the Container Cloud Lens extension as described below.
Install the Container Cloud Lens extension¶
Start Lens.
Verify that your Lens version is 4.2.4 or later.
Select Lens > Extensions.
Copy and paste the following text into the Install Extension field:
@mirantis/lens-extension-cc
Click Install.
Verify that the Container Cloud Lens extension appears in the Installed Extensions section.
Add a cluster to Lens¶
Enable your browser to open pop-ups for the Container Cloud web UI.
Log in to the Container Cloud web UI with the
m:kaas:namespace@operatororm:kaas:namespace@writerpermissions.Open the Clusters tab.
Verify that the target cluster is successfully deployed and is in the Ready status.
In the last column of the target cluster area, click the More action icon and select Add cluster to Lens.
In the Add Cluster To Lens window, click Add. The system redirects you to Lens that now contains the previously added cluster.
Caution
If prompted, allow your browser to open Lens.
Open a cluster in Lens¶
Add the target cluster to Lens as described above.
Log in to the Container Cloud web UI with the
m:kaas:namespace@operatororm:kaas:namespace@writerpermissions.Open the Clusters tab.
In the last column of the target cluster area, click the More action icon and select Open cluster in Lens.
OpenStack operations¶
The section covers the management aspects of an OpenStack cluster deployed on Kubernetes.
Upgrade OpenStack¶
This section provides instructions on how to upgrade OpenStack to a major version on a MOSK cluster.
Note
The update of the OpenStack components within the same major OpenStack version is performed seamlessly as part of the MOSK cluster update.
Prerequisites¶
Verify that your OpenStack cloud is running on the latest MOSK release. See Release Compatibility Matrix for the release matrix and supported upgrade paths.
Just before the upgrade, back up your OpenStack databases. See the following documentation for details:
Verify that OpenStack is healthy and operational. All OpenStack components in the
healthgroup in theOpenStackDeploymentStatusCR should be in theReadystate. See OpenStackDeploymentStatus custom resource for details.Verify the workability of your OpenStack deployment by running Tempest against the OpenStack cluster as described in Run Tempest tests. Verification of the testing pass rate before upgrading will help you measure your cloud quality before and after upgrade.
Read carefully through the Release Notes of your MOSK version paying attention to the Known issues section and the OpenStack upstream release notes for the target OpenStack version.
Calculate the maintenance window using Plan the cluster update and Calculate a maintenance window duration for update Deprecated and notify users.
When upgrading to OpenStack Yoga, remove the Panko service from the cloud by removing the
evententry from thespec:features:servicesstructure in theOpenStackDeploymentresource as described in Remove an OpenStack service.Note
The OpenStack Panko service has been removed from the product and is no longer maintained in the upstream OpenStack. See the project repository page for details.
Perform the upgrade¶
To start the OpenStack upgrade, change the value of the
spec:openstack_version parameter in the OpenStackDeployment object
to the target OpenStack release.
After you change the value of the spec:openstack_version parameter,
the OpenStack Controller initializes the upgrade process.
To verify the upgrade status, use:
Logs from the
osdplcontainer in the OpenStack Controllerrockoonpod.The
OpenStackDeploymentStatusobject.When upgrade starts, the
OPENSTACK VERSIONfield content changes to the target OpenStack version, andSTATEdisplaysAPPLYING:kubectl -n openstack get osdplst
Example of system output:
NAME OPENSTACK VERSION CONTROLLER VERSION STATE osh-dev antelope 0.15.6 APPLYING
When upgrade finishes, the
STATEfield should displayAPPLIED:kubectl -n openstack get osdplst
Example of system output:
NAME OPENSTACK VERSION CONTROLLER VERSION STATE osh-dev antelope 0.15.6 APPLIED
The maintenance window for the OpenStack upgrade usually takes from two to four hours, depending on the cloud size.
Verify the upgrade¶
Verify that OpenStack is healthy and operational. All OpenStack components in the
healthgroup in theOpenStackDeploymentStatusCR should be in theReadystate. See OpenStackDeploymentStatus custom resource for details.Verify the workability of your OpenStack deployment by running Tempest against the OpenStack cluster as described in Run Tempest tests.
Upgrade from Antelope to Caracal¶
Before upgrading, verify that you have completed the Prerequisites and removed the domains from federation mappings as described below.
Warning
If your MOSK cluster is running version
24.3 and includes the Instance High Availability service (OpenStack
Masakari), the OpenStack upgrade will fail due to an incorrect migration
of the Masakari database from legacy SQLAlchemy Migrate to Alembic caused
by a misconfigured alembic_table. To avoid this issue, follow the
workaround steps outlined in [47603] Masakari fails during the OpenStack upgrade to Caracal before proceeding
with the upgrade.
Warning
If you initially deployed your MOSK cluster
with OpenStack Victoria or earlier releases and gradually upgraded it to
Antelope, and you do not perform periodic cleanups of OpenStack databases
from soft-deleted rows, the upgrade will stuck due to the failing
cinder-db-sync job.
To prevent, diagnose, and fix this issue, perform the workaround steps outlined in [47695] Cinder database sync job fails during upgrade from Antelope to Caracal before proceeding with the upgrade.
MOSK enables you to upgrade directly from Antelope to
Caracal without the need to upgrade to the intermediate Bobcat release.
To upgrade the cloud, complete the upgrade steps
instruction changing the value of the spec:openstack_version parameter
in the OpenStackDeployment object from antelope to caracal.
Important
Perform the domains removal from the federation mappings if your MOSK cluster configuration includes federated identity management system, such as IAM or any other supported identity provider.
Before Caracal, Keystone does not properly handle domain specifications for users in mappings. Even though domains are specified for users, Keystone always creates users in the domain associated with the identity provider the user logs in from.
Starting with Caracal, Keystone honors the domains specified for users in mappings. Many example mappings, including the previous default mapping in MOSK, use domain specifications. After upgrading to Caracal, the new users logging in through federation may be assigned to a different Keystone domain, while existing users will retain their current domain. This behaviour may negatively impact monitoring, compliance, and overall cluster operations.
To maintain the same functionality after the upgrade, remove the domain
element from both the local.user element and local element, which
sets default domain values for user and group elements, from the previous
default mappings.
You can use the openstack mapping commands to manage mappings:
To list available mappings: openstack mapping list
To display the mapping rules: openstack mapping show <name>
To modify the mapping rules: openstack mapping set <name> --rules <rules>
Example mapping rules in Antelope:
[
{
"local": [
{
"user": {
"name": "{0}",
"email": "{1}",
"domain": {
"name": "Default"
}
}
},
{
"groups": "{2}",
"domain": {
"name": "Default"
}
},
{
"domain": {
"name": "Default"
}
}
],
"remote": [
{
"type": "OIDC-iam_username"
},
{
"type": "OIDC-email"
},
{
"type": "OIDC-iam_roles"
}
]
}
]
Example mapping rules in Caracal:
[
{
"local": [
{
"user": {
"name": "{0}",
"email": "{1}"
}
},
{
"groups": "{2}",
"domain": {
"name": "Default"
}
}
],
"remote": [
{
"type": "OIDC-iam_username"
},
{
"type": "OIDC-email"
},
{
"type": "OIDC-iam_roles"
}
]
}
]
Upgrade from Yoga to Antelope¶
MOSK enables you to upgrade directly from Yoga to Antelope without the need to upgrade to the intermediate Zed release.
Before upgrading, verify that you have completed the Prerequisites.
Important
There are several known issue affecting MOSK clusters running OpenStack Antelope that can disrupt the network connectivity of the cloud workloads.
If your cluster is still running OpenStack Yoga, update to the MOSK 24.2.1 patch release first and only then upgrade to OpenStack Antelope. If you have not been applying patch releases previously and would prefer to switch back to major releases-only mode, you will be able to do this when MOSK 24.3 is released.
If you have updated your cluster to OpenStack Antelope, apply the workarounds described in Release notes: OpenStack known issues for the following issues:
[45879] [Antelope] Incorrect packet handling between instance and its gateway
[44813] Traffic disruption observed on trunk ports
To upgrade the cloud, complete the upgrade steps
instruction changing the value of the spec:openstack_version parameter
in the OpenStackDeployment object from yoga to antelope.
Upgrade from Victoria to Yoga¶
Caution
If your cluster is running on top of the MOSK 23.1.2 patch version, the OpenStack upgrade to Yoga may fail due to the delay in the Cinder start. For the workaround, see 23.1.2 known issues: OpenStack upgrade failure.
Before upgrading, verify that you have completed the Prerequisites.
If your cloud runs on top of the OpenStack Victoria release, you must first upgrade to the technical OpenStack releases Wallaby and Xena before upgrading to Yoga.
Caution
The Wallaby and Xena releases are not recommended for a long-run production usage. These versions are transitional, so-called technical releases with limited testing scopes. For the OpenStack versions support cycle, refer to OpenStack support cycle.
To upgrade the cloud, complete the upgrade steps for each release version in line in the following strict order:
Upgrade the cloud from
victoriatowallabyUpgrade the cloud from
wallabytoxenaUpgrade the cloud from
xenatoyoga
Backup and restore OpenStack databases¶
Mirantis OpenStack for Kubernetes (MOSK) relies on the MariaDB Galera cluster to provide its OpenStack components with reliable storage of persistent data. Mirantis recommends backing up your OpenStack databases daily to ensure the safety of your cloud data. Also, you should always create an instant backup before updating your cloud or performing any kind of potentially disruptive experiment.
MOSK has a built-in automated backup routine that can be
triggered manually or by schedule. Periodic backups are suspended by default
but you can easily enable them through the OpenStackDeployment custom
resource. For the details about enablement and configuration of the periodic
backups, refer to Periodic OpenStack database backups in the Reference Architecture.
This section includes more intricate procedures that involve additional steps
beyond editing the OpenStackDeployment custom resource, such as restoring
the OpenStack database from a backup or configuring a remote storage for
backups.
Enable OpenStack database remote backups¶
TechPreview
By default, MOSK stores the OpenStack database backups locally in the Mirantis Ceph cluster, which is a part of the same cloud.
Alternatively, MOSK enables you to save the backup data to an external storage. This section contains the details on how you, as a cloud operator, can configure a remote storage backend for OpenStack database backups.
In general, the built-in automated backup mechanism saves the data to the
mariadb-phy-backup-data PersistentVolumeClaim (PVC), which is provisioned
from StorageClass specified in the spec.persistent_volume_storage_class
parameter of the OpenstackDeployment custom resource (CR).
If your MOSK cluster was originally deployed with the default backup storage, proceed with this step. Otherwise, skip it.
Copy the already existing backup data to a storage different from the
mariadb-phy-backup-dataPVC.Remove the
mariadb-phy-backup-dataPVC manually:kubectl -n openstack delete pvc mariadb-phy-backup-data
Enable the NFS backend in the
OpenStackDeploymentobject by editing the backup section of theOpenStackDeploymentCR as follows:spec: features: database: backup: enabled: true backend: pv_nfs pv_nfs: server: <ip-address/dns-name-of-the-server> path: <path-to-the-share-folder-on-the-server>
Optional. Set the required mount options for the NFS mount command. You can set as many options of mount as you need. For example:
spec: services: database: mariadb: values: volume: phy_backup: nfs: mountOptions: - "nfsvers=4" - "hard"
Verify the
mariadb-phy-backup-dataPVC and NFS persistent volume (PV):kubectl -n openstack get pvc mariadb-phy-backup-data -o wide
kubectl -n openstack get pv mariadb-phy-backup-data-nfs-pv -o yaml
An example of a positive system response:
NAME STATUS VOLUME CAPACITY ACCESS MODES STORAGECLASS AGE VOLUMEMODE mariadb-phy-backup-data Bound mariadb-phy-backup-data-nfs-pv 20Gi RWO 5m40s Filesystem
apiVersion: v1 kind: PersistentVolume metadata: annotations: meta.helm.sh/release-name: openstack-mariadb meta.helm.sh/release-namespace: openstack <<<skipped>>>> name: mariadb-phy-backup-data-nfs-pv resourceVersion: "2279204" uid: 60db9f89-afc4-417b-bf44-8acab844f17e spec: accessModes: - ReadWriteOnce capacity: storage: 20Gi claimRef: apiVersion: v1 kind: PersistentVolumeClaim name: mariadb-phy-backup-data namespace: openstack resourceVersion: "2279201" uid: e0e08d73-e56f-425a-ad4e-e5393aa3cdc1 mountOptions: - nfsvers=4 - hard nfs: path: / server: 10.10.0.116 persistentVolumeReclaimPolicy: Retain volumeMode: Filesystem status: phase: Bound
Remove NFS PVC and PV:
kubectl -n openstack delete pvc mariadb-phy-backup-data
kubectl -n openstack delete pv mariadb-phy-backup-data-nfs-pv
Re-enable the local backup in the
OpenStackDeploymentCR:spec: features: database: backup: enabled: true backend: pvc
Verify that the
mariadb-phy-backup-dataPVC uses the default PV:kubectl -n openstack get pvc mariadb-phy-backup-data
An example of a positive system response:
NAME STATUS VOLUME CAPACITY ACCESS MODES STORAGECLASS AGE mariadb-phy-backup-data Bound pvc-a4f6e24b-c05b-4a76-bca2-bb6a5c8ef5b5 20Gi RWO mirablock-k8s-block-hdd 80m
Restore OpenStack databases from a backup¶
During the OpenStack database restoration, the MariaDB cluster is unavailable due to the MariaDB StatefulSet being scaled down to 0 replicas. Therefore, to safely restore the state of the OpenStack database, plan the maintenance window thoroughly and in accordance with the database size.
The duration of the maintenance window may depend on the following:
Network throughput
Performance of the storage where backups are kept, which is Mirantis Ceph by default
Local disks performance of the nodes where MariaDB data resides
To restore OpenStack databases:
Obtain an image of the MariaDB container:
kubectl -n openstack get pods mariadb-server-0 -o jsonpath='{.spec.containers[0].image}'
Create the
check_pod.yamlfile to create the helper pod required to view the backup volume content:--- apiVersion: v1 kind: ServiceAccount metadata: name: check-backup-helper namespace: openstack --- apiVersion: v1 kind: Pod metadata: name: check-backup-helper namespace: openstack labels: application: check-backup-helper spec: nodeSelector: openstack-control-plane: enabled containers: - name: helper securityContext: allowPrivilegeEscalation: false runAsUser: 0 readOnlyRootFilesystem: true command: - sleep - infinity image: << image of mariadb container >> imagePullPolicy: IfNotPresent volumeMounts: - name: pod-tmp mountPath: /tmp - mountPath: /var/backup name: mysql-backup restartPolicy: Never serviceAccount: check-backup-helper serviceAccountName: check-backup-helper volumes: - name: pod-tmp emptyDir: {} - name: mariadb-secrets secret: secretName: mariadb-secrets defaultMode: 0444 - name: mariadb-bin configMap: name: mariadb-bin defaultMode: 0555 - name: mysql-backup persistentVolumeClaim: claimName: mariadb-phy-backup-data
Create the helper pod:
kubectl -n openstack apply -f check_pod.yaml
Obtain the name of the backup to restore:
kubectl -n openstack exec -t check-backup-helper -- tree /var/backup
Example of system response:
/var/backup |-- base | `-- 2020-09-09_11-35-48 | |-- backup.stream.gz | |-- backup.successful | |-- grastate.dat | |-- xtrabackup_checkpoints | `-- xtrabackup_info |-- incr | `-- 2020-09-09_11-35-48 | |-- 2020-09-10_01-02-36 | |-- 2020-09-11_01-02-02 | |-- 2020-09-12_01-01-54 | |-- 2020-09-13_01-01-55 | `-- 2020-09-14_01-01-55 `-- lost+found 10 directories, 5 files
If you want to restore the full backup, the name from the example above is
2020-09-09_11-35-48. To restore a specific incremental backup, the name from the example above is2020-09-09_11-35-48/2020-09-12_01-01-54.In the example above, the backups will be restored in the following strict order:
2020-09-09_11-35-48- full backup, path/var/backup/base/2020-09-09_11-35-482020-09-10_01-02-36- incremental backup, path/var/backup/incr/2020-09-09_11-35-48/2020-09-10_01-02-362020-09-11_01-02-02- incremental backup, path/var/backup/incr/2020-09-09_11-35-48/2020-09-11_01-02-022020-09-12_01-01-54- incremental backup, path/var/backup/incr/2020-09-09_11-35-48/2020-09-12_01-01-54
Delete the helper pod:
kubectl -n openstack delete -f check_pod.yaml
Pass the following parameters to the
mariadb_resque.pyscript from the OsDpl object:Parameter
Type
Default
Description
--backup-nameString
Name of a folder with backup in
<BASE_BACKUP>or<BASE_BACKUP>/<INCREMENTAL_BACKUP>.--replica-restore-timeoutInteger
3600Timeout in seconds for 1 replica data to be restored to the
mysqldata directory. Also, includes time for spawning a rescue runner pod in Kubernetes and extracting data from a backup archive.Edit the
OpenStackDeploymentobject as follows:spec: services: database: mariadb: values: manifests: job_mariadb_phy_restore: true conf: phy_restore: backup_name: "2020-09-09_11-35-48/2020-09-12_01-01-54" replica_restore_timeout: 7200
Wait until the
mariadb-phy-restorejob suceeds:kubectl -n openstack get jobs mariadb-phy-restore -o jsonpath='{.status}'Important
If
mariadb-phy-restorefails, the MariaDB Pods do not start automatically. For example, the failure may occur due to discrepancy between the current and backup versions of MariaDB, broken backup archive, and so on.Assess the
mariadb-phy-restorejob log to identify the issue:kubectl -n openstack logs --tail=10000 -l application=mariadb-phy-restore,job-name=mariadb-phy-restore
If the restoration process does not start due to the MariaDB versions discrepancy:
Use other backup file with the corresponding MariaDB version for restoration, if any.
Start MariaDB Pods without restoration:
kubectl scale --replicas=3 sts/mariadb-server -n openstack
The command above restores the previous cluster state.
The
mariadb-phy-restorejob is an immutable object. Therefore, remove the job after each successful execution. To correctly remove the job, clean up all the settings from theOpenStackDeploymentobject that you have configured during step 7 of this procedure. This will remove all related pods as well.Resolve database discrepancies by analysing the following resources that may be inconsistent in the restored snapshot as opposed to the original environment:
Leftover VMs, volumes, images, and other dynamic resources. For example:
A VM is removed after a snapshot for restoration is created. Such VM will be present as an orphan entry in the database and the OpenStack API after restoration.
A VM is created after a snapshot for restoration is created. Such VM will disappear from the OpenStack API after database restoration but will still be present as a process on the compute host.
Broken Octavia Amphorae that may become unresponsive after restoration, potentially requiring LoadBalancer failover
Other broken or leftover resources
Verify the periodic backup jobs for the OpenStack database¶
Verify pods in the
openstacknamespace. After the backup jobs have succeeded, the pods stay in theCompletedstate:kubectl -n openstack get pods -l application=mariadb-phy-backup
Example of a posistive system response:
NAME READY STATUS RESTARTS AGE mariadb-phy-backup-1599613200-n7jqv 0/1 Completed 0 43h mariadb-phy-backup-1599699600-d79nc 0/1 Completed 0 30h mariadb-phy-backup-1599786000-d5kc7 0/1 Completed 0 6h17m
Note
By default, the system keeps three latest successful and one latest failed pods.
Obtain an image of the MariaDB container:
kubectl -n openstack get pods mariadb-server-0 -o jsonpath='{.spec.containers[0].image}'
Create the
check_pod.yamlfile to create the helper pod required to view the backup volume content.Configuration example:
apiVersion: v1 kind: ServiceAccount metadata: name: check-backup-helper namespace: openstack --- apiVersion: v1 kind: Pod metadata: name: check-backup-helper namespace: openstack labels: application: check-backup-helper spec: nodeSelector: openstack-control-plane: enabled containers: - name: helper securityContext: allowPrivilegeEscalation: false runAsUser: 0 readOnlyRootFilesystem: true command: - sleep - infinity image: << image of mariadb container >> imagePullPolicy: IfNotPresent volumeMounts: - name: pod-tmp mountPath: /tmp - mountPath: /var/backup name: mysql-backup restartPolicy: Never serviceAccount: check-backup-helper serviceAccountName: check-backup-helper volumes: - name: pod-tmp emptyDir: {} - name: mariadb-secrets secret: secretName: mariadb-secrets defaultMode: 0444 - name: mariadb-bin configMap: name: mariadb-bin defaultMode: 0555 - name: mysql-backup persistentVolumeClaim: claimName: mariadb-phy-backup-data
Apply the helper service account and pod resources:
kubectl -n openstack apply -f check_pod.yaml kubectl -n openstack get pods -l application=check-backup-helper
Example of a positive system response:
NAME READY STATUS RESTARTS AGE check-backup-helper 1/1 Running 0 27s
Verify the directories structure within the
/var/backupdirectory of the spawned pod:kubectl -n openstack exec -t check-backup-helper -- tree /var/backup
Example of a system response:
/var/backup |-- base | `-- 2020-09-09_11-35-48 | |-- backup.stream.gz | |-- backup.successful | |-- grastate.dat | |-- xtrabackup_checkpoints | `-- xtrabackup_info |-- incr | `-- 2020-09-09_11-35-48 | |-- 2020-09-10_01-02-36 | | |-- backup.stream.gz | | |-- backup.successful | | |-- grastate.dat | | |-- xtrabackup_checkpoints | | `-- xtrabackup_info | `-- 2020-09-11_01-02-02 | |-- backup.stream.gz | |-- backup.successful | |-- grastate.dat | |-- xtrabackup_checkpoints | `-- xtrabackup_info
The
basedirectory contains full backups. Each directory in theincrfolder contains incremental backups related to a certain full backup in thebasefolder. All incremental backups always have the base backup name as parent folder.Delete the helper pod:
kubectl delete -f check_pod.yaml
Add a controller node¶
This section describes how to add a new control plane node to the existing MOSK deployment.
To add an OpenStack controller node:
Add a bare metal host to the MOSK cluster as described in Add a bare metal host.
When adding the bare metal host YAML file, specify the following OpenStack control plane node labels for the OpenStack control plane services such as database, messaging, API, schedulers, conductors, L3 and L2 agents:
openstack-control-plane=enabledopenstack-gateway=enabledopenvswitch=enabled
Create a Kubernetes machine in your cluster as described in Add a machine.
When adding the machine, verify that OpenStack control plane node has the following labels:
openstack-control-plane=enabledopenstack-gateway=enabledopenvswitch=enabled
Note
Depending on the applications that were colocated on the failed controller node, you may need to specify some additional labels, for example,
ceph_role_mgr=trueandceph_role_mon=true. To successfuly replace a failedmonandmgrnode, refer to Ceph operations.Verify that the node is in the
Readystate through the Kubernetes API:kubectl get node <NODE-NAME> -o wide | grep Ready
Verify that the node has all required labels described in the previous steps:
kubectl get nodes --show-labels
Configure new Octavia health manager resources:
Rerun the
octavia-create-resourcesjob:kubectl -n osh-system exec -t <OS-CONTROLLER-POD> -c osdpl osctl-job-rerun octavia-create-resources openstack
Wait until the Octavia health manager pod on the newly added control plane node appears in the
Runningstate:kubectl -n openstack get pods -o wide | grep <NODE_ID> | grep octavia-health-manager
Note
If the pod is in the
crashloopbackoffstate, remove it:kubectl -n openstack delete pod <OCTAVIA-HEALTH-MANAGER-POD-NAME>
Verify that an OpenStack port for the node has been created and the node is in the
Activestate:kubectl -n openstack exec -t <KEYSTONE-CLIENT-POD-NAME> openstack port show octavia-health-manager-listen-port-<NODE-NAME>
Replace a failed controller node¶
This section describes how to replace a failed control plane node in your
MOSK deployment. The procedure applies to the control plane
nodes that are, for example, permanently failed due to a hardware failure and
appear in the NotReady state:
kubectl get nodes <CONTAINER-CLOUD-NODE-NAME>
Example of system response:
NAME STATUS ROLES AGE VERSION
<CONTAINER-CLOUD-NODE-NAME> NotReady <none> 10d v1.18.8-mirantis-1
To replace a failed controller node:
Remove the Kubernetes labels from the failed node by editing the
.metadata.labelsnode object:kubectl edit node <CONTAINER-CLOUD-NODE-NAME>
If your cluster is deployed with a compact control plane, inspect precautions for a cluster machine deletion.
Add the control plane node to your deployment as described in Add a controller node.
Identify all stateful applications present on the failed node:
node=<CONTAINER-CLOUD-NODE-NAME> claims=$(kubectl -n openstack get pv -o jsonpath="{.items[?(@.spec.nodeAffinity.required.nodeSelectorTerms[0].matchExpressions[0].values[0] == '${node}')].spec.claimRef.name}") for i in $claims; do echo $i; done
Example of system response:
mysql-data-mariadb-server-2 openstack-operator-bind-mounts-rfr-openstack-redis-1 etcd-data-etcd-etcd-0
For MOSK 23.3 series or earlier, reschedule stateful applications pods to healthy controller nodes as described in Reschedule stateful applications. For the newer versions, MOSK performs the rescheduling of stateful applications automatically.
If the failed controller node had the
StackLightlabel, fix the StackLight volume node affinity conflict as described in Delete a cluster machine.Remove the OpenStack port related to the Octavia health manager pod of the failed node:
kubectl -n openstack exec -t <KEYSTONE-CLIENT-POD-NAME> openstack port delete octavia-health-manager-listen-port-<NODE-NAME>
For clouds using Open Virtual Network (OVN) as the networking backend, remove the Northbound and Southbound database members for the failed node:
Log in to the running
openvswitch-ovn-db-XXpod.Remove an old Northboud database member:
Identify the member to be removed:
ovs-appctl -t /var/run/ovn/ovnnb_db.ctl cluster/status OVN_Northbound
Example of system response:
5d02 Name: OVN_Northbound Cluster ID: 4d61 (4d61fde5-6cd5-449e-9846-34fcb470687b) Server ID: 5d02 (5d022977-982b-4de7-b125-e679746ece8d) Address: tcp:openvswitch-ovn-db-0.ovn-discovery.openstack.svc.cluster.local:6643 Status: cluster member Role: follower Term: 5402 Leader: c617 Vote: c617 Election timer: 10000 Log: [22917, 26535] Entries not yet committed: 0 Entries not yet applied: 0 Connections: ->c617 ->4d1e <-c617 <-0e28 Disconnections: 0 Servers: c617 (c617 at tcp:openvswitch-ovn-db-2.ovn-discovery.openstack.svc.cluster.local:6643) last msg 1153 ms ago 4d1e (4d1e at tcp:openvswitch-ovn-db-1.ovn-discovery.openstack.svc.cluster.local:6643) 0e28 (0e28 at tcp:openvswitch-ovn-db-1.ovn-discovery.openstack.svc.cluster.local:6643) last msg 109828 ms ago 5d02 (5d02 at tcp:openvswitch-ovn-db-0.ovn-discovery.openstack.svc.cluster.local:6643) (self)
In the above example output, the
4d1emember belongs to the failed node.Remove the old member:
ovs-appctl -t /var/run/ovn/ovnnb_db.ctl cluster/kick OVN_Northbound 4d1e sent removal request to leader
Verify that the old member has been removed successfully:
ovs-appctl -t /var/run/ovn/ovnnb_db.ctl cluster/status OVN_Northbound
Example of a successful system response:
5d02 Name: OVN_Northbound Cluster ID: 4d61 (4d61fde5-6cd5-449e-9846-34fcb470687b) Server ID: 5d02 (5d022977-982b-4de7-b125-e679746ece8d) Address: tcp:openvswitch-ovn-db-0.ovn-discovery.openstack.svc.cluster.local:6643 Status: cluster member Role: follower Term: 5402 Leader: c617 Vote: c617 Election timer: 10000 Log: [22917, 26536] Entries not yet committed: 0 Entries not yet applied: 0 Connections: ->c617 <-c617 <-0e28 ->0e28 Disconnections: 1 Servers: c617 (c617 at tcp:openvswitch-ovn-db-2.ovn-discovery.openstack.svc.cluster.local:6643) last msg 3321 ms ago 0e28 (0e28 at tcp:openvswitch-ovn-db-1.ovn-discovery.openstack.svc.cluster.local:6643) last msg 134877 ms ago 5d02 (5d02 at tcp:openvswitch-ovn-db-0.ovn-discovery.openstack.svc.cluster.local:6643) (self)
Remove an old Southbound database member by following the same steps used to remove an old Northbound database member:
Identify the member to be removed:
ovs-appctl -t /var/run/ovn/ovnsb_db.ctl cluster/status OVN_Southbound
Remove the old member:
ovs-appctl -t /var/run/ovn/ovnsb_db.ctl cluster/kick OVN_Southbound <SERVER-ID>
Add a compute node¶
This section describes how to add a new compute node to your existing Mirantis OpenStack for Kubernetes deployment.
To add a compute node:
Add a bare metal host to the MOSK cluster as described in Add a bare metal host.
Create a Kubernetes machine in your cluster as described in Add a machine.
When adding the machine, specify the node labels as required for an OpenStack compute node:
OpenStack node roles¶ Node role
Description
Kubernetes labels
Minimal count
OpenStack control plane
Hosts the OpenStack control plane services such as database, messaging, API, schedulers, conductors, L3 and L2 agents.
openstack-control-plane=enabledopenstack-gateway=enabledopenvswitch=enabled3
OpenStack compute
Hosts the OpenStack compute services such as libvirt and L2 agents.
openstack-compute-node=enabledopenvswitch=enabled(for a deployment with Open vSwitch as a backend for networking)Varies
If required, configure the compute host to enable DPDK, huge pages, SR-IOV, and other advanced features in your MOSK deployment. See Advanced OpenStack configuration (optional) for details.
Once the node is available in Kubernetes and when the
nova-computeandneutronpods are running on the node, verify that the compute service and Neutron Agents are healthy in OpenStack API.In the
keystone-clientpod, run:openstack network agent list --host <cmp_host_name> openstack compute service list --host <cmp_host_name>
Verify that the compute service is mapped to cell.
The OpenStack Controller triggers the
nova-cell-setupjob once it detects a new compute pod in theReadystate. This job sets mapping for new compute services to cells.In the
nova-api-osapipod, run:nova-manage cell_v2 list_hosts | grep <cmp_host_name>
Change oversubscription settings for existing compute nodes¶
Available since MOSK 23.1
MOSK enables you to control the oversubscription of compute node resources through the placement service API.
To manage the oversubscription through the placement API:
Obtain the host name of the hypervisor in question:
openstack hypervisor list -f yaml
Example of system response:
- Host IP: 10.10.0.78 Hypervisor Hostname: ps-ps-obnqilm4xxlu-0-gdy3x46euaeu-server-ftp6p7j6pyjl.cluster.local Hypervisor Type: QEMU ID: 1 State: up - Host IP: 10.10.0.118 Hypervisor Hostname: ps-ps-obnqilm4xxlu-1-n36gax6zqgef-server-xtby2leuercd.cluster.local Hypervisor Type: QEMU ID: 7 State: up
Determine the resource provider that corresponds to the hypervisor:
openstack resource provider list -f yaml --name <hypervisor_hostname>
Example of system response:
- generation: 4 name: ps-ps-obnqilm4xxlu-1-n36gax6zqgef-server-xtby2leuercd.cluster.local parent_provider_uuid: null root_provider_uuid: b16e9094-3f0e-4b8e-a138-e0b1f0a980db uuid: b16e9094-3f0e-4b8e-a138-e0b1f0a980db
Verify the current values in the resource provider by its UUID:
openstack resource provider inventory list <provider_uuid> -f yaml
Example of system response:
- allocation_ratio: 8.0 max_unit: 8 min_unit: 1 reserved: 0 resource_class: VCPU step_size: 1 total: 8 used: 0 - allocation_ratio: 1.0 max_unit: 7956 min_unit: 1 reserved: 512 resource_class: MEMORY_MB step_size: 1 total: 7956 used: 0 - allocation_ratio: 1.6 max_unit: 145 min_unit: 1 reserved: 0 resource_class: DISK_GB step_size: 1 total: 145 used: 0
Update the allocation ratio for the required resource class in the resource provider and inspect the system response to verify that the change has been applied:
openstack resource provider inventory set <provider_uuid> --amend --resource VCPU:allocation_ratio=10
Caution
To ensure accurate resource updates, it is crucial to specify the
--amendargument when making requests. Failure to do so will require the inclusion of values for all fields associated with the resource provider.Example of system response:
- allocation_ratio: 10.0 max_unit: 8 min_unit: 1 reserved: 0 resource_class: VCPU step_size: 1 total: 8 used: 0 - allocation_ratio: 1.0 max_unit: 7956 min_unit: 1 reserved: 512 resource_class: MEMORY_MB step_size: 1 total: 7956 used: 0 - allocation_ratio: 1.6 max_unit: 145 min_unit: 1 reserved: 0 resource_class: DISK_GB step_size: 1 total: 145 used: 0
Delete a compute node¶
Since MOSK 23.2, the OpenStack-related metadata is automatically removed during the graceful machine deletion through the Mirantis Container Cloud web UI. For the procedure, refer to Delete a cluster machine.
During the graceful machine deletion, the OpenStack Controller (Rockoon) performs the following operations:
Disables the OpenStack Compute and Block Storage services on the node to prevent further scheduling of workloads to it.
Verifies if any resources are present on the node, for example, instances and volumes. By default, the OpenStack Controller blocks the removal process until the resources are removed by the user. To adjust this behavior to the needs of your cluster, refer to OpenStack Controller configuration.
Removes OpenStack services metadata including compute services, Neutron agents, and volume services.
Caution
You cannot collocate the OpenStack compute node with other cluster components, such as Ceph. If done so, refer to the removal steps of the collocated components when planning the maintenance window.
If your cluster runs MOSK 23.1 or older version, perfrom the following steps before you remove the node from the cluster through the web UI to correctly remove the OpenStack-related metadata from it:
Disable the compute service to prevent spawning of new instances. In the
keystone-clientpod, run:openstack compute service set --disable <cmp_host_name> nova-compute --disable-reason "Compute is going to be removed."
Migrate all workloads from the node. For more information, follow Nova official documentation: Migrate instances.
Ensure that there are no pods running on the node to delete by draining the node as instructed in the Kubernetes official documentation: Safely drain node.
Delete the compute service using the OpenStack API. In the
keystone-clientpod, run:openstack compute service delete <service_id>
Note
To obtain
<service_id>, run:openstack compute service list --host <cmp_host_name>
Depending on the networking backend in use, proceed with one of the following:
Obtain the network agent ID:
openstack network agent list --host <cmp_host_name>
Delete the Neutron Agent service running the following command in the
keystone-clientpod:openstack network agent delete <agent_id>
Log in to the Tungsten Fabric web UI.
Navigate to Configure > Infrastructure > Virtual Routers.
Select the target compute node.
Click Delete.
Reschedule stateful applications¶
Note
The procedure applies to the MOSK clusters running MOSK 23.3 series or earlier versions. Starting from 24.1, MOSK performs the rescheduling of stateful applications automatically.
The rescheduling of stateful applications may be required when replacing a permanently failed node, decommissioning a node, migrating applications to nodes with a more suitable set of hardware, and in several other use cases.
MOSK deployment profiles include the following stateful applications:
OpenStack database (MariaDB)
OpenStack coordination (etcd)
OpenStack Time Series Database backend (Redis)
Each stateful application from the list above has a persistent volume claim (PVC) based on a local persistent volume per pod. Each of control plane nodes has a set of local volumes available. To migrate an application pod to another node, recreate a PVC with the persistent volume from the target node.
Caution
A stateful application pod can only be migrated to a node that does not contain other pods of this application.
Caution
When a PVC is removed, all data present in the related persistent volume is removed from the node as well.
Reschedule pods to another control plane node¶
This section describes how to reschedule pods for MariaDB, etcd, and Redis to another control plane node.
Important
Perform the pods rescheduling if you have to move a PVC to another node and the current node is still present in the cluster. If the current node has been removed already, MOSK reschedules pods automatically when a node with required labels is present in the cluster.
Recreate PVCs as described in Recreate a PVC on another control plane node.
Remove the pod:
Note
To remove a pod from a node in the
NotReadystate, add--grace-period=0 --forceto the following command.kubectl -n openstack delete pod <STATEFULSET-NAME>-<NUMBER>
Wait until the pod appears in the
Readystate.When the rescheduling is finalized, the
<STATEFULSET-NAME>-<NUMBER>pod rejoins the Galera cluster with a clean MySQL data directory and requests the Galera state transfer from the available nodes.
Important
Perform the pods rescheduling if you have to move a PVC to another node and the current node is still present in the cluster. If the current node has been removed already, MOSK reschedules pods automatically when a node with required labels is present in the cluster.
Recreate PVCs as described in Recreate a PVC on another control plane node.
Remove the pod:
Note
To remove a pod from a node in the
NotReadystate, add--grace-period=0 --forceto the following command.kubectl -n openstack-redis delete pod <STATEFULSET-NAME>-<NUMBER>
Wait until the pod is in the
Readystate.
Warning
During the reschedule procedure of the etcd LCM, a short cluster downtime is expected.
Before MOSK 23.1:
Identify the etcd replica ID that is a numeric suffix in a pod name. For example, the ID of the
etcd-etcd-0pod is0. This ID is required during the reschedule procedure.kubectl -n openstack get pods | grep etcd
Example of a system response:
etcd-etcd-0 0/1 Pending 0 3m52s etcd-etcd-1 1/1 Running 0 39m etcd-etcd-2 1/1 Running 0 39m
If the replica ID is
1or higher:Add the
coordinationsection to thespec.servicessection of the OsDpl object:spec: services: coordination: etcd: values: conf: etcd: ETCD_INITIAL_CLUSTER_STATE: existing
Wait for the etcd
statefulSetto update the new state parameter:kubectl -n openstack get sts etcd-etcd -o jsonpath='{.spec.template.spec.containers[0].env[?(@.name=="ETCD_INITIAL_CLUSTER_STATE")].value}'
Scale down the etcd
StatefulSetto0replicas. Verify that no replicas are running on the failed node.kubectl -n openstack scale sts etcd-etcd --replicas=0
Select from the following options:
If the current node is still present in the cluster and the PVC should be moved to another node, recreate the PVC as described in Recreate a PVC on another control plane node.
If the current node has been removed, remove the PVC related to the etcd replica of the failed node:
kubectl -n <NAMESPACE> delete pvc <PVC-NAME>
The PVC will be recreated automatically after the etcd
StatefulSetis scaled to the initial number of replicas.
Scale the etcd
StatefulSetto the initial number of replicas:kubectl -n openstack scale sts etcd-etcd --replicas=<NUMBER-OF-REPLICAS>
Wait until all etcd pods are in the
Readystate.Verify that the etcd cluster is healthy:
kubectl -n openstack exec -t etcd-etcd-1 -- etcdctl -w table endpoint --cluster status
Before MOSK 23.1, if the replica ID is
1or higher:Remove the
coordinationsection from thespec.servicessection of the OsDpl object.Wait until all etcd pods appear in the
Readystate.Verify that the etcd cluster is healthy:
kubectl -n openstack exec -t etcd-etcd-1 -- etcdctl -w table endpoint --cluster status
Recreate a PVC on another control plane node¶
This section describes how to recreate a PVC of a stateful application on another control plane node.
To recreate a PVC on another control plane node:
Select one of the persistent volumes available on the node:
Caution
A stateful application pod can only be migrated to the node that does not contain other pods of this application.
NODE_NAME=<NODE-NAME> STORAGE_CLASS=$(kubectl -n openstack get osdpl <OSDPL_OBJECT_NAME> -o jsonpath='{.spec.local_volume_storage_class}') kubectl -n openstack get pv -o json | jq --arg NODE_NAME $NODE_NAME --arg STORAGE_CLASS $STORAGE_CLASS -r '.items[] | select(.spec.nodeAffinity.required.nodeSelectorTerms[0].matchExpressions[0].values[0] == $NODE_NAME and .spec.storageClassName == $STORAGE_CLASS and .status.phase == "Available") | .metadata.name'
As the new PVC should contain the same parameters as the deleted one except for
volumeName, save the old PVC configuration in YAML:kubectl -n <NAMESPACE> get pvc <PVC-NAME> -o yaml > <OLD-PVC>.yaml
Note
<NAMESPACE>is a Kubernetes namespace where the PVC is created. For Redis, specifyopenstack-redis, for other applications specifyopenstack.Delete the old PVC:
kubectl -n <NAMESPACE> delete pvc <PVC-NAME>
Note
If a PVC has stuck in the
terminatingstate, runkubectl -n openstack edit pvc <PVC-NAME>and remove thefinalizerssection frommetadataof the PVC.Create a PVC with a new persistent volume:
cat <<EOF | kubectl apply -f - apiVersion: v1 kind: PersistentVolumeClaim metadata: name: <PVC-NAME> namespace: <NAMESPACE> spec: accessModes: - ReadWriteOnce resources: requests: storage: <STORAGE-SIZE> storageClassName: <STORAGE-CLASS> volumeMode: Filesystem volumeName: <PV-NAME> EOF
Caution
<STORAGE-SIZE>,<STORAGE-CLASS>, and<NAMESPACE>should correspond to thestorage,storageClassName, andnamespacevalues from the<OLD-PVC>.yamlfile with the old PVC configuration.
Run Tempest tests¶
The OpenStack Integration Test Suite (Tempest), is a set of integration tests to be run against a live OpenStack environment. This section instructs you on how to verify the workability of your OpenStack deployment using Tempest.
To verify an OpenStack deployment using Tempest:
Configure the Tempest run parameters using the
features:services:tempeststructure in theOpenStackDeploymentcustom resource.Note
To perform the smoke testing of your deployment, no additional configuration is required.
Configuration examples:
To perform the full Tempest testing:
spec: services: tempest: tempest: values: conf: script: | tempest run --config-file /etc/tempest/tempest.conf --concurrency 4 --blacklist-file /etc/tempest/test-blacklist --regex test
To set the image build timeout to
600:spec: services: tempest: tempest: values: conf: tempest: image: build_timeout: 600
Run Tempest. The OpenStack Tempest is deployed like other OpenStack services in a dedicated
openstack-tempestHelm release by addingtempesttospec:features:servicesin theOpenStackDeploymentcustom resource:spec: features: services: - tempest
Wait until Tempest is ready. The Tempest tests are launched by the
openstack-tempest-run-testsjob. To keep track of the tests execution, run:kubectl -n openstack logs -l application=tempest,component=run-tests
Get the Tempest results. The Tempest results can be stored in a
pvc-tempestPersistentVolumeClaim (PVC). To get them from a PVC, use:# Run pod and mount pvc to it cat <<EOF | kubectl apply -f - apiVersion: v1 kind: Pod metadata: name: tempest-test-results-pod namespace: openstack spec: nodeSelector: openstack-control-plane: enabled volumes: - name: tempest-pvc-storage persistentVolumeClaim: claimName: pvc-tempest containers: - name: tempest-pvc-container image: ubuntu command: ['sh', '-c', 'sleep infinity'] volumeMounts: - mountPath: "/var/lib/tempest/data" name: tempest-pvc-storage EOF
If required, copy the results locally:
kubectl -n openstack cp tempest-test-results-pod:/var/lib/tempest/data/report_file.xml .
Remove the Tempest test results pod:
kubectl -n openstack delete pod tempest-test-results-pod
To rerun Tempest:
Remove Tempest from the list of enabled services.
Wait until Tempest jobs are removed.
Add Tempest back to the list of the enabled services.
Remove an OpenStack cluster¶
This section instructs you on how to remove an OpenStack cluster, deployed on
top of Kubernetes, by deleting the openstackdeployments.lcm.mirantis.com
(OsDpl) CR.
To remove an OpenStack cluster:
Verify that the OsDpl object is present:
kubectl get osdpl -n openstack
Delete the OsDpl object:
kubectl delete osdpl osh-dev -n openstack
The deletion may take a certain amount of time.
Verify that all pods and jobs have been deleted and no objects are present in the command output:
kubectl get pods,jobs -n openstack
Delete Persistent Volume Claims (PVCs) using the following snippet. Deletion of PVCs causes data deletion on Persistent Volumes. The volumes themselves will become available for further operations.
Caution
Before deleting PVCs, save valuable data in a safe place.
#!/bin/bash PVCS=$(kubectl get pvc --all-namespaces |egrep "openstack|openstack-redis" | egrep "redis|etcd|mariadb" | awk '{print $1" "$2" "$4}'| column -t ) echo "$PVCS" | while read line; do PVC_NAMESPACE=$(echo "$line" | awk '{print $1}') PVC_NAME=$(echo "$line" | awk '{print $2}') echo "Deleting PVC ${PVC_NAME}" kubectl delete pvc ${PVC_NAME} -n ${PVC_NAMESPACE} done
Note
Deletion of PVCs may get stuck if a resource that uses the PVC is still running. Once the resource is deleted, the PVC deletion process will proceed.
Delete the MariaDB state ConfigMap:
kubectl delete configmap openstack-mariadb-mariadb-state -n openstack
Delete secrets using the following snippet:
#!/bin/bash SECRETS=$(kubectl get secret -n openstack | awk '{print $1}'| column -t | awk 'NR>1') echo "$SECRETS" | while read line; do echo "Deleting Secret ${line}" kubectl delete secret ${line} -n openstack done
Verify that OpenStack ConfigMaps and secrets have been deleted:
kubectl get configmaps,secrets -n openstack
Remove an OpenStack service¶
OpenStack Controller (Rockoon)
Since MOSK 25.1, the OpenStack Controller has been open-sourced under the name Rockoon and is maintained as an independent open-source project going forward.
As part of this transition, all openstack-controller pods are named
rockoon pods across the MOSK documentation and deployments. This change
does not affect functionality, but this is the reminder for the users to
utilize the new naming for pods and other related artifacts accordingly.
This section instructs you on how to remove an OpenStack service deployed on
top of Kubernetes. A service is typically removed by deleting a corresponding
entry in the spec.features.services section of the
openstackdeployments.lcm.mirantis.com (OsDpl) CR.
Caution
You cannot remove the default services built into the preset
section.
Remove a service¶
Verify that the
spec.features.servicessection is present in the OsDpl object:kubectl -n openstack get osdpl osh-dev -o jsonpath='{.spec.features.services}'
Example of system output:
[instance-ha object-storage]
Obtain the user name of the service database that will be required during Clean up OpenStack database leftovers after the service removal to substitute
SERVICE-DB-NAME:Note
For example, the
<SERVICE-NAME>for theinstance-haservice type ismasakari.kubectl -n osh-system exec -t <ROCKOON-POD-NAME> -- helm3 -n openstack get values openstack-<SERVICE-NAME> -o json | jq -r .endpoints.oslo_db.auth.<SERVICE-NAME>.username
Delete the service from the
spec.features.servicessection of the OsDpl CR:kubectl -n openstack edit osdpl osh-dev
The deletion may take a certain amount of time.
Verify that all related objects have been deleted and no objects are present in the output of the following command:
for i in $(kubectl api-resources --namespaced -o name | grep -v event); do kubectl -n openstack get $i 2>/dev/null | grep <SERVICE-NAME>; done
Clean up OpenStack API leftovers after the service removal¶
Log in to the Keystone client pod shell:
kubectl -n openstack exec -it <KEYSTONE-CLIENT-POD-NAME> -- bash
Remove service endpoints from the Keystone catalog:
for i in $(openstack endpoint list --service <SERVICE-NAME> -f value -c ID); do openstack endpoint delete $i; done
Remove the service user from the Keystone catalog:
openstack user list --project service | grep <SERVICE-NAME> openstack user delete <SERVICE-USER-ID>
Remove the service from the catalog:
openstack service list | grep <SERVICE-NAME> openstack service delete <SERVICE-ID>
Clean up OpenStack database leftovers after the service removal¶
Caution
The procedure below will permanently destroy the data of the removed service.
Log in to the
mariadb-serverpod shell:kubectl -n openstack exec -it mariadb-server-0 -- bash
Remove the service database user and its permissions:
Note
Use the user name for the service database obtained during the Remove a service procedure to substitute
SERVICE-DB-NAME:mysql -u root -p${MYSQL_DBADMIN_PASSWORD} -e "REVOKE ALL PRIVILEGES, GRANT OPTION FROM '<SERVICE-DB-USERNAME>'@'%';" mysql -u root -p${MYSQL_DBADMIN_PASSWORD} -e "DROP USER '<SERVICE-DB-USERNAME>'@'%';"
Remove the service database:
mysql -u root -p${MYSQL_DBADMIN_PASSWORD} -e "DROP DATABASE <SERVICE-NAME>;"
Enable uploading of an image through Horizon with untrusted SSL certificates¶
By default, the OpenStack Dashboard (Horizon) is configured to load images directly into Glance. However, if a MOSK cluster is deployed using untrusted certificates for public API endpoints and Horizon, uploading of images to Glance through the Horizon web UI may fail.
When accessing the Horizon web UI of such MOSK deployment
for the first time, a warning informs you that the site is insecure and you
must force trust the certificate of this site. However, when trying to upload
an image directly from a web browser, the certificate of the Glance API is
still not considered by the web browser as a trusted one since host:port
of the site is different. In this case, you must explicitly trust the
certificate of the Glance API.
To enable uploading of an image through Horizon with untrusted SSL certificates:
Navigate to the Horizon web UI.
Configure your web browser to trust the Horizon certificate if you have not done so yet:
In Google Chrome or Chromium, click Advanced > Proceed to <URL> (unsafe).
In Mozilla Firefox, navigate to Advanced > Add Exception, enter the URL in the Location field, and click Confirm Security Exception.
Note
For other web browsers, the steps may vary slightly.
Navigate to Project > API Access.
Copy the Service Endpoint URL of the Image service.
Open this URL in a new window or tab of the same web browser.
Configure your web browser to trust the certificate of this site as described in the step 2.
As a result, the version discovery document should appear with contents that varies depending on the OpenStack version. For example, for OpenStack Victoria:
{"versions": [{"id": "v2.9", "status": "CURRENT", "links": \ [{"rel": "self", "href": "https://glance.ic-eu.ssl.mirantis.net/v2/"}]}, \ {"id": "v2.7", "status": "SUPPORTED", "links": \ [{"rel": "self", "href": "https://glance.ic-eu.ssl.mirantis.net/v2/"}]}, \ {"id": "v2.6", "status": "SUPPORTED", "links": \ [{"rel": "self", "href": "https://glance.ic-eu.ssl.mirantis.net/v2/"}]}, \ {"id": "v2.5", "status": "SUPPORTED", "links": \ [{"rel": "self", "href": "https://glance.ic-eu.ssl.mirantis.net/v2/"}]}, \ {"id": "v2.4", "status": "SUPPORTED", "links": \ [{"rel": "self", "href": "https://glance.ic-eu.ssl.mirantis.net/v2/"}]}, \ {"id": "v2.3", "status": "SUPPORTED", "links": \ [{"rel": "self", "href": "https://glance.ic-eu.ssl.mirantis.net/v2/"}]}, \ {"id": "v2.2", "status": "SUPPORTED", "links": \ [{"rel": "self", "href": "https://glance.ic-eu.ssl.mirantis.net/v2/"}]}, \ {"id": "v2.1", "status": "SUPPORTED", "links": \ [{"rel": "self", "href": "https://glance.ic-eu.ssl.mirantis.net/v2/"}]}, \ {"id": "v2.0", "status": "SUPPORTED", "links": \ [{"rel": "self", "href": "https://glance.ic-eu.ssl.mirantis.net/v2/"}]}]}
Once done, you should be able to upload an image through Horizon with untrusted SSL certificates.
Rotate OpenStack credentials¶
OpenStack Controller (Rockoon)
Since MOSK 25.1, the OpenStack Controller has been open-sourced under the name Rockoon and is maintained as an independent open-source project going forward.
As part of this transition, all openstack-controller pods are named
rockoon pods across the MOSK documentation and deployments. This change
does not affect functionality, but this is the reminder for the users to
utilize the new naming for pods and other related artifacts accordingly.
The credential rotation procedure is designed to minimize the impact on service availability and workload downtime. It depends on the credential type and is based on the following principles:
Credentials for OpenStack admin database and messaging are immediately changed during one rotation cycle, without a transition period.
Credentials for OpenStack admin identity are rotated with a transition period of one extra rotation cycle. This means that the credentials become invalid after two rotations. MOSK exposes the latest valid credentials to the
openstack-externalnamespace. For details, refer to Access OpenStack through CLI from your local machine.Credentials for OpenStack service users, including those for messaging, identity, and database, undergo a transition period of one extra rotation cycle during rotation.
Note
If immediate inactivation of credentials is required, initiate the rotation procedure twice.
Impact on workloads availability¶
The restarts of the Networking service may cause workload downtimes. The exact lengths of these downtimes depend on the cloud density and scale.
Impact on APIs availability¶
Rotating both administrator and service credentials can potentially result in certain API operations failing.
Rotation prerequisites¶
Verify that the current
stateof the LCM action inOpenstackDeploymentStatusisAPPLIED:kubectl -n openstack get osdplst -o yaml
Example of an expected system response:
1 kind: OpenStackDeploymentStatus 2 metadata: 3 name: osh-dev 4 namespace: openstack 5 spec: {} 6 status: 7 ... 8 osdpl: 9 cause: update 10 changes: '((''add'', (''status'',), None, {''watched'': {''ceph'': {''secret'': 11 {''hash'': ''0fc01c5e2593bc6569562b451b28e300517ec670809f72016ff29b8cbaf3e729''}}}}),)' 12 controller_version: 0.5.3.dev12 13 fingerprint: a112a4a7d00c0b5b79e69a2c78c3b50b0caca76a15fe7d79a6ad1305b19ee5ec 14 openstack_version: ussuri 15 state: APPLIED 16 timestamp: "2021-09-08 17:01:45.633143"
Verify that there are no other LCM operations running on the OpenStack cluster.
Thoroughly plan the maintenance window taking into account the following considerations:
All OpenStack control plane services, components of the Networking service (OpenStack Neutron) responsible for the data plane and messaging services are restarted during service credentials rotation.
OpenStack database and OpenStack messaging services are restarted during administrator credentials rotation, as well as some of the Openstack control plane services, including the Instance High Availability service (OpenStack Masakari), Dashboard (OpenStack Horizon), and Identity service (OpenStack Keystone).
For approximate maintenance window duration, refer to Calculate a maintenance window duration for update Deprecated.
Rotate the credentials¶
Log in to the
osdplcontainer in therockoonpod:kubectl -n osh-system exec -it <rockoon-pod> -c osdpl -- bash
Use the osctl utility to trigger credentials rotation:
osctl credentials rotate --osdpl <osdpl-object-name> --type <credentials-type>
Where the
<credentials-type>value is eitheradminorservice.Note
Mirantis recommends rotating both admin and service credentials simultaneously to decrease the duration of the maintenance window and number of service restarts. You can do this by passing the
--typeargument twice:osctl credentials rotate --osdpl <osdpl-object-name> --type service --type admin
Wait until the
OpenStackDeploymentStatusobject has stateAPPLIEDand all OpenStack components in thehealthgroup in theOpenStackDeploymentStatuscustom resource are in theReadystate.Alternatively, you can launch the rotation command with the
--waitflag.
Now, the latest admin password for your OpenStack environment is available in
the openstack-identity-credentials secret in the openstack-external
namespace.
Customize OpenStack container images¶
This section provides instructions on how to customize the functionality of your MOSK OpenStack services by installing custom system or Python packages into their container images.
The MOSK services are running in Ubuntu-based containers, which can be extended to meet specific requirements or implement specific use cases, for example:
Enabling third-party storage driver for OpenStack Cinder
Implementing a custom scheduler for OpenStack Nova
Adding a custom dashboard to OpenStack Horizon
Building your own image importing workflow for OpenStack Glance
Warning
Mirantis cannot be held responsible for any consequences arising from using customized container images. Mirantis does not provide support for such actions, and any modifications to production systems are made at your own risk.
Note
Custom images are pinned in the OpenStackDeployment custom
resource. These images do not undergo automatic updates or upgrades.
Cloud administrator is responsible for image update during OpenStack
updating and upgrading.
Build a custom container image¶
Create a new directory and switch to it:
mkdir my-custom-image cd my-custom-image
Create a Dockerfile:
touch DockerfileSpecify the location for the base image in the Dockerfile.
A custom image can be derived from any OpenStack image shipped with MOSK. For locations of the images comprising a specific MOSK release, refer to a corresponding release artifacts page in the Release Notes.
ARG FROM=<images-base-url>/openstack/<image-name>:<tag> FROM $FROM
Note
Presuming the custom image will need to get rebuilt for every new MOSK release, Mirantis recommends parametrizing the location of its base by introducing the
$FROMargument to the Dockerfile.Instruct the Dockerfile to install additional system packages:
RUN apt-get update ;\ apt-get install --no-install-recommends -y <package name> ;\ apt-get clean -y
If you need to install packages from a third-party repository:
Make sure that the
add-apt-repositoryutility is installed:RUN apt-get install --no-install-recommends -y software-properties-common
Add the third-party repository:
RUN add-apt-repository <repository_name>
Instruct the Dockerfile to install additional Python packages:
Caution
Rules to comply with when extending MOSK container images with Python packages:
Use only Python wheel packaging standard, the older
*.eggpackage type is not supportedHonor upper constraints that MOSK defines for its OpenStack packages prerequisites
OpenStack components in every MOSK release are shipped together with their requirements packaged as Python wheels and constraints file. Download and extract these artifacts from the corresponding
requirementscontainer image, so that they can be used for building your packages as well. Use therequirementsimage with the same tag as the base image that you plan to customize:docker pull <images-base-url>/openstack/requirements:<tag> docker save -o requirements.tar <images-base-url>/openstack/requirements:<tag> mkdir requirements tar -xf requirements.tar -C requirements tar -xf requirements/<shasum>/layer.tar -C requirements
Build your Python wheel packages using one of the commands below depending on the place where the source code is stored:
Build from source:
pip wheel --no-binary <package> <package> --wheel-dir=custom-wheels -c requirements/dev-upper-constraints.txt
Build from an upstream pip repository:
pip wheel <package> --wheel-dir=custom-wheels -c requirements/dev-upper-constraints.txt
Build from a custom repository:
pip wheel <package> --extra-index-url <Repo-URL> --wheel-dir=custom-wheels -c requirements/dev-upper-constraints.txt
Include the built custom wheel packages and packages for the requirements into the Dockerfile:
COPY custom-wheels /tmp/custom-wheels COPY requirements /tmp/wheels
Install the necessary wheel packages to be stored along with other OpenStack components:
RUN source /var/lib/openstack/bin/activate ;\ pip install <package> --no-cache-dir --only-binary :all: --no-compile --find-links /tmp/wheels --find-links /tmp/custom-wheels -c /tmp/wheels/dev-upper-constraints.txt
Build the container image from your Dockerfile:
docker build --tag <user-name>/<repo-name>:<tag> .
When selecting the name for your image, Mirantis recommends following the common practice across major public Docker repositories, that is Docker Hub. The image name should be
<user-name>/<repo-name>, where<user-name>is a unique identifier of the user who authored it and<repo-name>is the name of the software shipped.Specify the current directory as the build context. Also, use the
--tagoption to assign the tag to your image. Assigning a tag:<tag>enables you to add multiple versions of the same image to the repository. Unless you assign a tag, it defaults tolatest.If you are adding Python packages, you can minimize the size of the custom image by building it with the
--squashflag. It merges all the image layers into one and instructs the system not to store the cache layers of the wheel packages.Verify that the image has been built and is present on your system:
docker image ls
Publish the image to the designated registry by its name and tag:
Note
Before pushing the image, make sure that you have authenticated with the registry using the docker login command.
docker push <user-name>/<repo-name>:<tag>
Attach a private Docker registry to MOSK Kubernetes underlay¶
To ensure that the Kubernetes worker nodes in your MOSK cluster can locate and download the custom image, it should be published to a container image registry that the cluster is configured to use.
To configure the MOSK Kubernetes underlay to use your
private registry, you need to create a ContainerRegistry resource in the
Mirantis Container Cloud API with the registry domain and CA certificate in it,
and specify the resource in the Cluster object that corresponds to
MOSK.
For the details, refer to Define a custom CA certificate for a private Docker registry and ContainerRegistry resource.
Inject a custom image into MOSK cluster¶
To inject a customized OpenStack container into your MOSK cluster:
Create a ConfigMap in the openstack namespace with the following
content, replacing <OPENSTACKDEPLOYMENT-NAME> with the name of your
OpenStackDeployment custom resource:
apiVersion: v1
kind: ConfigMap
metadata:
labels:
penstack.lcm.mirantis.com/watch: "true"
name: <OPENSTACKDEPLOYMENT-NAME>-artifacts
namespace: openstack
data:
caracal: |
horizon: <image_path>
Update the spec:services section in the OpenStackDeployment custom
resource to override the default location of the container image with your
own:
Open the
OpenStackDeploymentcustom resource for editing:kubectl -n openstack edit osdpl osh-dev
Specify the path to your custom image:
For MOSK Dashboard (OpenStack Horizon):
spec: services: dashboard: horizon: values: images: tags: horizon: <PATH-TO-IMAGE>
For the MOSK Block Storage service (OpenStack Cinder):
spec: services: block-storage: cinder: values: images: tags: <IMAGE-NAME>: <PATH-TO-IMAGE>
How to examples¶
To help you better understand the process, this section provides a few examples illustrating how to add various plugins to MOSK services.
Warning
Mirantis cannot be held responsible for any consequences arising from using storage drivers, plugins, or features that are not explicitly tested or documented with MOSK. Mirantis does not provide support for such configurations as a part of standard product subscription.
Troubleshoot orphaned resource allocations¶
Available since MOSK 24.3
OpenStack Controller (Rockoon)
Since MOSK 25.1, the OpenStack Controller has been open-sourced under the name Rockoon and is maintained as an independent open-source project going forward.
As part of this transition, all openstack-controller pods are named
rockoon pods across the MOSK documentation and deployments. This change
does not affect functionality, but this is the reminder for the users to
utilize the new naming for pods and other related artifacts accordingly.
Orphaned resource allocations are entries in the Placement database that track resource consumption, but the corresponding consumer (instance) no longer exists on the compute nodes. As a result, the Nova scheduler mistakenly believes that compute nodes have more resources allocated than they actually have.
For example, orphaned resource allocations may occur when an instance
is evacuated from a hypervisor while the related nova-compute service
is down.
This section provides instructions on how to resolve orphaned resource allocations in Nova if they are detected on compute nodes.
Detect orphaned allocations¶
Orphaned allocations are detected by the nova-placement-audit CronJob that
runs every four hours.
The osdpl-exporter service processes the nova-placement-audit CronJob
output and exports current number of orphaned allocations to StackLight as an
osdpl_nova_audit_orphaned_allocations metric. If the value of this metric
is greater than 0, StackLight raises a major alert
NovaOrphanedAllocationsDetected.
Collect logging data from the cluster¶
Obtain the mapping with IDs of resource providers and related orphaned consumers:
kubectl -n openstack logs -l application=nova,component=placement-audit -c placement-audit-report | \ jq .orphaned_allocations.detected
Example of a system response:
{ "12ed66d0-00d8-40e5-a28b-19cdecd2211d": [ { "consumer": "1e616d60-bc5b-436d-8d71-503d15de5c55", "resources": { "DISK_GB": 5, "MEMORY_MB": 512, "VCPU": 1 } } ] }
Obtain the list of the
nova-computeservices that have issues with orphaned allocations:Obtain the UUIDs of the resource providers containing orphaned allocations:
rp_uuids=$(kubectl -n openstack logs -l application=nova,component=placement-audit -c placement-audit-report | \ jq -c '.orphaned_allocations.detected|keys')
Obtain the hostnames of the compute nodes that correspond to the resource providers obtained in the previous step:
cmp_fqdns=$(kubectl -n openstack exec -t deployment/keystone-client -- \ openstack resource provider list -f json | \ jq --argjson rp_uuids "$rp_uuids" -c ' .[] | select( [.uuid] | inside($rp_uuids) ) | .name') cmp_hostnames=$(for n in $(echo ${cmp_fqdns} | tr -d \"); do echo ${n%%.*}; done)
List the
nova-computeservices that contain orphaned allocations:kubectl -n openstack exec -t deployment/keystone-client -- \ openstack compute service list --service nova-compute --long -f json | \ jq --arg hosts "$cmp_hostnames" -r '.[] | select( .Host | inside($hosts) )'
Example of a system response:
[{ "ID": "14a1685a-798e-40f1-b490-a09a5c8f6f66", "Binary": "nova-compute", "Host": "mk-ps-7bqjjdq7o53q-0-rlocum3rumf4-server-ir4n3ag33erk", "Zone": "nova", "Status": "enabled", "State": "down", "Updated At": "2024-08-12T07:52:46.000000", "Disabled Reason": null, "Forced Down": true }]
Analyze the list of the
nova-computeservices obtained during the previous step:For the
nova-computeservices in thedownstate, most probably there were evacuations of instances from the correspoding nodes when the services were down. If this is the case, proceed directly to Remove orphaned allocations. Otherwise, proceed with collecting the logs.To verify if the evacuations were performed:
openstack server migration list --type evacuation --host <CMP_HOSTNAME> -f json
Example of a system response:
[{ "Id": 3, "UUID": "d7c29e99-2f69-4f85-80ed-72e1ef71c099", "Source Node": "mk-ps-7bqjjdq7o53q-1-snwv3ahiip6i-server-3axzquptwsao.cluster.local", "Dest Node": "mk-ps-7bqjjdq7o53q-0-rlocum3rumf4-server-ir4n3ag33erk.cluster.local", "Source Compute": "mk-ps-7bqjjdq7o53q-1-snwv3ahiip6i-server-3axzquptwsao", "Dest Compute": "mk-ps-7bqjjdq7o53q-0-rlocum3rumf4-server-ir4n3ag33erk", "Dest Host": "10.10.0.61", "Status": "completed", "Server UUID": "1e616d60-bc5b-436d-8d71-503d15de5c55", "Old Flavor": null, "New Flavor": null, "Type": "evacuation", "Created At": "2024-08-07T09:01:08.000000", "Updated At": "2024-08-07T16:11:54.000000" }]
For the
nova-computeservices in theUPstate, proceed with collecting the logs.
Collect the following logs from the environment:
Caution
The log data can be significant in size. Ensure that there is sufficient space available in the
/tmp/directory of the OpenStack Controller (Rockoon) pod. Create a separate report for each node.Logs from compute nodes for a 3-day period around the time of the alert:
From the node with the orphaned allocation
From the node with the actual allocation (where the instance exists, if any)
kubectl -n osh-system exec -it deployment/rockoon -- bash osctl sos --between <REPORT_PERIOD_TIMESTAMPS> \ --host <CMP_HOSTNAME> \ --component nova \ --collector elastic \ --collector nova \ --workspace /tmp/ report
For example, if the alert was raised on 2024-08-12, set
<REPORT_PERIOD_TIMESTAMPS>to2024-08-11,2024-08-13.Logs from the
nova-scheduler,nova-api,nova-conductor,placement-apipods for a 3-day period around the time of the alert:ctl_nodes=$(kubectl get nodes -l openstack-control-plane=enabled -o name) kubectl -n osh-system exec -it deployment/rockoon -- bash # for each node in ctl_nodes execute: osctl sos --between <REPORT_PERIOD_TIMESTAMPS> \ --host <CTL_HOSTNAME> \ --component nova \ --component placement \ --collector elastic \ --workspace /tmp/ report
Logs from the Kubernetes objects:
kubectl -n osh-system exec -it deployment/rockoon -- bash osctl sos --collector k8s --workspace /tmp/ report
Nova service data from the API:
kubectl -n openstack exec -it deployment/keystone-client -- bash openstack server migration list openstack compute service list --long openstack resource provider list # Get the server event list for each orphaned consumer id openstack server event list <SERVER_ID>
Note
SERVER_IDis the orphaned consumer ID from thenova-placement-auditlogs.
Create a support case and attach the obtained information.
Remove orphaned allocations¶
Log in to the
nova-api-osapipod:kubectl -n openstack exec -it deployment/nova-api-osapi -- bash
Remove orphaned allocations:
To remove all found orphaned allocations:
nova-manage placement audit --verbose --delete
To remove orphaned allocations on a specific resource provider:
nova-manage placement audit --verbose --delete --resource_provider <RESOURCE_PROVIDER_UUID>
Verify that no orphaned allocations exist:
nova-manage placement audit --verbose
Start monitoring IP address capacity¶
Available since MOSK 25.1
Note
The MOSK deployments with Tungsten Fabric do not support IP address capacity monitoring.
Monitoring IP address capacity helps cloud operators allocate routable IP addresses efficiently for dynamic workloads in the clouds. This capability provides insights for predicting future needs for IP addresses, ensuring seamless communication between workloads, users, and services while optimizing IP address usage.
By monitoring IP address capacity, cloud operators can:
Predict when to add new IP address blocks to prevent service disruptions.
Identify networks or subnets nearing capacity to prevent issues.
Optimize the allocation of costly external IP address pools.
To start monitoring IP address capacity in your cloud:
Verify that all required networks and subnets are monitored.
By default, MOSK monitors IP address capacity for the external networks that have the
router:external=Externalattribute and segmentation type ofvlanorflat.To include additional networks and subnets in the monitoring:
Tag the network with the
openstack.lcm.mirantis.com:prometheustag. When a network is tagged, all its subnets are automatically included in the monitoring:openstack network set <NETWORK-ID> --tag openstack.lcm.mirantis.com:prometheus
Tag individual subnets with the
openstack.lcm.mirantis.com:prometheustag. This includes the subnet in the monitoring regardless of the network tagging:openstack subnet set <SUBNET_ID> --tag openstack.lcm.mirantis.com:prometheus
View and analyze the monitoring data.
The metrics of IP address capacity are collected by the OpenStack Exporter and are available in Prometheus within the StackLight monitoring suite as:
osdpl_neutron_network_total_ipsosdpl_neutron_network_free_ipsosdpl_neutron_subnet_total_ipsosdpl_neutron_subnet_free_ips
See also
See also
OpenStack services configuration¶
This section covers post-deployment configuration of OpenStack services and
is intended for cloud operators responsible for maintaining a functional
cloud infrastructure for end users. It focuses on more complex procedures
that require additional steps beyond simply editing the OpenStackDeployment
custom resource.
For an overview of the capabilities provided by MOSK
OpenStack services and instructions on enabling and configuring them
at the OpenStackDeployment level, refer to Cloud services.
Configure high availability with Masakari¶
Instances High Availability Service or Masakari is an OpenStack project designed to ensure high availability of instances and compute processes running on hosts.
Before the end user can start enjoying the benefits of Masakari, the cloud operator has to configure the service properly. This section includes instructions on how to create segments and host through the Masakari API as well as provides the list of additional settings that can be useful in certain use cases.
Group compute nodes into segments¶
The segment object is a logical grouping of compute nodes into zones
also known as availability zones. The segment object enables
the cloud operator to list, create, show details for, update,
and delete segments.
To create a segment named allcomputes with service_type = compute,
and recovery_method = auto, run:
openstack segment create allcomputes auto compute
Example of a positive system response:
+-----------------+--------------------------------------+
| Field | Value |
+-----------------+--------------------------------------+
| created_at | 2021-07-06T07:34:23.000000 |
| updated_at | None |
| uuid | b8b0d7ca-1088-49db-a1e2-be004522f3d1 |
| name | allcomputes |
| description | None |
| id | 2 |
| service_type | compute |
| recovery_method | auto |
+-----------------+--------------------------------------+
Create hosts under segments¶
The host object represents compute service hypervisors. A host belongs
to a segment. The host can be any kind of virtual machine that has
compute service running on it. The host object enables the operator
to list, create, show details for, update, and delete hosts.
To create a host under a given segment:
Obtain the hypervisor hostname:
openstack hypervisor list
Example of a positive system response:
+----+-------------------------------------------------------+-----------------+------------+-------+ | ID | Hypervisor Hostname | Hypervisor Type | Host IP | State | +----+-------------------------------------------------------+-----------------+------------+-------+ | 2 | vs-ps-vyvsrkrdpusv-1-w2mtagbeyhel-server-cgpejthzbztt | QEMU | 10.10.0.39 | up | | 5 | vs-ps-vyvsrkrdpusv-0-ukqbpy2pkcuq-server-s4u2thvgxdfi | QEMU | 10.10.0.14 | up | +----+-------------------------------------------------------+-----------------+------------+-------+
Create the host under previously created segment. For example, with
uuid = b8b0d7ca-1088-49db-a1e2-be004522f3d1:Caution
The segment under which you create a host must exist.
openstack segment host create \ vs-ps-vyvsrkrdpusv-1-w2mtagbeyhel-server-cgpejthzbztt \ compute \ SSH \ b8b0d7ca-1088-49db-a1e2-be004522f3d1
Positive system response:
+---------------------+-------------------------------------------------------+ | Field | Value | +---------------------+-------------------------------------------------------+ | created_at | 2021-07-06T07:37:26.000000 | | updated_at | None | | uuid | 6f1bd5aa-0c21-446a-b6dd-c1b4d09759be | | name | vs-ps-vyvsrkrdpusv-1-w2mtagbeyhel-server-cgpejthzbztt | | type | compute | | control_attributes | SSH | | reserved | False | | on_maintenance | False | | failover_segment_id | b8b0d7ca-1088-49db-a1e2-be004522f3d1 | +---------------------+-------------------------------------------------------+
Enable notifications¶
The alerting API is used by Masakari monitors to notify about a failure
of either a host, process, or instance. The notification object enables
the operator to list, create, and show details of notifications.
Useful tunings¶
The list of useful tunings for the Masakari service includes:
[host_failure]\evacuate_all_instancesEnables the operator to decide whether to evacuate all instances or only the instances that have
[host_failure]\ha_enabled_instance_metadata_keyset toTrue. By default, the parameter is set toFalse.[host_failure]\ha_enabled_instance_metadata_keyEnables the operator to decide on the instance metadata key naming that affects the per instance behavior of
[host_failure]\evacuate_all_instances. The default is the same for both failure types, which includehostandinstance, but the value can be overridden to make the metadata key different per failure type.[host_failure]\ignore_instances_in_error_stateEnables the operator to decide whether error instances should be allowed for evacuation from a failed source compute node or not. If set to
True, it will ignore error instances from evacuation from a failed source compute node. Otherwise, it will evacuate error instances along with other instances from a failed source compute node.Available since MOSK 24.2
[host_failure]\ha_enabled_project_tagBy default, instances belonging to any project are evacuated. However, if the operator needs to restrict this functionality to specific projects, they can tag these projects with a designated tag and pass this tag as the value for this Masakari option. Consequently, instances from projects that do not have the specified tag are not considered for evacuation, even if they have the corresponding metadata key and value set.
[instance_failure]\process_all_instancesEnables the operator to decide whether all instances or only the ones that have
[instance_failure]\ha_enabled_instance_metadata_keyset toTrueshould be recovered from instance failure events. If set toTrue, it will execute instance failure recovery actions for an instance irrespective of whether that particular instance has[instance_failure]\ha_enabled_instance_metadata_keyset toTrueor not. Otherwise, it will only execute instance failure recovery actions for an instance which has[instance_failure]\ha_enabled_instance_metadata_keyset toTrue.[instance_failure]\ha_enabled_instance_metadata_keyEnables the operators to decide on the instance metadata key naming that affects the per-instance behavior of
[instance_failure]\process_all_instances. The default is the same for both failure types, which includehostandinstance, but you can override the value to make the metadata key different per failure type.
Configure monitoring of cloud workload availability¶
MOSK enables cloud operators to oversee the availability of workloads hosted in their OpenStack infrastructure through the monitoring of floating IP addresses availability (Cloudpprober) and network port availability (Portprober).
For the feature description and usage requirements, refer to Workload monitoring.
Configure floating IP address availability monitoring¶
Available since MOSK 23.2 TechPreview
MOSK allows you to monitor the floating IP address availability through the Cloudprober service. This section explains the details of the service configuration.
Enable the Cloudprober service in the
OpenStackDeploymentcustom resource:spec: features: services: - cloudprober
Wait untill the
OpenStackDeploymentstate becomesapplied:kubectl -n openstack get osdplst
Example of a positive system response:
NAME OPENSTACK VERSION CONTROLLER VERSION STATE osh-dev yoga 0.13.1.dev54 APPLIED
Verify that the Cloudprober service is running:
kubectl -n openstack get pods -l application=cloudprober
Example of a positive system response:
NAME READY STATUS RESTARTS AGE openstack-cloudprober-587b4bf7c4-lwmxx 2/2 Running 2 3d1h openstack-cloudprober-587b4bf7c4-v9tt9 2/2 Running 0 3d1h
Verify that the Cloudprober service is sending data to StackLight:
Log in to the StackLight Prometheus web UI.
Navigate to Status - Targets.
Search for the
openstack-cloudprobertarget and verify that it isUP.
By default, for outgoing traffic, the IP address for the Cloudprober Pod is translated to the node IP address. In this procedure, we assume no further translation of that node IP address on the path between the node and floating network.
Identify the node IP address used for traffic destined to floating network by selecting the IP address from the floating network and running the following command on each OpenStack control plane node:
ip r get <floating ip> | grep -E -o '(src .*)' | awk '{print $2}'
In the project where monitored virtual machines are running, create a security group:
openstack security group create --project <project_id> instance-monitoring
Create the rule for each IP address you obtain in step 1:
openstack security group rule create --proto icmp --ingress --remote-ip <node ip> instance-monitoring
Log in to the
keystone-clientPod to assign theopenstack.lcm.mirantis.com:probertag to each instance to be added to monitoring:openstack --os-compute-api-version 2.26 server set --tag openstack.lcm.mirantis.com:prober <INSTANCE_ID>
Assign the
instance-monitoringsecurity group to the server:openstack server add security group <SERVER_ID> <SECURITY_GROUP_ID>
Verify that the instances have been added successfully.
Cloudprober uses auto-discovery of instances on periodic basis. Therefore, wait for the discovery interval to pass (defaults to 600 seconds) and execute the following command inside the
keystone-clientPod:curl -s http://cloudprober.openstack.svc.cluster.local:9313/metrics | grep <INSTANCE_ID>
Example of a positive system response:
cloudprober_total{ptype="ping",probe="openstack-instances-icmp-probe",dst="d34a0c6b-91a2-4bd3-95ea-772da49b90c3-10.11.12.122",openstack_hypervisor_hostname="mk-ps-xp4m27lfl56j-1-w74pc7cinu67-server-42kx24m22xop.cluster.local",openstack_instance_id="d34a0c6b-91a2-4bd3-95ea-772da49b90c3",openstack_instance_name="test-vm-proj",openstack_project_id="1eb031db8add42fda2fdb0ef2c2ad8d7"} 266388 1685963215202 cloudprober_success{ptype="ping",probe="openstack-instances-icmp-probe",dst="d34a0c6b-91a2-4bd3-95ea-772da49b90c3-10.11.12.122",openstack_hypervisor_hostname="mk-ps-xp4m27lfl56j-1-w74pc7cinu67-server-42kx24m22xop.cluster.local",openstack_instance_id="d34a0c6b-91a2-4bd3-95ea-772da49b90c3",openstack_instance_name="test-vm-proj",openstack_project_id="1eb031db8add42fda2fdb0ef2c2ad8d7"} 266386 1685963215202 cloudprober_latency{ptype="ping",probe="openstack-instances-icmp-probe",dst="d34a0c6b-91a2-4bd3-95ea-772da49b90c3-10.11.12.122",openstack_hypervisor_hostname="mk-ps-xp4m27lfl56j-1-w74pc7cinu67-server-42kx24m22xop.cluster.local",openstack_instance_id="d34a0c6b-91a2-4bd3-95ea-772da49b90c3",openstack_instance_name="test-vm-proj",openstack_project_id="1eb031db8add42fda2fdb0ef2c2ad8d7"} 315484742.137 1685963215202 cloudprober_validation_failure{ptype="ping",probe="openstack-instances-icmp-probe",dst="d34a0c6b-91a2-4bd3-95ea-772da49b90c3-10.11.12.122",openstack_hypervisor_hostname="mk-ps-xp4m27lfl56j-1-w74pc7cinu67-server-42kx24m22xop.cluster.local",openstack_instance_id="d34a0c6b-91a2-4bd3-95ea-772da49b90c3",openstack_instance_name="test-vm-proj",openstack_project_id="1eb031db8add42fda2fdb0ef2c2ad8d7",validator="data-integrity"} 0 1685963215202
Note
You can adjust the instance auto-discovery interval in the
OpenStackDeploymentobject. However, Mirantis does not recommend setting it to too low values to avoid high load on the OpenStack API:spec: features: cloudprober: discovery: interval: 300
Now, you can start seeing the availability of instances floating IP addresses per OpenStack compute node and project, as well as viewing the probe statistics for individual instance floating IP addresses through the OpenStack Instances Availability dashboard in Grafana.
See also
Enable network port availability monitoring¶
Available since MOSK 24.2 TechPreview
MOSK allows you to monitor the network port availability through the Portprober service.
The Portprober service is enabled by default when the Cloudprober service is enabled as described above, on clouds running OpenStack Antelope or newer version and using Neutron OVS backend for networking.
Also, you can enable Portprober explicitly, regardless of whether Cloudprober is enabled or not. See Network port availability monitoring (Portprober) for details.
When the service is enabled, you can monitor the network port availability through the OpenStack PortProber dashboard in Grafana.
See also
Ceph operations¶
This section outlines Ceph LCM operations such as adding Ceph Monitor, Ceph nodes, and RADOS Gateway nodes to an existing Ceph cluster or removing them, as well as removing or replacing Ceph OSDs. The section also includes OpenStack-specific operations for Ceph.
The following sections describe the Ceph cluster configuration options:
Ceph default configuration options¶
Ceph Controller provides the capability to specify configuration options for
the Ceph cluster through the spec.cephClusterSpec.rookConfig key-value
parameter of the KaaSCephCluster resource as if they were set in a usual
ceph.conf file. For details, see Ceph advanced configuration.
However, if rookConfig is empty, Ceph Controller still specifies the
following default configuration options for each Ceph cluster:
Required network parameters that you can change through the
spec.cephClusterSpec.networksection:cluster network = <spec.cephClusterSpec.network.clusterNet> public network = <spec.cephClusterSpec.network.publicNet>
General default configuration options that you can override using the
rookConfigparameter:mon target pg per osd = 200 mon max pg per osd = 600 # Workaround configuration option to avoid the # https://github.com/rook/rook/issues/7573 issue # when updating to Rook 1.6.x versions: rgw_data_log_backing = omap
If rookConfig is empty but the spec.cephClsuterSpec.objectStore.rgw
section is defined, Ceph Controller specifies the following OpenStack-related
default configuration options for each Ceph cluster:
Ceph Object Gateway options, which you can override using the
rookConfigparameter:rgw swift account in url = true rgw keystone accepted roles = '_member_, Member, member, swiftoperator' rgw keystone accepted admin roles = admin rgw keystone implicit tenants = true rgw swift versioning enabled = true rgw enforce swift acls = true rgw_max_attr_name_len = 64 rgw_max_attrs_num_in_req = 32 rgw_max_attr_size = 1024 rgw_bucket_quota_ttl = 0 rgw_user_quota_bucket_sync_interval = 0 rgw_user_quota_sync_interval = 0 rgw s3 auth use keystone = true
Additional parameters for the Keystone integration:
Warning
All values with the
keystoneprefix are programmatically specified for each MOSK deployment. Do not modify these parameters manually.rgw keystone api version = 3 rgw keystone url = <keystoneAuthURL> rgw keystone admin user = <keystoneUser> rgw keystone admin password = <keystonePassword> rgw keystone admin domain = <keystoneProjectDomain> rgw keystone admin project = <keystoneProjectName>
See also
Ceph advanced configuration¶
This section describes how to configure a Ceph cluster through the
KaaSCephCluster (kaascephclusters.kaas.mirantis.com) CR during or
after the deployment of a MOSK cluster.
The KaaSCephCluster CR spec has two sections, cephClusterSpec and
k8sCluster and specifies the nodes to deploy as Ceph components. Based on
the roles definitions in the KaaSCephCluster CR, Ceph Controller
automatically labels nodes for Ceph Monitors and Managers. Ceph OSDs are
deployed based on the storageDevices parameter defined for each Ceph node.
For a default KaaSCephCluster CR, see Example of a complete template configuration for cluster creation.
Configure a Ceph cluster¶
Select from the following options:
If you do not have a cluster yet, open
kaascephcluster.yaml.templatefor editing.If the cluster is already deployed, open the
KaasCephClusterCR for editing:kubectl edit kaascephcluster -n <ClusterProjectName>
Substitute
<ClusterProjectName>with a corresponding value.
Using the tables below, configure the Ceph cluster as required.
Select from the following options:
If you are creating a cluster, save the updated
KaaSCephClustertemplate to the corresponding file and proceed with the cluster creation.If you are configuring
KaaSCephClusterof an existing cluster, exit the text editor to apply the change.
Ceph configuration options¶
Parameter |
Description |
|---|---|
|
Describes a Ceph cluster in the MOSK cluster. For details
on |
|
Defines the cluster on which the spec:
k8sCluster:
name: kaas-mgmt
namespace: default
|
Parameter |
Description |
|---|---|
|
Specifies networks for the Ceph cluster:
|
|
Specifies the list of Ceph nodes. For details, see
Node parameters. The nodes:
master-0:
<node spec>
master-1:
<node spec>
...
worker-0:
<node spec>
|
|
Specifies the list of Ceph nodes grouped by node lists or node
labels. For details, see NodeGroups parameters. The nodes:
group-1:
spec: <node spec>
nodes: ["master-0", "master-1"]
group-2:
spec: <node spec>
label: <nodeLabelExpression>
...
group-3:
spec: <node spec>
nodes: ["worker-2", "worker-3"]
The |
|
Specifies the list of Ceph pools. For details, see Pool parameters. |
|
Specifies the parameters for Object Storage, such as RADOS Gateway, the Ceph Object Storage. Also specifies the RADOS Gateway Multisite configuration. For details, see RADOS Gateway parameters and Multisite parameters. |
|
Optional. String key-value parameter that allows overriding Ceph configuration options. Since MOSK 24.2, use the The use of this option enables restart of only specific daemons related
to the corresponding section. If you do not specify the section,
a parameter is set in the For example: rookConfig:
"osd_max_backfills": "64"
"mon|mon_health_to_clog": "true"
"osd|osd_journal_size": "8192"
"osd.14|osd_journal_size": "6250"
|
|
Available since MOSK 23.3. Enables specification of
extra options for a setup, includes the |
|
In MOSK 25.1, is automatically replaced with |
|
Available since MOSK 25.1 to automatically replace the
|
|
Enables pools mirroring between two interconnected clusters. For details, see Enable Ceph RBD mirroring. |
|
List of Ceph clients. For details, see Clients parameters. |
|
Disables autogeneration of shared Ceph values for OpenStack
deployments. Set to |
|
Contains the
For example: mgr:
mgrModules:
- name: balancer
enabled: true
settings:
balancerMode: upmap
- name: pg_autoscaler
enabled: true
The Note Most Ceph Manager modules require additional configuration
that you can perform through the |
|
Configures health checks and liveness probe settings for Ceph daemons. For details, see HealthCheck parameters.
Example configurationspec:
cephClusterSpec:
network:
clusterNet: 10.10.10.0/24
publicNet: 10.10.11.0/24
nodes:
master-0:
<node spec>
...
pools:
- <pool spec>
...
rookConfig:
"mon max pg per osd": "600"
...
|
Parameter |
Description |
|---|---|
|
Specifies the
If a Ceph node contains a If a Ceph node contains a
|
|
Specifies the list of devices to use for Ceph OSD deployment. Includes the following parameters: Note Since MOSK 23.3, Mirantis recommends migrating
all For details, refer to Container Cloud documentation: Addressing storage devices.
|
|
Specifies the explicit key-value CRUSH topology for a node. For details, see Ceph official documentation: CRUSH maps. Includes the following parameters:
Example configuration: crush:
datacenter: dc1
room: room1
pdu: pdu1
row: row1
rack: rack1
chassis: ch1
region: region1
zone: zone1
|
|
Optional. Available since MOSK 25.1. Specifies the custom monitor endpoint for the node on which the monitor is placed. The custom monitor endpoint can be equal, for example, to an IP address from the Ceph public network range. Example configuration: monitorIP: "192.168.13.1"
|
Parameter |
Description |
|---|---|
|
Specifies a Ceph node specification. For the entire spec, see Node parameters. |
|
Specifies a list of names of machines to which the Ceph node nodeGroups:
group-1:
spec: <node spec>
nodes:
- master-0
- master-1
- worker-0
|
|
Specifies a string with a valid label selector expression to
select machines to which the node spec must be applied. Mutually
exclusive with nodeGroup:
group-2:
spec: <node spec>
label: "ceph-storage-node=true,!ceph-control-node"
|
Parameter |
Description |
|---|---|
|
Mandatory. Specifies the pool name as a prefix for each Ceph block pool. The
resulting Ceph block pool name will be |
|
Optional. Enables Ceph block pool to use only the |
|
Mandatory. Specifies the pool role and is used mostly for (MOSK) pools. |
|
Mandatory. Defines if the pool and dependent StorageClass should be set as default. Must be enabled only for one pool. |
|
Mandatory. Specifies the device class for the defined pool. Possible values are HDD, SSD, and NVMe. |
|
Mandatory, mutually exclusive with
|
|
Mandatory, mutually exclusive with |
|
Mandatory. The failure domain across which the replicas or chunks of data will
be spread. Set to Caution Mirantis does not recommend using the following
intermediate topology keys: |
|
Optional. Enables the mirroring feature for the defined pool.
Includes the |
|
Optional. Not updatable as it applies only once. Enables expansion of
persistent volumes based on Note A Kubernetes cluster only supports increase of storage size. |
|
Optional. Not updatable as it applies only once. Specifies custom
|
|
Optional. Available since MOSK 23.1. Specifies the key-value map for the parameters of the Ceph pool. For details, see Ceph documentation: Set Pool values. |
|
Optional. Available since MOSK 23.3. Specifies reclaim
policy for the underlying |
Example configuration:
pools:
- name: kubernetes
role: kubernetes
deviceClass: hdd
replicated:
size: 3
targetSizeRatio: 10.0
default: true
To configure additional required pools for MOSK, see Add a Ceph cluster.
Caution
Since Ceph Pacific, Ceph CSI driver does not propagate the
777 permission on the mount point of persistent volumes based on any
StorageClass of the Ceph pool.
Parameter |
Description |
|---|---|
|
Ceph client name. |
|
Key-value parameter with Ceph client capabilities. For details
about |
Example configuration:
clients:
- name: glance
caps:
mon: allow r, allow command "osd blacklist"
osd: profile rbd pool=images
Parameter |
Description |
|---|---|
|
Ceph Object Storage instance name. |
|
Mutually exclusive with the cephClusterSpec:
objectStorage:
rgw:
dataPool:
erasureCoded:
codingChunks: 1
dataChunks: 2
|
|
Mutually exclusive with the cephClusterSpec:
objectStorage:
rgw:
metadataPool:
replicated:
size: 3
failureDomain: host
where Warning When using the non-recommended Ceph pools For example, if |
|
The gateway settings corresponding to the
For example: cephClusterSpec:
objectStorage:
rgw:
gateway:
allNodes: false
instances: 1
port: 80
securePort: 8443
|
|
Defines whether to delete the data and metadata pools in the |
|
Optional. To create new Ceph RGW resources, such as buckets or users, specify the following keys. Ceph Controller will automatically create the specified object storage users and buckets in the Ceph cluster.
|
|
Optional. Mutually exclusive with For example: cephClusterSpec:
objectStorage:
multisite:
zones:
- name: master-zone
...
rgw:
zone:
name: master-zone
|
|
Optional. Custom TLS certificate parameters used to access the Ceph RGW endpoint. If not specified, a self-signed certificate will be generated. For example: cephClusterSpec:
objectStorage:
rgw:
SSLCert:
cacert: |
-----BEGIN CERTIFICATE-----
ca-certificate here
-----END CERTIFICATE-----
tlsCert: |
-----BEGIN CERTIFICATE-----
private TLS certificate here
-----END CERTIFICATE-----
tlsKey: |
-----BEGIN RSA PRIVATE KEY-----
private TLS key here
-----END RSA PRIVATE KEY-----
|
|
Optional. Available since {{ product_name_abbr }} 25.1. Flag to determine
that a TLS certificate for accessing the Ceph RGW endpoint is used but
not exposed in cephClusterSpec:
objectStorage:
rgw:
SSLCertInRef: true
The operator must manually provide TLS configuration using the
data:
cacert: <base64encodedCaCertificate>
cert: <base64encodedCertificate>
When removing an already existing When adding a new secret directly without exposing it in
|
For configuration example, see Enable Ceph RGW Object Storage.
Parameter |
Description |
|---|---|
|
Available since MOSK 23.3.
A key-value setting used to assign a specification label to any
available device on a specific node. These labels can then be
utilized within Usage: extraOpts:
deviceLabels:
<node-name>:
<dev-label>: /dev/disk/by-id/<unique_ID>
...
<node-name-n>:
<dev-label-n>: /dev/disk/by-id/<unique_ID>
nodesGroup:
<group-name>:
spec:
storageDevices:
- devLabel: <dev_label>
- devLabel: <dev_label_n>
nodes:
- <node_name>
- <node_name_n>
Before MOSK 23.3, you need to specify the device labels for each node separately: nodes:
<node-name>:
- storageDevices:
- fullPath: /dev/disk/by-id/<unique_ID>
<node-name-n>:
- storageDevices:
- fullPath: /dev/disk/by-id/<unique_ID>
|
|
Available since MOSK 23.3 as TechPreview.
A list of custom device class names to use in the specification.
Enables you to specify the custom names different from the default
ones, which include Usage: extraOpts:
customDeviceClasses:
- <custom_class_name>
nodes:
kaas-node-5bgk6:
storageDevices:
- config: # existing item
deviceClass: <custom_class_name>
fullPath: /dev/disk/by-id/<unique_ID>
pools:
- default: false
deviceClass: <custom_class_name>
erasureCoded:
codingChunks: 1
dataChunks: 2
failureDomain: host
Before MOSK 23.3, you cannot specify custom class names in the specification. |
Parameter |
Description |
|---|---|
|
List of realms to use, represents the realm namespaces. Includes the following parameters:
|
|
The list of zone groups for realms. Includes the following parameters:
|
|
The list of zones used within one zone group. Includes the following parameters:
|
For configuration example, see Enable multisite for Ceph RGW Object Storage.
Parameter |
Description |
|---|---|
|
Specifies health check settings for Ceph daemons. Contains the following parameters:
Each parameter allows defining the following settings:
|
|
Key-value parameter with liveness probe settings for the defined
daemon types. Can be one of the following:
Note Ceph Controller specifies the following
|
|
Key-value parameter with startup probe settings for the defined
daemon types. Can be one of the following:
|
Example configuration
healthCheck:
daemonHealth:
mon:
disabled: false
interval: 45s
timeout: 600s
osd:
disabled: false
interval: 60s
status:
disabled: true
livenessProbe:
mon:
disabled: false
probe:
timeoutSeconds: 10
periodSeconds: 3
successThreshold: 3
mgr:
disabled: false
probe:
timeoutSeconds: 5
failureThreshold: 5
osd:
probe:
initialDelaySeconds: 5
timeoutSeconds: 10
failureThreshold: 7
startupProbe:
mon:
disabled: true
mgr:
probe:
successThreshold: 3
The following sections describe the OpenStack-related Ceph operations:
Configure Ceph Object Gateway TLS¶
Once you enable Ceph Object Gateway (radosgw) as described in
Enable Ceph RGW Object Storage, you can configure the Transport Layer Security (TLS)
protocol for a Ceph Object Gateway public endpoint using the following options:
Using MOSK TLS, if it is enabled and exposes its certificates and domain for Ceph. In this case, Ceph Object Gateway will automatically create an ingress rule with MOSK certificates and domain to access the Ceph Object Gateway public endpoint. Therefore, you only need to reach the Ceph Object Gateway public and internal endpoints and set the CA certificates for a trusted TLS connection.
Using custom
ingressspecified in theKaaSCephClusterCR. In this case, Ceph Object Gateway public endpoint will use the public domain specified using theingressparameters.
Caution
External Ceph Object Gateway service is not supported and will be deleted during update. If your system already uses endpoints of an external Ceph Object Gateway service, reconfigure them to the ingress endpoints.
Caution
When using a custom or OpenStack ingress, ensure to configure the DNS name for RGW to target an external IP address of that ingress. If there is no OpenStack or custom ingress available, point the DNS to an external load balancer of RGW.
Note
Since MOSK 23.3, if the cluster has
tls-proxy enabled, TLS certificates specified in ingress objects,
including those configured in the KaaSCephCluster specification,
are disregarded. Instead, common certificates are applied to all ingresses
from the OpenStackDeployment object. This implies that tlsCert and
other ingress certificates specified in KaaSCephCluster are ignored,
and the common certificate from the OpenStackDeployment object is used.
This section also describes how to specify a custom public endpoint for the Object Storage service.
To configure Ceph Object Gateway TLS:
Verify whether MOSK TLS is enabled. The
spec.features.ssl.public_endpointssection should be specified in theOpenStackDeploymentCR.To generate an SSL certificate for internal usage, verify that the gateway
securePortparameter is specified in theKaasCephClusterCR. For details, see Enable Ceph RGW Object Storage.Select from the following options:
Configure TLS for Ceph Object Gateway using a custom
ingressConfig:Open the
KaasCephClusterCR for editing.Specify the
ingressConfigparameters:Description of the tlsConfig section parameters¶ certsTLS configuration for ingress including certificates. Contains the following parameters:
cacertThe Certificate Authority (CA) certificate, used for the ingress rule TLS support.
tlsCertThe TLS certificate, used for the ingress rule TLS support.
tlsKeyThe TLS private key, used for the ingress rule TLS support.
publicDomainMandatory. The domain name to use for public endpoints.
Caution
The default ingress controller does not support
publicDomainvalues different from the OpenStack ingress public domain. Therefore, if you intend to use the default OpenStack Ingress Controller for your Ceph Object Storage public endpoint, plan to use the same public domain as your OpenStack endpoints.hostnameCustom name to override the Objectstore RGW name for public RGW access. Public RGW endpoint has the
https://<hostname>.<publicDomain>format.tlsSecretRefNameOptional. Secret name with TLS certs on the managed cluster in the
rook-cephnamespace prepared by the operator. Allows avoiding exposure of certs directly inspec. Must contain the following format:data: ca.cert: <base64encodedCaCertificate> tls.crt: <base64encodedTlsCert> tls.key: <base64encodedTlsKey>
Caution
When using
tlsSecretRefName, remove the following fields:cacert,tlsCert, andtlsKey.Description of optional parameters in the ingressConfig section¶ controllerClassNameName of the custom Ingress Controller. By default, the
openstack-ingress-nginxclass name is specified and Ceph uses the OpenStack Ingress Controller based on NGINX.annotationsExtra annotations for the ingress proxy that are a key-value mapping of strings to add or override ingress rule annotations. For details, see NGINX Ingress Controller: Annotations.
By default, the following annotations are set:
nginx.ingress.kubernetes.io/rewrite-targetis set to/nginx.ingress.kubernetes.io/upstream-vhostis set to<rgwName>.rook-ceph.svc
The value for
<rgwName>is located inspec.cephClusterSpec.objectStorage.rgw.name.Optional annotations:
nginx.ingress.kubernetes.io/proxy-request-buffering: "off"that disables buffering foringressto prevent the 413 (Request Entity Too Large) error when uploading large files usingradosgw.nginx.ingress.kubernetes.io/proxy-body-size: <size>that increases the default uploading size limit to prevent the 413 (Request Entity Too Large) error when uploading large files usingradosgw. Set the value in MB (m) or KB (k). For example,100m.
Note
By default, an ingress rule is created with an internal Ceph Object Gateway service endpoint as a backend. Also,
rgw dns nameis specified in the Ceph configuration and is set to<rgwName>.rook-ceph.svcby default.You can override
rgw dns nameusing thespec.cephClusterSpec.rookConfigkey-value parameter. In this case, also change the corresponding ingress annotation.Configuration example with the
rgw dns nameoverridespec: cephClusterSpec: objectStorage: rgw: name: rgw-store ingressConfig: tlsConfig: publicDomain: public.domain.name certs: cacert: | -----BEGIN CERTIFICATE----- ... -----END CERTIFICATE----- tlsCert: | -----BEGIN CERTIFICATE----- ... -----END CERTIFICATE----- tlsKey: | -----BEGIN RSA PRIVATE KEY----- ... -----END RSA PRIVATE KEY----- controllerClassName: openstack-ingress-nginx annotations: nginx.ingress.kubernetes.io/rewrite-target: / nginx.ingress.kubernetes.io/upstream-vhost: rgw-store.public.domain.name nginx.ingress.kubernetes.io/proxy-body-size: 100m rookConfig: "rgw dns name": rgw-store.public.domain.name
For clouds with the
publicDomainparameter specified, align theupstream-vhostingress annotation with the name of the Ceph Object Storage and the specified public domain.Ceph Object Storage requires the
upstream-vhostandrgw dns nameparameters to be equal. Therefore, override the defaultrgw dns namewith the corresponding ingress annotation value.
Configure Ceph Object Gateway TLS using a custom
ingress:Warning
The
rgwsection is deprecated and theingressparameters are moved undercephClusterSpec.ingress. If you continue usingrgw.ingress, it will be automatically translated intocephClusterSpec.ingressduring the MOSK cluster release update.Open the
KaasCephClusterCR for editing.Specify the
ingressparameters:publicDomain- domain name to use for the external service.Caution
Since MOSK 23.3, the default ingress controller does not support
publicDomainvalues different from the OpenStack ingress public domain. Therefore, if you intend to use the default OpenStack ingress controller for your Ceph Object Storage public endpoint, plan to use the same public domain as your OpenStack endpoints.cacert- Certificate Authority (CA) certificate, used for the ingress rule TLS support.tlsCert- TLS certificate, used for the ingress rule TLS support.tlsKey- TLS private key, used for the ingress rule TLS support.customIngressOptional - includes the following custom Ingress Controller parameters:className- the custom Ingress Controller class name. If not specified, theopenstack-ingress-nginxclass name is used by default.annotations- extra annotations for the ingress proxy. For details, see NGINX Ingress Controller: Annotations.By default, the following annotations are set:
nginx.ingress.kubernetes.io/rewrite-targetis set to/nginx.ingress.kubernetes.io/upstream-vhostis set to<rgwName>.rook-ceph.svc.The value for
<rgwName>isspec.cephClusterSpec.objectStorage.rgw.name.
Optional annotations:
nginx.ingress.kubernetes.io/proxy-request-buffering: "off"that disables buffering foringressto prevent the 413 (Request Entity Too Large) error when uploading large files usingradosgw.nginx.ingress.kubernetes.io/proxy-body-size: <size>that increases the default uploading size limit to prevent the 413 (Request Entity Too Large) error when uploading large files usingradosgw. Set the value in MB (m) or KB (k). For example,100m.
For example:
customIngress: className: openstack-ingress-nginx annotations: nginx.ingress.kubernetes.io/rewrite-target: / nginx.ingress.kubernetes.io/upstream-vhost: openstack-store.rook-ceph.svc nginx.ingress.kubernetes.io/proxy-body-size: 100m
Note
An ingress rule is by default created with an internal Ceph Object Gateway service endpoint as a backend. Also,
rgw dns nameis specified in the Ceph configuration and is set to<rgwName>.rook-ceph.svcby default. You can override this option using thespec.cephClusterSpec.rookConfigkey-value parameter. In this case, also change the corresponding ingress annotation.For example:
spec: cephClusterSpec: objectStorage: rgw: name: rgw-store ingress: publicDomain: public.domain.name cacert: | -----BEGIN CERTIFICATE----- ... -----END CERTIFICATE----- tlsCert: | -----BEGIN CERTIFICATE----- ... -----END CERTIFICATE----- tlsKey: | -----BEGIN RSA PRIVATE KEY----- ... -----END RSA PRIVATE KEY----- customIngress: annotations: "nginx.ingress.kubernetes.io/upstream-vhost": rgw-store.public.domain.name rookConfig: "rgw dns name": rgw-store.public.domain.name
Warning
For clouds with the
publicDomainparameter specified, align theupstream-vhostingress annotation with the name of the Ceph Object Storage and the specified public domain.Ceph Object Storage requires the
upstream-vhostandrgw dns nameparameters to be equal. Therefore, override the defaultrgw dns nameto the corresponding ingress annotation value.
Obtain the MOSK CA certificate for a trusted connection:
kubectl -n openstack-ceph-shared get secret openstack-rgw-creds -o jsonpath="{.data.ca_cert}" | base64 -d
Access internal and public Ceph Object Gateway endpoints by selecting one of the following options:
Note
If you are using the HTTP scheme instead of HTTPS:
In the
KaaSCephClusterobject on the management cluster, find theingressConfigsection and add custom annotations:spec: cephClusterSpec: ingressConfig: annotations: "nginx.ingress.kubernetes.io/force-ssl-redirect": "false" "nginx.ingress.kubernetes.io/ssl-redirect": "false"
You can omit TLS configuration for the default settings provided by OpenStack to be applied.
If both HTTP and HTTPS must be used, apply the following configuration in the
KaaSCephClusterobject:spec: cephClusterSpec: ingressConfig: tlsConfig: publicDomain: public.domain.name cacert: | -----BEGIN CERTIFICATE----- ... -----END CERTIFICATE----- tlsCert: | -----BEGIN CERTIFICATE----- ... -----END CERTIFICATE----- tlsKey: | -----BEGIN RSA PRIVATE KEY----- ... -----END RSA PRIVATE KEY----- annotations: "nginx.ingress.kubernetes.io/force-ssl-redirect": "false" "nginx.ingress.kubernetes.io/ssl-redirect": "false"
In the
KaaSCephClusterobject on the management cluster, find theingresssection and add custom annotations:spec: cephClusterSpec: ingress: publicDomain: public.domain.name cacert: | -----BEGIN CERTIFICATE----- ... -----END CERTIFICATE----- tlsCert: | -----BEGIN CERTIFICATE----- ... -----END CERTIFICATE----- tlsKey: | -----BEGIN RSA PRIVATE KEY----- ... -----END RSA PRIVATE KEY----- customIngress: annotations: "nginx.ingress.kubernetes.io/force-ssl-redirect": "false" "nginx.ingress.kubernetes.io/ssl-redirect": "false"
In the
KaaSCephClusterobject on the management cluster, find theingresssection and add custom annotations:spec: cephClusterSpec: rookConfig: rgw_remote_addr_param: "HTTP_X_FORWARDED_FOR"
Obtain the Ceph Object Gateway public endpoint:
kubectl -n rook-ceph get ingress
Obtain the public endpoint TLS CA certificate:
kubectl -n rook-ceph get secret $(kubectl -n rook-ceph get ingress -o jsonpath='{.items[0].spec.tls[0].secretName}{"\n"}') -o jsonpath='{.data.ca\.crt}' | base64 -d; echo
Obtain the internal endpoint name for Ceph Object Gateway:
kubectl -n rook-ceph get svc -l app=rook-ceph-rgw
The internal endpoint for Ceph Object Gateway has the
https://<internal-svc-name>.rook-ceph.svc:<rgw-secure-port>/format, where<rgw-secure-port>isspec.rgw.gateway.securePortspecified in theKaaSCephClusterCR.Obtain the internal endpoint TLS CA certificate:
kubectl -n rook-ceph get secret rgw-ssl-certificate -o jsonpath="{.data.cacert}" | base64 -d
Update the zonegroup
hostnamesof Ceph Object Gateway:The
hostnameszonegroup of Ceph Object Gateway updates automatically and you can skip this step if one of the following requirements is met:The public hostname matches the public domain name set by the
spec.cephClusterSpec.ingressConfig.tlsConfig.publicDomainfieldThe OpenStack configuration has been applied
Otherwise, complete the steps for MOSK 22.4 and earlier product versions.
Note
Prior to MOSK 25.1, use the
spec.cephClusterSpec.ingress.publicDomainfield instead.Enter the
rook-ceph-toolspod:kubectl -n rook-ceph exec -it deployment/rook-ceph-tools -- bash
Obtain Ceph Object Gateway default zonegroup configuration:
radosgw-admin zonegroup get --rgw-zonegroup=<objectStorageName> --rgw-zone=<objectStorageName> | tee zonegroup.json
Substitute
<objectStorageName>with the Ceph Object Storage name fromspec.cephClusterSpec.objectStorage.rgw.name.Inspect
zonegroup.jsonand verify that thehostnameskey is a list that contains two endpoints: an internal endpoint and a custom public endpoint:"hostnames": ["rook-ceph-rgw-<objectStorageName>.rook-ceph.svc", <customPublicEndpoint>]
Substitute
<objectStorageName>with the Ceph Object Storage name and<customPublicEndpoint>with the public endpoint with a custom public domain.If one or both endpoints are omitted in the list, add the missing endpoints to the
hostnameslist in thezonegroup.jsonfile and update Ceph Object Gateway zonegroup configuration:radosgw-admin zonegroup set --rgw-zonegroup=<objectStorageName> --rgw-zone=<objectStorageName> --infile zonegroup.json radosgw-admin period update --commit
Verify that the
hostnameslist contains both the internal and custom public endpoint:radosgw-admin --rgw-zonegroup=<objectStorageName> --rgw-zone=<objectStorageName> zonegroup get | jq -r ".hostnames"
Example of system response:
[ "rook-ceph-rgw-obj-store.rook-ceph.svc", "obj-store.mcc1.cluster1.example.com" ]
Exit the
rook-ceph-toolspod:exit
Once done, Ceph Object Gateway becomes available by the custom public endpoint with an S3 API client, OpenStack Swift CLI, and OpenStack Horizon Containers plugin.
Use object storage server-side encryption¶
TechPreview
When you use Ceph Object Gateway server-side encryption (SSE), unencrypted data sent over HTTPS is stored encrypted by the Ceph Object Gateway in the Ceph cluster. The current implementation integrates Barbican as a key management service.
The object storage SSE feature is enabled by default in MOSK deployments with Barbican. To use object storage SSE, the AWS CLI S3 client is used.
To use object storage server-side encryption:
Create Amazon Elastic Compute Cloud (EC2) credentials:
openstack ec2 credentials create
Configure AWS CLI with
accessandsecretcreated in the previous step:aws configureCreate a secret key in Barbican secret key:
openstack secret order create --name <name> --algorithm <algorithm> --mode <mode> --bit-length 256 --payload-content-type=<payload-content-type> key
Substitute the parameters enclosed in angle brackets:
<name>- human-friendly name.<algorithm>- algorithm to use with the requested key. For example,aes.<mode>- algorithm mode to use with the requested key. For example,ctr.<payload-content-type>- type/format of the secret to generate. For example,application/octet-stream.
Verify that the key has been created:
openstack secret order get <order-href>
Substitute
<order-href>with the corresponding value from the output of the previous command.Specify the
ceph-rgwuser in the Barbican secret Access Control List (ACL):Obtain the list of
ceph-rgwusers:openstack user list --domain service | grep ceph-rgw
Example output:
| c63b70134e0845a2b13c3f947880f66a | ceph-rgwZ6ycK3dY |
In the output, capture the first value as the
<user-id>, which isc63b70134e0845a2b13c3f947880f66ain the above example.Specify the
ceph-rgwuser in the Barbican ACL:openstack acl user add --user <user-id> <secret-href>
Substitute
<user-id>with the corresponding value from the output of the previous command and<secret-href>with the corresponding value obtained in step 3.
Create an S3 bucket:
aws --endpoint-url <rgw-endpoint-url> --ca-bundle <ca-bundle> s3api create-bucket --bucket <bucket-name>
Substitute the parameters enclosed in angle brackets:
<rgw-endpoint-url>- Ceph Object Gateway endpoint DNS name<ca-bundle>- CA Certificate Bundle<bucket-name>- human-friendly bucket name
Upload a file using object storage SSE:
aws --endpoint-url <rgw-endpoint-url> --ca-bundle <ca-bundle> s3 cp <path-to-file> "s3://<bucket-name>/<filename>" --sse aws:kms --sse-kms-key-id <key-id>
Substitute the parameters enclosed in angle brackets:
<path-to-file>- path to the file that you want to upload<filename>- name under which the uploaded file will be stored in the bucket<key-id>- Barbican secret key ID
Select from the following options to download the file:
Download the file using a key:
aws --endpoint-url <rgw-endpoint-url> --ca-bundle <ca-bundle> s3 cp "s3://<bucket-name>/<filename>" <path-to-output-file> --sse aws:kms --sse-kms-key-id <key-id>
Substitute
<path-to-output-file>with the path to the file you want to download.Download the file without a key:
aws --endpoint-url <rgw-endpoint-url> --ca-bundle <ca-bundle> s3 cp "s3://<bucket-name>/<filename>" <output-filename>
Set an Amazon S3 bucket policy¶
This section explains how to create an Amazon Simple Storage Service (Amazon S3 or S3) bucket and set an S3 bucket policy between two Ceph Object Storage users.
Create Ceph Object Storage users¶
Ceph Object Storage users can create Amazon S3 buckets and bucket policies that grant access to other users.
This section describes how to create two Ceph Object Storage users and configure their S3 credentials.
To create and configure Ceph Object Storage users:
Open the
KaaSCephClusterCR:kubectl --kubeconfig <managementKubeconfig> -n <managedClusterProject> edit kaascephcluster
Substitute
<managementKubeconfig>with a management clusterkubeconfigfile and<managedClusterProject>with a managed cluster project name.In the
cephClusterSpecsection, add new Ceph Object Storage users.Caution
For user
name, apply the UUID format with no capital letters.For example:
spec: cephClusterSpec: objectStorage: rgw: objectUsers: - name: user-b displayName: user-a capabilities: bucket: "*" user: read - name: user-t displayName: user-t capabilities: bucket: "*" user: read
Verify that
rgwUserSecretsare created for both users:kubectl --kubeconfig <managementKubeconfig> -n <managedClusterProject> get kaascephcluster -o yaml
Substitute
<managementKubeconfig>with a management clusterkubeconfigfile and<managedClusterProject>with a managed cluster project name.Example of a positive system response:
status: miraCephSecretsInfo: secretInfo: rgwUserSecrets: - name: user-a secretName: <user-aCredSecretName> secretNamespace: <user-aCredSecretNamespace> - name: user-t secretName: <user-tCredSecretName> secretNamespace: <user-tCredSecretNamespace>
Obtain S3 user credentials from the cluster secrets. Specify an access key and a secret key for both users:
kubectl --kubeconfig <managedKubeconfig> -n <user-aCredSecretNamespace> get secret <user-aCredSecretName> -o jsonpath='{.data.AccessKey}' | base64 -d kubectl --kubeconfig <managedKubeconfig> -n <user-aCredSecretNamespace> get secret <user-aCredSecretName> -o jsonpath='{.data.SecretKey}' | base64 -d kubectl --kubeconfig <managedKubeconfig> -n <user-tCredSecretNamespace> get secret <user-tCredSecretName> -o jsonpath='{.data.AccessKey}' | base64 -d kubectl --kubeconfig <managedKubeconfig> -n <user-tCredSecretNamespace> get secret <user-tCredSecretName> -o jsonpath='{.data.SecretKey}' | base64 -d
Substitute
<managementKubeconfig>with a management clusterkubeconfigand specify the correspondingsecretNamespaceandsecretNamefor both users.Obtain Ceph Object Storage public endpoint from the
KaaSCephClusterstatus:kubectl --kubeconfig <managementKubeconfig> -n <managedClusterProject> get kaascephcluster -o yaml | grep PublicEndpoint
Substitute
<managementKubeconfig>with a management clusterkubeconfigfile and<managedClusterProject>with a managed cluster project name.Example of a positive system response:
objectStorePublicEndpoint: https://object-storage.mirantis.example.comObtain the CA certificate to use an HTTPS endpoint:
kubectl --kubeconfig <managedKubeconfig> -n rook-ceph get secret $(kubectl -n rook-ceph get ingress -o jsonpath='{.items[0].spec.tls[0].secretName}{"\n"}') -o jsonpath='{.data.ca\.crt}' | base64 -d; echo
Save the output to
ca.crt.
Set a bucket policy for a Ceph Object Storage user¶
Available since 2.23.1 (Cluster release 12.7.0)
Amazon S3 is an object storage service with different access policies. A bucket policy is a resource-based policy that grants permissions to a bucket and objects in it. For more details, see Amazon S3 documentation: Using bucket policies .
The following procedure illustrates the process of setting a bucket policy for
a bucket (test01) stored in a Ceph Object Storage. The bucket policy
requires at least two users: a bucket owner (user-a) and a bucket user
(user-t). The bucket owner creates the bucket and sets the policy that
regulates access for the bucket user.
Caution
For user name, apply the UUID format with no capital letters.
To configure an Amazon S3 bucket policy:
Note
The s3cmd is a free command-line tool and client for uploading, retrieving, and managing data in Amazon S3 and other cloud storage service providers that use the S3 protocol. You can download the s3cmd CLI tool from Amazon S3 tools: Download s3cmd.
Configure the s3cmd client with the
user-acredentials:s3cmd --configure --ca-certs=ca.crt
Specify the bucket access parameters as required:
Bucket access parameters¶ Parameter
Description
Comment
Access KeyPublic part of access credentials.
Specify a user access key.
Secret KeySecret part of access credentials.
Specify a user secret key.
Default RegionRegion of AWS servers where requests are sent by default.
Use the default value.
S3 EndpointConnection point to the Ceph Object Storage.
Specify the Ceph Object Storage public endpoint.
DNS-style bucket+hostname:port template for accessing a bucketBucket location.
Specify the Ceph Object Storage public endpoint.
Path to GPG programPath to the GNU Privacy Guard encryption suite.
Use the default value.
Use HTTPS protocolHTTPS protocol switch.
Specify
Yes.HTTP Proxy server nameHTTP Proxy server name.
Skip this parameter.
When configured correctly, the s3cmd tool connects to the Ceph Object Storage. Save new settings when prompted by the system.
As
user-a, create a new buckettest01:s3cmd mb s3://test01
Example of a positive system response:
Bucket 's3://test01/' created
Upload an object to the bucket:
touch test.txt s3cmd put test.txt s3://test01
Example of a positive system response:
upload: 'test.txt' -> 's3://test01/test.txt' [1 of 1] 0 of 0 0% in 0s 0.00 B/s done
Verify that the object is in the
test01bucket:s3cmd ls s3://test01
Example of a positive system response:
2022-09-02 13:06 0 s3://test01/test.txt
Create the bucket policy file and add bucket CRUD permissions for
user-t:{ "Version": "2012-10-17", "Id": "S3Policy1", "Statement": [ { "Sid": "BucketAllow", "Effect": "Allow", "Principal": { "AWS": ["arn:aws:iam:::user/user-t"] }, "Action": [ "s3:ListBucket", "s3:PutObject", "s3:GetObject" ], "Resource": [ "arn:aws:s3:::test01", "arn:aws:s3:::test01/*" ] } ] }
Set the bucket policy for the
test01bucket:s3cmd setpolicy policy.json s3://test01
Example of a positive system response:
s3://test01/: Policy updated
Verify that the
user-thas access for thetest01bucket by reconfiguring the s3cmd client with theuser-tcredentials:s3cmd --ca-certs=ca.crt --configure
Specify the bucket access parameters in a similar to the step 1 manner.
When configured correctly, the s3cmd tool connects to the Ceph Object Storage. Save new settings when prompted by the system.
Verify that the
user-tcan read the buckettest01content:s3cmd ls s3://test01
Example of a positive system response:
2022-09-02 13:06 0 s3://test01/test.txt
Download the object from the
test01bucket:s3cmd get s3://test01/test.txt check.txt
Example of a positive system response:
download: 's3://test01/test.txt' -> 'check.txt' [1 of 1] 0 of 0 0% in 0s 0.00 B/s done
Upload a new object to the
test01bucket:s3cmd put check.txt s3://test01
Example of a positive system response:
upload: 'check.txt' -> 's3://test01/check.txt' [1 of 1] 0 of 0 0% in 0s 0.00 B/s done
Verify that the object is in the
test01bucket:s3cmd ls s3://test01
Example of a positive system response:
2022-09-02 14:41 0 s3://test01/check.txt 2022-09-02 13:06 0 s3://test01/test.txt
Verify the new object by reconfiguring the s3cmd client with the
user-acredentials:s3cmd --configure --ca-certs=ca.crt
List
test01bucket objects:s3cmd ls s3://test01
Example of a positive system response:
2022-09-02 14:41 0 s3://test01/check.txt 2022-09-02 13:06 0 s3://test01/test.txt
Set a bucket policy for OpenStack users¶
The following procedure illustrates the process of setting a bucket policy for a bucket between two OpenStack users.
Due to specifics of the Ceph integration with OpenStack projects, you should configure the bucket policy for OpenStack users indirectly through setting permissions for corresponding OpenStack projects.
For illustration purposes, we use the following names in the procedure:
test01for the bucketuser-a,user-tfor the OpenStack usersproject-a,project-tfor the OpenStack projects
To configure an Amazon S3 bucket policy for OpenStack users:
Specify the
rookConfigparameter in thecephClusterSpecsection of theKaaSCephClustercustom resource:spec: cephClusterSpec: rookConfig: rgw keystone implicit tenants: "swift"
Prepare the Ceph Object Storage similarly to the procedure described in Create Ceph Object Storage users.
Create two OpenStack projects:
openstack project create project-a openstack project create project-t
Example of system response:
+-------------+----------------------------------+ | Field | Value | +-------------+----------------------------------+ | description | | | domain_id | default | | enabled | True | | id | faf957b776874a2e80384cb882ebf6ab | | is_domain | False | | name | project-a | | options | {} | | parent_id | default | | tags | [] | +-------------+----------------------------------+
You can also use existing projects. Save the ID of each project for the bucket policy specification.
Note
For details how to access OpenStack CLI, refer Access your OpenStack environment.
Create an OpenStack user for each project:
openstack user create user-a --project project-a openstack user create user-t --project project-t
Example of system response:
+---------------------+----------------------------------+ | Field | Value | +---------------------+----------------------------------+ | default_project_id | faf957b776874a2e80384cb882ebf6ab | | domain_id | default | | enabled | True | | id | cc2607dc383e4494948d68eeb556f03b | | name | user-a | | options | {} | | password_expires_at | None | +---------------------+----------------------------------+
You can also use existing project users.
Assign the
memberrole to the OpenStack users:openstack role add member --user user-a --project project-a openstack role add member --user user-t --project project-t
Verify that the OpenStack users have obtained the
memberroles paying attention to the role IDs:openstack role show member
Example of system response:
+-------------+----------------------------------+ | Field | Value | +-------------+----------------------------------+ | description | None | | domain_id | None | | id | 8f0ce4f6cd61499c809d6169b2b5bd93 | | name | member | | options | {'immutable': True} | +-------------+----------------------------------+
List the role assignments for the
user-aanduser-t:openstack role assignment list --user user-a --project project-a openstack role assignment list --user user-t --project project-t
Example of system response:
+----------------------------------+----------------------------------+-------+----------------------------------+--------+--------+-----------+ | Role | User | Group | Project | Domain | System | Inherited | +----------------------------------+----------------------------------+-------+----------------------------------+--------+--------+-----------+ | 8f0ce4f6cd61499c809d6169b2b5bd93 | cc2607dc383e4494948d68eeb556f03b | | faf957b776874a2e80384cb882ebf6ab | | | False | +----------------------------------+----------------------------------+-------+----------------------------------+--------+--------+-----------+
Create Amazon EC2 credentials for
user-aanduser-t:openstack ec2 credentials create --user user-a --project project-a openstack ec2 credentials create --user user-t --project project-t
Example of system response:
+------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------+ | Field | Value | +------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------+ | access | d03971aedc2442dd9a79b3b409c32046 | | links | {'self': 'http://keystone-api.openstack.svc.cluster.local:5000/v3/users/cc2607dc383e4494948d68eeb556f03b/credentials/OS-EC2/d03971aedc2442dd9a79b3b409c32046'} | | project_id | faf957b776874a2e80384cb882ebf6ab | | secret | 0a9fd8d9e0d24aecacd6e75951154d0f | | trust_id | None | | user_id | cc2607dc383e4494948d68eeb556f03b | +------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------+
Obtain the values from the
accessandsecretfields to connect with Ceph Object Storage trough the s3cmd tool.Note
The s3cmd is a free command-line tool for uploading, retrieving, and managing data in Amazon S3 and other cloud storage service providers that use the S3 protocol. You can download the s3cmd tool from Amazon S3 tools: Download s3cmd.
Create bucket users and configure a bucket policy for the
project-tOpenStack project similarly to the procedure described in Set a bucket policy for a Ceph Object Storage user. Ceph integration does not allow providing permissions for OpenStack users directly. Therefore, you need to set permissions for the project that corresponds to the user:{ "Version": "2012-10-17", "Id": "S3Policy1", "Statement": [ { "Sid": "BucketAllow", "Effect": "Allow", "Principal": { "AWS": ["arn:aws:iam::<PROJECT-T_ID>:root"] }, "Action": [ "s3:ListBucket", "s3:PutObject", "s3:GetObject" ], "Resource": [ "arn:aws:s3:::test01", "arn:aws:s3:::test01/*" ] } ] }
You can configure different bucket policies for various situations. See examples below.
Provide access to a bucket from one OpenStack project to another
{
"Version": "2012-10-17",
"Id": "S3Policy1",
"Statement": [
{
"Sid": "BucketAllow",
"Effect": "Allow",
"Principal": {
"AWS": ["arn:aws:iam::<osProjectId>:root"]
},
"Action": [
"s3:ListBucket",
"s3:PutObject",
"s3:GetObject"
],
"Resource": [
"arn:aws:s3:::<bucketName>",
"arn:aws:s3:::<bucketName>/*"
]
}
]
}
Substitute the following parameters:
<osProjectId>- the target OpenStack project ID<bucketName>- the target bucket name where the policy will be set
Provide access to a bucket from a Ceph Object Storage user to an OpenStack project
{
"Version": "2012-10-17",
"Id": "S3Policy1",
"Statement": [
{
"Sid": "BucketAllow",
"Effect": "Allow",
"Principal": {
"AWS": ["arn:aws:iam::<osProjectId>:root"]
},
"Action": [
"s3:ListBucket",
"s3:PutObject",
"s3:GetObject"
],
"Resource": [
"arn:aws:s3:::<bucketName>",
"arn:aws:s3:::<bucketName>/*"
]
}
]
}
Substitute the following parameters:
<osProjectId>- the target OpenStack project ID<bucketName>- the target bucket name where policy will be set
Provide access to a bucket from an OpenStack user to a Ceph Object Storage user
{
"Version": "2012-10-17",
"Id": "S3Policy1",
"Statement": [
{
"Sid": "BucketAllow",
"Effect": "Allow",
"Principal": {
"AWS": ["arn:aws:iam:::user/<userName>"]
},
"Action": [
"s3:ListBucket",
"s3:PutObject",
"s3:GetObject"
],
"Resource": [
"arn:aws:s3:::<bucketName>",
"arn:aws:s3:::<bucketName>/*"
]
}
]
}
Substitute the following parameters:
<userName>- the target Ceph Object Storage user name<bucketName>- the target bucket name where policy will be set
Provide access to a bucket from one Ceph Object Storage user to another
{
"Version": "2012-10-17",
"Id": "S3Policy1",
"Statement": [
{
"Sid": "BucketAllow",
"Effect": "Allow",
"Principal": {
"AWS": ["arn:aws:iam:::user/<userName>"]
},
"Action": [
"s3:ListBucket",
"s3:PutObject",
"s3:GetObject"
],
"Resource": [
"arn:aws:s3:::<bucketName>",
"arn:aws:s3:::<bucketName>/*"
]
}
]
}
Substitute the following parameters:
<userName>- the target Ceph Object Storage user name<bucketName>- the target bucket name where policy will be set
Calculate target ratio for Ceph pools¶
Ceph pool target ratio defines for the Placement Group (PG) autoscaler the amount of data the pools are expected to acquire over time in relation to each other. You can set initial PG values for each Ceph pool. Otherwise, the autoscaler starts with the minimum value and scales up, causing a lot of data to move in the background.
You can allocate several pools to use the same device class, which is a solid
block of available capacity in Ceph. For example, if three pools
(kubernetes-hdd, images-hdd, and volumes-hdd) are set to use the
same device class hdd, you can set the target ratio for Ceph pools to
provide 80% of capacity to the volumes-hdd pool and distribute the
remaining capacity evenly between the two other pools. This way, Ceph pool
target ratio instructs Ceph on when to warn that a pool is running out of free
space and, at the same time, instructs Ceph on how many placement groups Ceph
should allocate/autoscale for a pool for better data distribution.
Ceph pool target ratio is not a constant value and you can change it according to new capacity plans. Once you specify target ratio, if the PG number of a pool scales, other pools with specified target ratio will automatically scale accordingly.
For details, see Ceph Documentation: Autoscaling Placement Groups.
To calculate target ratio for each Ceph pool:
Define raw capacity of the entire storage by device class:
kubectl -n rook-ceph exec -it $(kubectl -n rook-ceph get pod -l "app=rook-ceph-tools" -o name) -- ceph df
For illustration purposes, the procedure below uses raw capacity of 185 TB or 189440 GB.
Design Ceph pools with the considered device class upper bounds of the possible capacity. For example, consider the
hdddevice class that contains the following pools:The
kubernetes-hddpool will contain not more than 2048 GB.The
images-hddpool will contain not more than 2048 GB.The
volumes-hddpool will contain 50 GB per VM. The upper bound of used VMs on the cloud is 204, the pool replicated size is3. Therefore, calculate the upper bounds forvolumes-hdd:50 GB per VM * 204 VMs * 3 replicas = 30600 GB
The
backup-hddpool can be calculated as a relative ofvolumes-hdd. For example, 1volumes-hddstorage unit per 5backup-hddunits.The
vms-hddis a pool for ephemeral storage Copy on Write (CoW). We recommend designing the amount of ephemeral data it should store. For example purposes, we use 500 GB. However, in reality, despite the CoW data reduction, this value is very optimistic.
Note
If
dataPoolis replicated and Ceph Object Store is planned for intensive use, also calculate upper bounds fordataPool.Calculate target ratio for each considered pool. For example:
Example bounds and capacity¶ Pools upper bounds
Pools capacity
kubernetes-hdd= 2048 GBimages-hdd= 2048 GBvolumes-hdd= 30600 GBbackup-hdd= 30600 GB * 5 = 153000 GBvms-hdd= 500 GB
Summary capacity = 188196 GB
Total raw capacity = 189440 GB
Calculate pools fit factor using the (total raw capacity) / (pools summary capacity) formula. For example:
pools fit factor = 189440 / 188196 = 1.0066
Calculate pools upper bounds size using the (pool upper bounds) * (pools fit factor) formula. For example:
kubernetes-hdd = 2048 GB * 1.0066 = 2061.5168 GB images-hdd = 2048 GB * 1.0066 = 2061.5168 GB volumes-hdd = 30600 GB * 1.0066 = 30801.96 GB backup-hdd = 153000 GB * 1.0066 = 154009.8 GB vms-hdd = 500 GB * 1.0066 = 503.3 GB
Calculate pool target ratio using the (pool upper bounds) * 100 / (total raw capacity) formula. For example:
kubernetes-hdd = 2061.5168 GB * 100 / 189440 GB = 1.088 images-hdd = 2061.5168 GB * 100 / 189440 GB = 1.088 volumes-hdd = 30801.96 GB * 100 / 189440 GB = 16.259 backup-hdd = 154009.8 GB * 100 / 189440 GB = 81.297 vms-hdd = 503.3 GB * 100 / 189440 GB = 0.266
If required, calculate the target ratio for erasure-coded pools.
Due to erasure-coded pools splitting each object into
Kdata parts andMcoding parts, the total used storage for each object is less than that in replicated pools. Indeed,Mis equal to the number of OSDs that can be missing from the cluster without the cluster experiencing data loss. This means that planned data is stored with an efficiency of(K+M)/2on the Ceph cluster. For example, if an erasure-coded data pool withK=2, M=2planned capacity is 200 GB, then the total used capacity is200*(2+2)/2, which is 400 GB.Open the
KaasCephClusterCR of a managed cluster for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.In the
spec.cephClusterSpec.poolssection, specify the calculated relatives astargetSizeRatiofor each considered replicated pool. For example:spec: cephClusterSpec: pools: - name: kubernetes deviceClass: hdd ... replicated: size: 3 targetSizeRatio: 1.088 - name: images deviceClass: hdd ... replicated: size: 3 targetSizeRatio: 1.088 - name: volumes deviceClass: hdd ... replicated: size: 3 targetSizeRatio: 16.259 - name: backup deviceClass: hdd ... replicated: size: 3 targetSizeRatio: 81.297 - name: vms deviceClass: hdd ... replicated: size: 3 targetSizeRatio: 0.266
If Ceph Object Store
dataPoolisreplicatedand a proper value is calculated, also specify it:spec: cephClusterSpec: objectStorage: rgw: name: rgw-store ... dataPool: deviceClass: hdd ... replicated: size: 3 targetSizeRatio: <relative>
In the
spec.cephClusterSpec.poolssection, specify the calculated relatives asparameters.target_size_ratiofor each considered erasure-coded pool. For example:Note
The
parameterssection is a key-value mapping where the value is of the string type and should be quoted.spec: cephClusterSpec: pools: - name: ec-pool deviceClass: hdd ... parameters: target_size_ratio: "<relative>"
If Ceph Object Store
dataPooliserasure-codedand a proper value is calculated, also specify it:spec: cephClusterSpec: objectStorage: rgw: name: rgw-store ... dataPool: deviceClass: hdd ... parameters: target_size_ratio: "<relative>"
Verify that all target ratio has been successfully applied to the Ceph cluster:
kubectl -n rook-ceph exec -it $(kubectl -n rook-ceph get pod -l "app=rook-ceph-tools" -o name) -- ceph osd pool autoscale-status
Example of system response:
POOL SIZE TARGET SIZE RATE RAW CAPACITY RATIO TARGET RATIO EFFECTIVE RATIO BIAS PG_NUM NEW PG_NUM AUTOSCALE device_health_metrics 0 2.0 149.9G 0.0000 1.0 1 on kubernetes-hdd 2068 2.0 149.9G 0.0000 1.088 1.0885 1.0 32 on volumes-hdd 19 2.0 149.9G 0.0000 16.259 16.2591 1.0 256 on vms-hdd 19 2.0 149.9G 0.0000 0.266 0.2661 1.0 128 on backup-hdd 19 2.0 149.9G 0.0000 81.297 81.2972 1.0 256 on images-hdd 888.8M 2.0 149.9G 0.0116 1.088 1.0881 1.0 32 on
Optional. Repeat the steps above for other device classes.
Ceph pools for Cinder multi-backend¶
Available since MOSK 23.2
The KaaSCephCluster object supports multiple Ceph pools with the
volumes role to configure Cinder multiple backends.
To define Ceph pools for Cinder multiple backends:
In the
KaaSCephClusterobject, add the desired number of Ceph pools to thepoolssection with thevolumesrole:kubectl -n <MOSKClusterProject> edit kaascephcluster
Substitute
<MOSKClusterProject>with corresponding namespace of the MOSK cluster.Example configuration:
spec: cephClusterSpec: pools: - default: false deviceClass: hdd name: volumes replicated: size: 3 role: volumes - default: false deviceClass: hdd name: volumes-backend-1 replicated: size: 3 role: volumes - default: false deviceClass: hdd name: volumes-backend-2 replicated: size: 3 role: volumes
Verify that Cinder backend pools are created and ready:
kubectl -n <managedClusterProject> get kaascephcluster -o yaml
Example output:
status: fullClusterStatus: blockStorageStatus: poolsStatus: volumes-hdd: present: true status: observedGeneration: 1 phase: Ready volumes-backend-1-hdd: present: true status: observedGeneration: 1 phase: Ready volumes-backend-2-hdd: present: true status: observedGeneration: 1 phase: Ready
Verify that the added Ceph pools are accessible from the Cinder service. For example:
kubectl -n openstack exec -it cinder-volume-0 -- rbd ls -p volumes-backend-1-hdd -n client.cinder kubectl -n openstack exec -it cinder-volume-0 -- rbd ls -p volumes-backend-2-hdd -n client.cinder
After the Ceph pool becomes available, it is automatically specified as an additional Cinder backend and registered as a new volume type, which you can use to create Cinder volumes.
The following sections describe how to configure, manage, and verify specific aspects of a Ceph cluster.
Caution
Before you proceed with any reading or writing operation, first verify the cluster status using the ceph tool as described in Verify the Ceph core services.
Automated Ceph LCM¶
This section describes the supported automated Ceph lifecycle management (LCM) operations.
High-level workflow of Ceph OSD or node removal¶
The Ceph LCM automated operations such as Ceph OSD or Ceph node removal are
performed by creating a corresponding KaaSCephOperationRequest CR that
creates separate CephOsdRemoveRequest requests. It allows for automated
removal of healthy or non-healthy Ceph OSDs from a Ceph cluster and covers the
following scenarios:
Reducing hardware - all Ceph OSDs are up/in but you want to decrease the number of Ceph OSDs by reducing the number of disks or hosts.
Hardware issues. For example, if a host unexpectedly goes down and will not be restored, or if a disk on a host goes down and requires replacement.
This section describes the KaaSCephOperationRequest CR creation workflow,
specification, and request status.
For step-by-step procedures, refer to Automated Ceph LCM.
The workflow of creating a Ceph OSD removal request includes the following steps:
Removing obsolete nodes or disks from the
spec.nodessection of theKaaSCephClusterCR as described in Ceph advanced configuration.Note
Note the names of the removed nodes, devices or their paths exactly as they were specified in
KaaSCephClusterfor further usage.Creating a YAML template for the
KaaSCephOperationRequestCR. For details, see KaaSCephOperationRequest OSD removal specification.If
KaaSCephOperationRequestcontains information about Ceph OSDs to remove in a proper format, the information will be validated to eliminate human error and avoid a wrong Ceph OSD removal.If the
osdRemove.nodessection ofKaaSCephOperationRequestis empty, the Ceph Request Controller will automatically detect Ceph OSDs for removal, if any. Auto-detection is based not only on the information provided in theKaaSCephClusterbut also on the information from the Ceph cluster itself.
Once the validation or auto-detection completes, the entire information about the Ceph OSDs to remove appears in the
KaaSCephOperationRequestobject: hosts they belong to, OSD IDs, disks, partitions, and so on. The request then moves to theApproveWaitingphase until the Operator manually specifies theapproveflag in the spec.Manually adding an affirmative
approveflag in theKaaSCephOperationRequestspec. Once done, the Ceph Status Controller reconciliation pauses until the request is handled and executes the following:Stops regular Ceph Controller reconciliation
Removes Ceph OSDs
Runs batch jobs to clean up the device, if possible
Removes host information from the Ceph cluster if the entire Ceph node is removed
Marks the request with an appropriate result with a description of occurred issues
Note
If the request completes successfully, Ceph Controller reconciliation resumes. Otherwise, it remains paused until the issue is resolved.
Reviewing the Ceph OSD removal status. For details, see KaaSCephOperationRequest OSD removal status.
Manual removal of device cleanup jobs.
Note
Device cleanup jobs are not removed automatically and are kept in the
ceph-lcm-mirantisnamespace along with pods containing information about the executed actions. The jobs have the following labels:labels: app: miraceph-cleanup-disks host: <HOST-NAME> osd: <OSD-ID> rook-cluster: <ROOK-CLUSTER-NAME>
Additionally, jobs are labeled with disk names that will be cleaned up, such as
vdb=true. You can remove a single job or a group of jobs using any label described above, such as host, disk, and so on.
Example of KaaSCephOperationRequest resource
apiVersion: kaas.mirantis.com/v1alpha1
kind: KaaSCephOperationRequest
metadata:
name: remove-osd-3-4-request
namespace: managed-namespace
spec:
osdRemove:
approve: true
nodes:
worker-3:
cleanupByDevice:
- name: sdb
- path: /dev/disk/by-path/pci-0000:00:1t.9
kaasCephCluster:
name: ceph-cluster-managed-cluster
namespace: managed-namespace
Example of Ceph OSDs ready for removal
apiVersion: kaas.mirantis.com/v1alpha1
kind: KaaSCephOperationRequest
metadata:
generateName: remove-osds
namespace: managed-ns
spec:
osdRemove: {}
kaasCephCluster:
name: ceph-cluster-managed-cl
namespace: managed-ns
This section describes the KaaSCephOperationRequest CR specification used
to automatically create a CephOsdRemoveRequest request. For the procedure
workflow, see Creating a Ceph OSD removal request.
Parameter |
Description |
|---|---|
|
Describes the definition for the |
|
Defines spec:
kaasCephCluster:
name: kaas-mgmt
namespace: default
|
Parameter |
Description |
|---|---|
|
Map of Kubernetes nodes that specifies how to remove Ceph OSDs: by host-devices or OSD IDs. For details, see KaaSCephOperationRequest ‘nodes’ parameters spec. |
|
Flag that indicates whether a request is ready to execute removal. Can only be manually enabled by the Operator. For example: spec:
osdRemove:
approve: true
|
|
Flag used to keep requests in handling and not to proceed to the next
request if the If the For example: spec:
osdRemove:
keepOnFail: true
|
|
Optional. Flag that marks a finished request, even if it failed, to keep it in history. It allows resuming the Ceph Controller reconciliation without removing the failed request. The flag is used only by Ceph Controller and has no effect on request processing. Can only be manually specified. For example: spec:
osdRemove:
resolved: true
|
|
Optional. Flag used to resume a failed request and proceed with Ceph OSD
removal if the spec:
osdRemove:
resumeFailed: true
|
Parameter |
Description |
|---|---|
|
Flag used to clean up an entire node and drop it from the CRUSH map.
Mutually exclusive with |
|
List that describes devices to clean up by name or device path as they
were specified in
Warning Since MOSK 23.3, Mirantis does not recommend setting device
For details, see Container Cloud documentation: Addressing storage devices. |
|
List of Ceph OSD IDs to remove. Mutually exclusive with
|
Example of KaaSCephOperationRequest with
spec.osdRemove.nodes
apiVersion: kaas.mirantis.com/v1alpha1
kind: KaaSCephOperationRequest
metadata:
name: remove-osd-request
namespace: default
spec:
kaasCephCluster:
name: kaas-mgmt
namespace: default
osdRemove:
nodes:
"node-a":
completeCleanUp: true
"node-b":
cleanupByOsdId: [1, 15, 25]
"node-c":
cleanupByDevice:
- name: "sdb"
- path: "/dev/disk/by-path/pci-0000:00:1c.5"
- path: "/dev/disk/by-id/scsi-SATA_HGST_HUS724040AL_PN1334PEHN18ZS"
The example above includes the following actions:
For
node-a, full cleanup, including all OSDs on the node, node drop from the CRUSH map, and cleanup of all disks used for Ceph OSDs on this node.For
node-b, cleanup of Ceph OSDs with IDs1,15, and25along with the related disk information.For
node-c, cleanup of the device with namesdb, the device with path ID/dev/disk/by-path/pci-0000:00:1c.5, and the device withby-id/dev/disk/by-id/scsi-SATA_HGST_HUS724040AL_PN1334PEHN18ZS, dropping of OSDs running on these devices.
This section describes the status.osdRemoveStatus.removeInfo fields of the
KaaSCephOperationRequest CR that you can use to review a Ceph OSD or node
removal phases. The following diagram represents the phases flow:
Parameter |
Description |
|---|---|
|
Describes the status of the current |
|
The key-value mapping that reflects the management cluster machine names with their corresponding Kubernetes node names. |
Parameter |
Description |
|---|---|
|
Describes the current request phase that can be one of:
|
|
The overall information about the Ceph OSDs to remove: final removal
map, issues, and warnings. Once the The
|
|
Informational messages describing the reason for the request transition to the next phase. |
|
History of spec updates for the request. |
Example of status.osdRemoveStatus.removeInfo after
successful Validation
removeInfo:
cleanUpMap:
"node-a":
completeCleanUp: true
osdMapping:
"2":
deviceMapping:
"sdb":
path: "/dev/disk/by-path/pci-0000:00:0a.0"
partition: "/dev/ceph-a-vg_sdb/osd-block-b-lv_sdb"
type: "block"
class: "hdd"
zapDisk: true
"6":
deviceMapping:
"sdc":
path: "/dev/disk/by-path/pci-0000:00:0c.0"
partition: "/dev/ceph-a-vg_sdc/osd-block-b-lv_sdc-1"
type: "block"
class: "hdd"
zapDisk: true
"11":
deviceMapping:
"sdc":
path: "/dev/disk/by-path/pci-0000:00:0c.0"
partition: "/dev/ceph-a-vg_sdc/osd-block-b-lv_sdc-2"
type: "block"
class: "hdd"
zapDisk: true
"node-b":
osdMapping:
"1":
deviceMapping:
"sdb":
path: "/dev/disk/by-path/pci-0000:00:0a.0"
partition: "/dev/ceph-b-vg_sdb/osd-block-b-lv_sdb"
type: "block"
class: "ssd"
zapDisk: true
"15":
deviceMapping:
"sdc":
path: "/dev/disk/by-path/pci-0000:00:0b.1"
partition: "/dev/ceph-b-vg_sdc/osd-block-b-lv_sdc"
type: "block"
class: "ssd"
zapDisk: true
"25":
deviceMapping:
"sdd":
path: "/dev/disk/by-path/pci-0000:00:0c.2"
partition: "/dev/ceph-b-vg_sdd/osd-block-b-lv_sdd"
type: "block"
class: "ssd"
zapDisk: true
"node-c":
osdMapping:
"0":
deviceMapping:
"sdb":
path: "/dev/disk/by-path/pci-0000:00:1t.9"
partition: "/dev/ceph-c-vg_sdb/osd-block-c-lv_sdb"
type: "block"
class: "hdd"
zapDisk: true
"8":
deviceMapping:
"sde":
path: "/dev/disk/by-path/pci-0000:00:1c.5"
partition: "/dev/ceph-c-vg_sde/osd-block-c-lv_sde"
type: "block"
class: "hdd"
zapDisk: true
"sdf":
path: "/dev/disk/by-path/pci-0000:00:5a.5",
partition: "/dev/ceph-c-vg_sdf/osd-db-c-lv_sdf-1",
type: "db",
class: "ssd"
The example above is based on the example spec provided in
KaaSCephOperationRequest OSD removal specification.
During the Validation phase, the provided information was validated and
reflects the final map of the Ceph OSDs to remove:
For
node-a, Ceph OSDs with IDs2,6, and11will be removed with the related disk and its information: all block devices, names, paths, and disk class.For
node-b, the Ceph OSDs with IDs1,15, and25will be removed with the related disk information.For
node-c, the Ceph OSD with ID8will be removed, which is placed on the specifiedsdbdevice. The related partition on thesdfdisk, which is used as the BlueStore metadata device, will be cleaned up keeping the disk itself untouched. Other partitions on that device will not be touched.
Example of removeInfo with removeStatus succeeded
removeInfo:
cleanUpMap:
"node-a":
completeCleanUp: true
hostRemoveStatus:
status: Removed
osdMapping:
"2":
removeStatus:
osdRemoveStatus:
status: Removed
deploymentRemoveStatus:
status: Removed
name: "rook-ceph-osd-2"
deviceCleanUpJob:
status: Finished
name: "job-name-for-osd-2"
deviceMapping:
"sdb":
path: "/dev/disk/by-path/pci-0000:00:0a.0"
partition: "/dev/ceph-a-vg_sdb/osd-block-b-lv_sdb"
type: "block"
class: "hdd"
zapDisk: true
Example of removeInfo with removeStatus failed
removeInfo:
cleanUpMap:
"node-a":
completeCleanUp: true
osdMapping:
"2":
removeStatus:
osdRemoveStatus:
errorReason: "retries for cmd ‘ceph osd ok-to-stop 2’ exceeded"
status: Failed
deviceMapping:
"sdb":
path: "/dev/disk/by-path/pci-0000:00:0a.0"
partition: "/dev/ceph-a-vg_sdb/osd-block-b-lv_sdb"
type: "block"
class: "hdd"
zapDisk: true
Example of removeInfo with removeStatus failed by timeout
removeInfo:
cleanUpMap:
"node-a":
completeCleanUp: true
osdMapping:
"2":
removeStatus:
osdRemoveStatus:
errorReason: Timeout (30m0s) reached for waiting pg rebalance for osd 2
status: Failed
deviceMapping:
"sdb":
path: "/dev/disk/by-path/pci-0000:00:0a.0"
partition: "/dev/ceph-a-vg_sdb/osd-block-b-lv_sdb"
type: "block"
class: "hdd"
zapDisk: true
Note
In case of failures similar to the examples above, review the
ceph-request-controller logs and the Ceph cluster status. Such failures
may simply indicate timeout and retry issues. If no other issues were found,
re-create the request with a new name and skip adding successfully removed
Ceph OSDS or Ceph nodes.
Add, remove, or reconfigure Ceph nodes¶
Mirantis Ceph Controller simplifies a Ceph cluster management by automating LCM operations. This section describes how to add, remove, or reconfigure Ceph nodes.
Note
When adding a Ceph node with the Ceph Monitor role, if any issues occur with
the Ceph Monitor, rook-ceph removes it and adds a new Ceph Monitor instead,
named using the next alphabetic character in order. Therefore, the Ceph Monitor
names may not follow the alphabetical order. For example, a, b, d,
instead of a, b, c.
Prepare a new machine for the required managed cluster as described in Add a machine. During machine preparation, update the settings of the related bare metal host profile for the Ceph node being replaced with the required machine devices as described in Create a custom bare metal host profile.
Open the
KaasCephClusterCR of a managed cluster for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.In the
nodessection, specify the parameters for a Ceph node as required. For the parameters description, see Node parameters.The example configuration of the
nodessection with the new node:nodes: kaas-node-5bgk6: roles: - mon - mgr storageDevices: - config: deviceClass: hdd fullPath: /dev/disk/by-id/scsi-SATA_HGST_HUS724040AL_PN1334PEHN18ZS
nodes: kaas-node-5bgk6: roles: - mon - mgr storageDevices: - config: deviceClass: hdd name: sdb
Warning
Since MOSK 23.3, Mirantis highly recommends using the non-wwn
by-idsymlinks to specify storage devices in thestorageDeviceslist.For details, see Container Cloud documentation: Addressing storage devices.
Note
To use a new Ceph node for a Ceph Monitor or Ceph Manager deployment, also specify the
rolesparameter.Reducing the number of Ceph Monitors is not supported and causes the Ceph Monitor daemons removal from random nodes.
Removal of the
mgrrole in thenodessection of theKaaSCephClusterCR does not remove Ceph Managers. To remove a Ceph Manager from a node, remove it from thenodesspec and manually delete themgrpod in the Rook namespace.
Verify that all new Ceph daemons for the specified node have been successfully deployed in the Ceph cluster. The
fullClusterInfosection should not contain any issues.kubectl -n <managedClusterProjectName> get kaascephcluster -o yaml
Example of system response
status: fullClusterInfo: daemonsStatus: mgr: running: a is active mgr status: Ok mon: running: '3/3 mons running: [a b c] in quorum' status: Ok osd: running: '3/3 running: 3 up, 3 in' status: Ok
Note
Ceph node removal presupposes usage of a KaaSCephOperationRequest
CR. For workflow overview, spec and phases description, see
High-level workflow of Ceph OSD or node removal.
Note
To remove a Ceph node with a mon role, first move the Ceph
Monitor to another node and remove the mon role from the Ceph node as
described in Move a Ceph Monitor daemon to another node.
Open the
KaasCephClusterCR of a managed cluster for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.In the
spec.cephClusterSpec.nodessection, remove the required Ceph node specification.For example:
spec: cephClusterSpec: nodes: worker-5: # remove the entire entry for the required node storageDevices: {...} roles: [...]
Create a YAML template for the
KaaSCephOperationRequestCR. For example:apiVersion: kaas.mirantis.com/v1alpha1 kind: KaaSCephOperationRequest metadata: name: remove-osd-worker-5 namespace: <managedClusterProjectName> spec: osdRemove: nodes: worker-5: completeCleanUp: true kaasCephCluster: name: <kaasCephClusterName> namespace: <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding cluster namespace and<kaasCephClusterName>with the correspondingKaaSCephClustername.Apply the template on the management cluster in the corresponding namespace:
kubectl apply -f remove-osd-worker-5.yaml
Verify that the corresponding request has been created:
kubectl get kaascephoperationrequest remove-osd-worker-5 -n <managedClusterProjectName>
Verify that the
removeInfosection appeared in theKaaSCephOperationRequestCRstatus:kubectl -n <managedClusterProjectName> get kaascephoperationrequest remove-osd-worker-5 -o yaml
Example of system response
status: childNodesMapping: kaas-node-d4aac64d-1721-446c-b7df-e351c3025591: worker-5 osdRemoveStatus: removeInfo: cleanUpMap: kaas-node-d4aac64d-1721-446c-b7df-e351c3025591: osdMapping: "10": deviceMapping: sdb: path: "/dev/disk/by-path/pci-0000:00:1t.9" partition: "/dev/ceph-b-vg_sdb/osd-block-b-lv_sdb" type: "block" class: "hdd" zapDisk: true "16": deviceMapping: sdc: path: "/dev/disk/by-path/pci-0000:00:1t.10" partition: "/dev/ceph-b-vg_sdb/osd-block-b-lv_sdc" type: "block" class: "hdd" zapDisk: true
Verify that the
cleanUpMapsection matches the required removal and wait for theApproveWaitingphase to appear instatus:kubectl -n <managedClusterProjectName> get kaascephoperationrequest remove-osd-worker-5 -o yaml
Example of system response:
status: phase: ApproveWaiting
Edit the
KaaSCephOperationRequestCR and set theapproveflag totrue:kubectl -n <managedClusterProjectName> edit kaascephoperationrequest remove-osd-worker-5
For example:
spec: osdRemove: approve: true
Review the status of the
KaaSCephOperationRequestresource request processing. The valuable parameters are as follows:status.phase- the current state of request processingstatus.messages- the description of the current phasestatus.conditions- full history of request processing before the current phasestatus.removeInfo.issuesandstatus.removeInfo.warnings- contain error and warning messages occurred during request processing
Verify that the
KaaSCephOperationRequesthas been completed. For example:status: phase: Completed # or CompletedWithWarnings if there are non-critical issues
Remove the device cleanup jobs:
kubectl delete jobs -n ceph-lcm-mirantis -l app=miraceph-cleanup-disks
There is no hot reconfiguration procedure for existing Ceph OSDs and Ceph Monitors. To reconfigure an existing Ceph node, follow the steps below:
Remove the Ceph node from the Ceph cluster as described in Remove a Ceph node from a managed cluster.
Add the same Ceph node but with a modified configuration as described in Add Ceph nodes on a managed cluster.
Add, remove, or reconfigure Ceph OSDs¶
Mirantis Ceph Controller simplifies Ceph cluster management by automating LCM operations. This section describes how to add, remove, or reconfigure Ceph OSDs.
Manually prepare the required machine devices with LVM2 on the existing node because
BareMetalHostProfiledoes not support in-place changes.To add a Ceph OSD to an existing or hot-plugged raw device
If you want to add a Ceph OSD on top of a raw device that already exists on a node or is hot-plugged, add the required device using the following guidelines:
You can add a raw device to a node during node deployment.
If a node supports adding devices without node reboot, you can hot plug a raw device to a node.
If a node does not support adding devices without node reboot, you can hot plug a raw device during node shutdown. In this case, complete the following steps:
Enable maintenance mode on the managed cluster.
Turn off the required node.
Attach the required raw device to the node.
Turn on the required node.
Disable maintenance mode on the managed cluster.
Open the
KaasCephClusterCR of a managed cluster for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.In the
nodes.<machineName>.storageDevicessection, specify the parameters for a Ceph OSD as required. For the parameters description, see Node parameters.The example configuration of the
nodessection with the new node:nodes: kaas-node-5bgk6: roles: - mon - mgr storageDevices: - config: # existing item deviceClass: hdd fullPath: /dev/disk/by-id/scsi-SATA_HGST_HUS724040AL_PN1334PEHN18ZS - config: # new item deviceClass: hdd fullPath: /dev/disk/by-id/scsi-0ATA_HGST_HUS724040AL_PN1334PEHN1VBC
nodes: kaas-node-5bgk6: roles: - mon - mgr storageDevices: - config: # existing item deviceClass: hdd name: sdb - config: # new item deviceClass: hdd name: sdc
Warning
Since MOSK 23.3, Mirantis highly recommends using the non-wwn
by-idsymlinks to specify storage devices in thestorageDeviceslist.For details, see Container Cloud documentation: Addressing storage devices.
Verify that the Ceph OSD on the specified node is successfully deployed. The
fullClusterInfosection should not contain any issues.kubectl -n <managedClusterProjectName> get kaascephcluster -o yaml
For example:
status: fullClusterInfo: daemonsStatus: ... osd: running: '3/3 running: 3 up, 3 in' status: Ok
Note
Since MOSK 23.2,
cephDeviceMappingis removed because its large size can potentially exceed the Kubernetes 1.5 MB quota.Verify the Ceph OSD on the managed cluster:
kubectl -n rook-ceph get pod -l app=rook-ceph-osd -o wide | grep <machineName>
Note
Ceph OSD removal presupposes usage of a KaaSCephOperationRequest
CR. For workflow overview, spec and phases description, see
High-level workflow of Ceph OSD or node removal.
Warning
When using the non-recommended Ceph pools replicated.size of
less than 3, Ceph OSD removal cannot be performed. The minimal replica
size equals a rounded up half of the specified replicated.size.
For example, if replicated.size is 2, the minimal replica size is
1, and if replicated.size is 3, then the minimal replica size
is 2. The replica size of 1 allows Ceph having PGs with only one
Ceph OSD in the acting state, which may cause a PG_TOO_DEGRADED
health warning that blocks Ceph OSD removal. Mirantis recommends setting
replicated.size to 3 for each Ceph pool.
Open the
KaasCephClusterCR of a managed cluster for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.Remove the required Ceph OSD specification from the
spec.cephClusterSpec.nodes.<machineName>.storageDeviceslist:The example configuration of the
nodessection with the new node:nodes: kaas-node-5bgk6: roles: - mon - mgr storageDevices: - config: deviceClass: hdd fullPath: /dev/disk/by-id/scsi-SATA_HGST_HUS724040AL_PN1334PEHN18ZS - config: # remove the entire item entry from storageDevices list deviceClass: hdd fullPath: /dev/disk/by-id/scsi-0ATA_HGST_HUS724040AL_PN1334PEHN1VBC
nodes: kaas-node-5bgk6: roles: - mon - mgr storageDevices: - config: deviceClass: hdd name: sdb - config: # remove the entire item entry from storageDevices list deviceClass: hdd name: sdc
Create a YAML template for the
KaaSCephOperationRequestCR. Select from the following options:Remove Ceph OSD by device name,
by-pathsymlink, orby-idsymlink:apiVersion: kaas.mirantis.com/v1alpha1 kind: KaaSCephOperationRequest metadata: name: remove-osd-<machineName>-sdb namespace: <managedClusterProjectName> spec: osdRemove: nodes: <machineName>: cleanupByDevice: - name: sdb kaasCephCluster: name: <kaasCephClusterName> namespace: <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding cluster namespace and<kaasCephClusterName>with the correspondingKaaSCephClustername.Warning
Since MOSK 23.3, Mirantis does not recommend setting device
nameor deviceby-pathsymlink in thecleanupByDevicefield as these identifiers are not persistent and can change at node boot. Remove Ceph OSDs withby-idsymlinks specified in thepathfield or usecleanupByOsdIdinstead.For details, see Container Cloud documentation: Addressing storage devices.
Note
Since MOSK 23.1,
cleanupByDeviceis not supported if a device was physically removed from a node. Therefore, usecleanupByOsdIdinstead. For details, see Remove a failed Ceph OSD by Ceph OSD ID.Before MOSK 23.1, if the
storageDeviceitem was specified withby-id, specify thepathparameter in thecleanupByDevicesection instead ofname.If the
storageDeviceitem was specified with aby-pathdevice path, specify thepathparameter in thecleanupByDevicesection instead ofname.
Remove Ceph OSD by OSD ID:
apiVersion: kaas.mirantis.com/v1alpha1 kind: KaaSCephOperationRequest metadata: name: remove-osd-<machineName>-sdb namespace: <managedClusterProjectName> spec: osdRemove: nodes: <machineName>: cleanupByOsdId: - 2 kaasCephCluster: name: <kaasCephClusterName> namespace: <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding cluster namespace and<kaasCephClusterName>with the correspondingKaaSCephClustername.
Apply the template on the management cluster in the corresponding namespace:
kubectl apply -f remove-osd-<machineName>-sdb.yaml
Verify that the corresponding request has been created:
kubectl get kaascephoperationrequest remove-osd-<machineName>-sdb -n <managedClusterProjectName>
Verify that the
removeInfosection appeared in theKaaSCephOperationRequestCRstatus:kubectl -n <managedClusterProjectName> get kaascephoperationrequest remove-osd-<machineName>-sdb -o yaml
Example of system response:
status: childNodesMapping: kaas-node-d4aac64d-1721-446c-b7df-e351c3025591: <machineName> osdRemoveStatus: removeInfo: cleanUpMap: kaas-node-d4aac64d-1721-446c-b7df-e351c3025591: osdMapping: "10": deviceMapping: sdb: path: "/dev/disk/by-path/pci-0000:00:1t.9" partition: "/dev/ceph-b-vg_sdb/osd-block-b-lv_sdb" type: "block" class: "hdd" zapDisk: true
Verify that the
cleanUpMapsection matches the required removal and wait for theApproveWaitingphase to appear instatus:kubectl -n <managedClusterProjectName> get kaascephoperationrequest remove-osd-<machineName>-sdb -o yaml
Example of system response:
status: phase: ApproveWaiting
Edit the
KaaSCephOperationRequestCR and set theapproveflag totrue:kubectl -n <managedClusterProjectName> edit kaascephoperationrequest remove-osd-<machineName>-sdb
For example:
spec: osdRemove: approve: true
Review the following
statusfields of theKaaSCephOperationRequestCR request processing:status.phase- current state of request processingstatus.messages- description of the current phasestatus.conditions- full history of request processing before the current phasestatus.removeInfo.issuesandstatus.removeInfo.warnings- error and warning messages occurred during request processing, if any
Verify that the
KaaSCephOperationRequesthas been completed. For example:status: phase: Completed # or CompletedWithWarnings if there are non-critical issues
Remove the device cleanup jobs:
kubectl delete jobs -n ceph-lcm-mirantis -l app=miraceph-cleanup-disks
There is no hot reconfiguration procedure for existing Ceph OSDs. To reconfigure an existing Ceph node, follow the steps below:
Remove a Ceph OSD from the Ceph cluster as described in Remove a Ceph OSD from a managed cluster.
Add the same Ceph OSD but with a modified configuration as described in Add a Ceph OSD on a managed cluster.
Add, remove, or reconfigure Ceph OSDs with metadata devices¶
Mirantis Ceph Controller simplifies Ceph cluster management by automating LCM operations. This section describes how to add, remove, or reconfigure Ceph OSDs with a separate metadata device.
From the Ceph disks defined in the
BareMetalHostProfileobject that was configured using the Configure Ceph disks in a host profile procedure, select one disk for data and one logical volume for metadata of a Ceph OSD to be added to the Ceph cluster.Note
If you add a new disk after machine provisioning, manually prepare the required machine devices using Logical Volume Manager (LVM) 2 on the existing node because
BareMetalHostProfiledoes not support in-place changes.To add a Ceph OSD to an existing or hot-plugged raw device
If you want to add a Ceph OSD on top of a raw device that already exists on a node or is hot-plugged, add the required device using the following guidelines:
You can add a raw device to a node during node deployment.
If a node supports adding devices without node reboot, you can hot plug a raw device to a node.
If a node does not support adding devices without node reboot, you can hot plug a raw device during node shutdown. In this case, complete the following steps:
Enable maintenance mode on the managed cluster.
Turn off the required node.
Attach the required raw device to the node.
Turn on the required node.
Disable maintenance mode on the managed cluster.
Open the
KaasCephClusterobject for editing:kubectl -n <managedClusterProjectName> edit kaascephcluster
Substitute
<managedClusterProjectName>with the corresponding value.In the
nodes.<machineName>.storageDevicessection, specify the parameters for a Ceph OSD as required. For the parameters description, see Node parameters.The example configuration of the
nodessection with the new node:nodes: kaas-node-5bgk6: roles: - mon - mgr storageDevices: - config: # existing item deviceClass: hdd fullPath: /dev/disk/by-id/scsi-SATA_HGST_HUS724040AL_PN1334PEHN18ZS - config: # new item deviceClass: hdd metadataDevice: /dev/bluedb/meta_1 fullPath: /dev/disk/by-id/scsi-0ATA_HGST_HUS724040AL_PN1334PEHN1VBC
nodes: kaas-node-5bgk6: roles: - mon - mgr storageDevices: - config: # existing item deviceClass: hdd name: sdb - config: # new item deviceClass: hdd metadataDevice: /dev/bluedb/meta_1 name: sdc
Warning
Since MOSK 23.3, Mirantis highly recommends using the non-wwn
by-idsymlinks to specify storage devices in thestorageDeviceslist.For details, see Container Cloud documentation: Addressing storage devices.
Verify that the Ceph OSD is successfully deployed on the specified node:
kubectl -n <managedClusterProjectName> get kaascephcluster -o yaml
In the system response, the
fullClusterInfosection should not contain any issues.Example of a successful system response:
status: fullClusterInfo: daemonsStatus: ... osd: running: '4/4 running: 4 up, 4 in' status: Ok
Obtain the name of the node on which the machine with the Ceph OSD is running:
kubectl -n <managedClusterProjectName> get machine <machineName> -o jsonpath='{.status.nodeRef.name}'
Substitute
<managedClusterProjectName>and<machineName>with corresponding values.Verify the Ceph OSD status:
kubectl -n rook-ceph get pod -l app=rook-ceph-osd -o wide | grep <nodeName>
Substitute
<nodeName>with the value obtained on the previous step.Example of system response:
rook-ceph-osd-0-7b8d4d58db-f6czn 1/1 Running 0 42h 10.100.91.6 kaas-node-6c5e76f9-c2d2-4b1a-b047-3c299913a4bf <none> <none> rook-ceph-osd-1-78fbc47dc5-px9n2 1/1 Running 0 21h 10.100.91.6 kaas-node-6c5e76f9-c2d2-4b1a-b047-3c299913a4bf <none> <none> rook-ceph-osd-3-647f8d6c69-87gxt 1/1 Running 0 21h 10.100.91.6 kaas-node-6c5e76f9-c2d2-4b1a-b047-3c299913a4bf <none> <none>
Note
Ceph OSD removal implies the usage of the
KaaSCephOperationRequest custom resource (CR). For workflow overview,
spec and phases description, see High-level workflow of Ceph OSD or node removal.
Warning
When using the non-recommended Ceph pools replicated.size of
less than 3, Ceph OSD removal cannot be performed. The minimal replica
size equals a rounded up half of the specified replicated.size.
For example, if replicated.size is 2, the minimal replica size is
1, and if replicated.size is 3, then the minimal replica size
is 2. The replica size of 1 allows Ceph having PGs with only one
Ceph OSD in the acting state, which may cause a PG_TOO_DEGRADED
health warning that blocks Ceph OSD removal. Mirantis recommends setting
replicated.size to 3 for each Ceph pool.
Open the
KaasCephClusterobject of the managed cluster for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.Remove the required Ceph OSD specification from the
spec.cephClusterSpec.nodes.<machineName>.storageDeviceslist:The example configuration of the
nodessection with the new node:nodes: kaas-node-5bgk6: roles: - mon - mgr storageDevices: - config: deviceClass: hdd fullPath: /dev/disk/by-id/scsi-SATA_HGST_HUS724040AL_PN1334PEHN18ZS - config: # remove the entire item entry from storageDevices list deviceClass: hdd metadataDevice: /dev/bluedb/meta_1 fullPath: /dev/disk/by-id/scsi-0ATA_HGST_HUS724040AL_PN1334PEHN1VBC
nodes: kaas-node-5bgk6: roles: - mon - mgr storageDevices: - config: deviceClass: hdd name: sdb - config: # remove the entire item entry from storageDevices list deviceClass: hdd metadataDevice: /dev/bluedb/meta_1 name: sdc
Create a YAML template for the
KaaSCephOperationRequestCR. For example:apiVersion: kaas.mirantis.com/v1alpha1 kind: KaaSCephOperationRequest metadata: name: remove-osd-<machineName>-sdb namespace: <managedClusterProjectName> spec: osdRemove: nodes: <machineName>: cleanupByDevice: - name: sdb kaasCephCluster: name: <kaasCephClusterName> namespace: <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding cluster namespace and<kaasCephClusterName>with the correspondingKaaSCephClustername.Warning
Since MOSK 23.3, Mirantis does not recommend setting device
nameor deviceby-pathsymlink in thecleanupByDevicefield as these identifiers are not persistent and can change at node boot. Remove Ceph OSDs withby-idsymlinks specified in thepathfield or usecleanupByOsdIdinstead.For details, see Container Cloud documentation: Addressing storage devices.
Note
Since MOSK 23.1,
cleanupByDeviceis not supported if a device was physically removed from a node. Therefore, usecleanupByOsdIdinstead. For details, see Remove a failed Ceph OSD by Ceph OSD ID.Before MOSK 23.1, if the
storageDeviceitem was specified withby-id, specify thepathparameter in thecleanupByDevicesection instead ofname.If the
storageDeviceitem was specified with aby-pathdevice path, specify thepathparameter in thecleanupByDevicesection instead ofname.
Apply the template on the management cluster in the corresponding namespace:
kubectl apply -f remove-osd-<machineName>-sdb.yaml
Verify that the corresponding request has been created:
kubectl get kaascephoperationrequest remove-osd-<machineName>-sdb -n <managedClusterProjectName>
Verify that the
removeInfosection appeared in theKaaSCephOperationRequestCRstatus:kubectl -n <managedClusterProjectName> get kaascephoperationrequest remove-osd-<machineName>-sdb -o yaml
Example of system response:
status: childNodesMapping: kaas-node-d4aac64d-1721-446c-b7df-e351c3025591: <machineName> osdRemoveStatus: removeInfo: cleanUpMap: kaas-node-d4aac64d-1721-446c-b7df-e351c3025591: osdMapping: "10": deviceMapping: sdb: path: "/dev/disk/by-path/pci-0000:00:1t.9" partition: "/dev/ceph-b-vg_sdb/osd-block-b-lv_sdb" type: "block" class: "hdd" zapDisk: true "5": deviceMapping: /dev/sdc: deviceClass: hdd devicePath: /dev/disk/by-path/pci-0000:00:0f.0 devicePurpose: block usedPartition: /dev/ceph-2d11bf90-e5be-4655-820c-fb4bdf7dda63/osd-block-e41ce9a8-4925-4d52-aae4-e45167cfcf5c zapDisk: true /dev/sdf: deviceClass: hdd devicePath: /dev/disk/by-path/pci-0000:00:12.0 devicePurpose: db usedPartition: /dev/bluedb/meta_1
Verify that the
cleanUpMapsection matches the required removal and wait for theApproveWaitingphase to appear instatus:kubectl -n <managedClusterProjectName> get kaascephoperationrequest remove-osd-<machineName>-sdb -o yaml
Example of system response:
status: phase: ApproveWaiting
In the
KaaSCephOperationRequestCR, set theapproveflag totrue:kubectl -n <managedClusterProjectName> edit kaascephoperationrequest remove-osd-<machineName>-sdb
Configuration snippet:
spec: osdRemove: approve: true
Review the following
statusfields of theKaaSCephOperationRequestCR request processing:status.phase- current state of request processingstatus.messages- description of the current phasestatus.conditions- full history of request processing before the current phasestatus.removeInfo.issuesandstatus.removeInfo.warnings- error and warning messages occurred during request processing, if any
Verify that the
KaaSCephOperationRequesthas been completed.Example of the positive
status.phasefield:status: phase: Completed # or CompletedWithWarnings if there are non-critical issues
Remove the device cleanup jobs:
kubectl delete jobs -n ceph-lcm-mirantis -l app=miraceph-cleanup-disks
There is no hot reconfiguration procedure for existing Ceph OSDs. To reconfigure an existing Ceph node, remove and re-add a Ceph OSD with a metadata device using the following options:
Since Container Cloud 2.24.0, if metadata device partitions are specified in the
BareMetalHostProfileobject as described in Configure Ceph disks in a host profile, the metadata device definition is an LVM path inmetadataDeviceof theKaaSCephClusterobject.Therefore, automated LCM will clean up the logical volume without removal and it can be reused. For this reason, to reconfigure a partition of a Ceph OSD metadata device:
Remove a Ceph OSD from the Ceph cluster as described in Remove a Ceph OSD with a metadata device.
Add the same Ceph OSD but with a modified configuration as described in Add a Ceph OSD with a metadata device.
Before MOSK 23.2 or if metadata device partitions are not specified in the
BareMetalHostProfileobject as described in Configure Ceph disks in a host profile, the most common definition of a metadata device is a full device name (by-pathorby-id) inmetadataDeviceof theKaaSCephClusterobject for Ceph OSD. For example,metadataDevice: /dev/nvme0n1. In this case, to reconfigure a partition of a Ceph OSD metadata device:Remove a Ceph OSD from the Ceph cluster as described in Remove a Ceph OSD with a metadata device. Automated LCM will clean up the data device and will remove the metadata device partition for the required Ceph OSD.
Reconfigure the metadata device partition manually to use it during addition of a new Ceph OSD.
Manual reconfiguration of a metadata device partition
Log in to the Ceph node running a Ceph OSD to reconfigure.
Find the required metadata device used for Ceph OSDs that should have LVM partitions with the
osd--dbsubstring:lsblk
Example of system response:
... vdf 252:80 0 32G 0 disk ├─ceph--7831901d--398e--415d--8941--e78486f3b019-osd--db--4bdbb0a0--e613--416e--ab97--272f237b7eab │ 253:3 0 16G 0 lvm └─ceph--7831901d--398e--415d--8941--e78486f3b019-osd--db--8f439d5c--1a19--49d5--b71f--3c25ae343303 253:5 0 16G 0 lvm
Capture the volume group UUID and logical volume sizes. In the example above, the volume group UUID is
ceph--7831901d--398e--415d--8941--e78486f3b019and the size is16G.Find the volume group of the metadata device:
vgs
Example of system response:
VG #PV #LV #SN Attr VSize VFree ceph-508c7a6d-db01-4873-98c3-52ab204b5ca8 1 1 0 wz--n- <32.00g 0 ceph-62d84b29-8de5-440c-a6e9-658e8e246af7 1 1 0 wz--n- <32.00g 0 ceph-754e0772-6d0f-4629-bf1d-24cb79f3ee82 1 1 0 wz--n- <32.00g 0 ceph-7831901d-398e-415d-8941-e78486f3b019 1 2 0 wz--n- <48.00g <17.00g lvm_root 1 1 0 wz--n- <61.03g 0
Capture the volume group with the name that matches the prefix of LVM partitions of the metadata device. In the example above, the required volume group is
ceph-7831901d-398e-415d-8941-e78486f3b019.Make a manual LVM partitioning for the new Ceph OSD. Create a new logical volume in the obtained volume group:
lvcreate -L <lvSize> -n <lvName> <vgName>
Substitute the following parameters:
<lvSize>with the previously obtained logical volume size. In the example above, it is16G.<lvName>with a new logical volume name. For example,meta_1.<vgName>with the previously obtained volume group name. In the example above, it isceph-7831901d-398e-415d-8941-e78486f3b019.
Note
Manually created partitions can be removed only manually, or during a complete metadata disk removal, or during the
Machineobject removal or re-provisioning.
Add the same Ceph OSD but with a modified configuration and manually created logical volume of the metadata device as described in Add a Ceph OSD with a metadata device.
For example, instead of
metadataDevice: /dev/bluedb/meta_1definemetadataDevice: /dev/ceph-7831901d-398e-415d-8941-e78486f3b019/meta_1that was manually created in the previous step.
Replace a failed Ceph OSD¶
After a physical disk replacement, you can use Ceph LCM API to redeploy a failed Ceph OSD. The common flow of replacing a failed Ceph OSD is as follows:
Remove the obsolete Ceph OSD from the Ceph cluster by device name, by Ceph OSD ID, or by path.
Add a new Ceph OSD on the new disk to the Ceph cluster.
Note
Ceph OSD replacement presupposes usage of a
KaaSCephOperationRequest CR. For workflow overview, spec and phases
description, see High-level workflow of Ceph OSD or node removal.
Warning
The procedure below presuppose that the operator knows the exact
device name, by-path, or by-id of the replaced device, as well as on
which node the replacement occurred.
Warning
Since Container Cloud 2.23.1 (Cluster release 12.7.0),
a Ceph OSD removal using by-path, by-id, or device name is not
supported if a device was physically removed from a node. Therefore, use
cleanupByOsdId instead. For details, see
Remove a failed Ceph OSD by Ceph OSD ID.
Warning
Since MOSK 23.3, Mirantis does not recommend setting device
name or device by-path symlink in the cleanupByDevice field
as these identifiers are not persistent and can change at node boot. Remove
Ceph OSDs with by-id symlinks specified in the path field or use
cleanupByOsdId instead.
For details, see Container Cloud documentation: Addressing storage devices.
Open the
KaasCephClusterCR of a managed cluster for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.In the
nodessection, remove the required device:spec: cephClusterSpec: nodes: <machineName>: storageDevices: - name: <deviceName> # remove the entire item from storageDevices list # fullPath: <deviceByPath> if device is specified with symlink instead of name config: deviceClass: hdd
Substitute
<machineName>with the machine name of the node where the device<deviceName>or<deviceByPath>is going to be replaced.Save
KaaSCephClusterand close the editor.Create a
KaaSCephOperationRequestCR template and save it asreplace-failed-osd-<machineName>-<deviceName>-request.yaml:apiVersion: kaas.mirantis.com/v1alpha1 kind: KaaSCephOperationRequest metadata: name: replace-failed-osd-<machineName>-<deviceName> namespace: <managedClusterProjectName> spec: osdRemove: nodes: <machineName>: cleanupByDevice: - name: <deviceName> # If a device is specified with by-path or by-id instead of # name, path: <deviceByPath> or <deviceById>. kaasCephCluster: name: <kaasCephClusterName> namespace: <managedClusterProjectName>
Substitute
<kaasCephClusterName>with the correspondingKaaSCephClusterresource from the<managedClusterProjectName>namespace.Apply the template to the cluster:
kubectl apply -f replace-failed-osd-<machineName>-<deviceName>-request.yaml
Verify that the corresponding request has been created:
kubectl get kaascephoperationrequest -n <managedClusterProjectName>
Verify that the
removeInfosection appeared in theKaaSCephOperationRequestCRstatus:kubectl -n <managedClusterProjectName> get kaascephoperationrequest replace-failed-osd-<machineName>-<deviceName> -o yaml
Example of system response:
status: childNodesMapping: <nodeName>: <machineName> osdRemoveStatus: removeInfo: cleanUpMap: <nodeName>: osdMapping: <osdId>: deviceMapping: <dataDevice>: deviceClass: hdd devicePath: <dataDeviceByPath> devicePurpose: block usedPartition: /dev/ceph-d2d3a759-2c22-4304-b890-a2d87e056bd4/osd-block-ef516477-d2da-492f-8169-a3ebfc3417e2 zapDisk: true
Definition of values in angle brackets:
<machineName>- name of the machine on which the device is being replaced, for example,worker-1<nodeName>- underlying node name of the machine, for example,kaas-node-5a74b669-7e53-4535-aabd-5b509ec844af<osdId>- Ceph OSD ID for the device being replaced, for example,1<dataDeviceByPath>-by-pathof the device placed on the node, for example,/dev/disk/by-path/pci-0000:00:1t.9<dataDevice>- name of the device placed on the node, for example,/dev/sdb
Verify that the
cleanUpMapsection matches the required removal and wait for theApproveWaitingphase to appear instatus:kubectl -n <managedClusterProjectName> get kaascephoperationrequest replace-failed-osd-<machineName>-<deviceName> -o yaml
Example of system response:
status: phase: ApproveWaiting
Edit the
KaaSCephOperationRequestCR and set theapproveflag totrue:kubectl -n <managedClusterProjectName> edit kaascephoperationrequest replace-failed-osd-<machineName>-<deviceName>
For example:
spec: osdRemove: approve: true
Review the following
statusfields of theKaaSCephOperationRequestCR request processing:status.phase- current state of request processingstatus.messages- description of the current phasestatus.conditions- full history of request processing before the current phasestatus.removeInfo.issuesandstatus.removeInfo.warnings- error and warning messages occurred during request processing, if any
Verify that the
KaaSCephOperationRequesthas been completed. For example:status: phase: Completed # or CompletedWithWarnings if there are non-critical issues
Remove the device cleanup jobs:
kubectl delete jobs -n ceph-lcm-mirantis -l app=miraceph-cleanup-disks
Caution
The procedure below presupposes that the operator knows only the failed Ceph OSD ID.
Identify the node and device names used by the affected Ceph OSD:
Using the Ceph CLI in the
rook-ceph-toolsPod, run:kubectl -n rook-ceph exec -it deploy/rook-ceph-tools -- ceph osd metadata <osdId>
Substitute
<osdId>with the affected OSD ID.Example output:
{ "id": 1, ... "bluefs_db_devices": "vdc", ... "bluestore_bdev_devices": "vde", ... "devices": "vdc,vde", ... "hostname": "kaas-node-6c5e76f9-c2d2-4b1a-b047-3c299913a4bf", ... },
In the example above,
hostnameis the node name anddevicesare all devices used by the affected Ceph OSD.In the
statussection of theKaaSCephClusterCR, obtain theosd-devicemapping:kubectl get kaascephcluster -n <managedClusterProjectName> -o yaml
Substitute
<managedClusterProjectName>with the corresponding value.For example:
status: fullClusterInfo: cephDetails: cephDeviceMapping: <nodeName>: <osdId>: <deviceName>
In the system response, capture the following parameters:
<nodeName>- the corresponding node name that hosts the Ceph OSD<osdId>- the ID of the Ceph OSD to replace<deviceName>- an actual device name to replace
Obtain
<machineName>for<nodeName>where the Ceph OSD is placed:kubectl -n rook-ceph get node -o jsonpath='{range .items[*]}{@.metadata.name}{" "}{@.metadata.labels.kaas\.mirantis\.com\/machine-name}{"\n"}{end}'
Open the
KaasCephClusterCR of a managed cluster for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.In the
nodessection, remove the required device:spec: cephClusterSpec: nodes: <machineName>: storageDevices: - name: <deviceName> # remove the entire item from storageDevices list config: deviceClass: hdd
Substitute
<machineName>with the machine name of the node where the device<deviceName>is going to be replaced.Save
KaaSCephClusterand close the editor.Create a
KaaSCephOperationRequestCR template and save it asreplace-failed-<machineName>-osd-<osdId>-request.yaml:apiVersion: kaas.mirantis.com/v1alpha1 kind: KaaSCephOperationRequest metadata: name: replace-failed-<machineName>-osd-<osdId> namespace: <managedClusterProjectName> spec: osdRemove: nodes: <machineName>: cleanupByOsdId: - <osdId> kaasCephCluster: name: <kaasCephClusterName> namespace: <managedClusterProjectName>
Substitute
<kaasCephClusterName>with the correspondingKaaSCephClusterresource from the<managedClusterProjectName>namespace.Apply the template to the cluster:
kubectl apply -f replace-failed-<machineName>-osd-<osdId>-request.yaml
Verify that the corresponding request has been created:
kubectl get kaascephoperationrequest -n <managedClusterProjectName>
Verify that the
removeInfosection appeared in theKaaSCephOperationRequestCRstatus:kubectl -n <managedClusterProjectName> get kaascephoperationrequest replace-failed-<machineName>-osd-<osdId>-request -o yaml
Example of system response
status: childNodesMapping: <nodeName>: <machineName> osdRemoveStatus: removeInfo: cleanUpMap: <nodeName>: osdMapping: <osdId>: deviceMapping: <dataDevice>: deviceClass: hdd devicePath: <dataDeviceByPath> devicePurpose: block usedPartition: /dev/ceph-d2d3a759-2c22-4304-b890-a2d87e056bd4/osd-block-ef516477-d2da-492f-8169-a3ebfc3417e2 zapDisk: true
Definition of values in angle brackets:
<machineName>- name of the machine on which the device is being replaced, for example,worker-1<nodeName>- underlying node name of the machine, for example,kaas-node-5a74b669-7e53-4535-aabd-5b509ec844af<osdId>- Ceph OSD ID for the device being replaced, for example,1<dataDeviceByPath>-by-pathof the device placed on the node, for example,/dev/disk/by-path/pci-0000:00:1t.9<dataDevice>- name of the device placed on the node, for example,/dev/sdb
Verify that the
cleanUpMapsection matches the required removal and wait for theApproveWaitingphase to appear instatus:kubectl -n <managedClusterProjectName> get kaascephoperationrequest replace-failed-<machineName>-osd-<osdId>-request -o yaml
Example of system response:
status: phase: ApproveWaiting
Edit the
KaaSCephOperationRequestCR and set theapproveflag totrue:kubectl -n <managedClusterProjectName> edit kaascephoperationrequest replace-failed-<machineName>-osd-<osdId>-request
For example:
spec: osdRemove: approve: true
Review the following
statusfields of theKaaSCephOperationRequestCR request processing:status.phase- current state of request processingstatus.messages- description of the current phasestatus.conditions- full history of request processing before the current phasestatus.removeInfo.issuesandstatus.removeInfo.warnings- error and warning messages occurred during request processing, if any
Verify that the
KaaSCephOperationRequesthas been completed. For example:status: phase: Completed # or CompletedWithWarnings if there are non-critical issues
Remove the device cleanup jobs:
kubectl delete jobs -n ceph-lcm-mirantis -l app=miraceph-cleanup-disks
Note
You can spawn Ceph OSD on a raw device, but it must be clean and without any data or partitions. If you want to add a device that was in use, also ensure it is raw and clean. To clean up all data and partitions from a device, refer to official Rook documentation.
If you want to add a Ceph OSD on top of a raw device that already exists on a node or is hot-plugged, add the required device using the following guidelines:
You can add a raw device to a node during node deployment.
If a node supports adding devices without node reboot, you can hot plug a raw device to a node.
If a node does not support adding devices without node reboot, you can hot plug a raw device during node shutdown. In this case, complete the following steps:
Enable maintenance mode on the managed cluster.
Turn off the required node.
Attach the required raw device to the node.
Turn on the required node.
Disable maintenance mode on the managed cluster.
Open the
KaasCephClusterCR of a managed cluster for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.In the
nodessection, add a new device:spec: cephClusterSpec: nodes: <machineName>: storageDevices: - fullPath: <deviceByID> # Since 2.25.0 (17.0.0) if device is supposed to be added with by-id # name: <deviceByID> # Prior MCC 2.25.0 if device is supposed to be added with by-id # fullPath: <deviceByPath> # if device is supposed to be added with by-path config: deviceClass: hdd
Substitute
<machineName>with the machine name of the node where device<deviceName>or<deviceByPath>is going to be added as a Ceph OSD.Verify that the new Ceph OSD has appeared in the Ceph cluster and is
inandup. ThefullClusterInfosection should not contain any issues.kubectl -n <managedClusterProjectName> get kaascephcluster -o yaml
For example:
status: fullClusterInfo: daemonStatus: osd: running: '3/3 running: 3 up, 3 in' status: Ok
Replace a failed Ceph OSD with a metadata device¶
The document describes various scenarios of a Ceph OSD outage and recovery or replacement. More specifically, this section describes how to replace a failed Ceph OSD with a metadata device:
If the metadata device is specified as a logical volume in the
BareMetalHostProfileobject and defined in theKaaSCephClusterobject as a logical volume pathIf the metadata device is specified in the
KaaSCephClusterobject as a device name
Note
Ceph OSD replacement implies the usage of the
KaaSCephOperationRequest custom resource (CR). For workflow overview,
spec and phases description, see High-level workflow of Ceph OSD or node removal.
You can apply the below procedure in the following cases:
A Ceph OSD failed without data or metadata device outage. In this case, first remove a failed Ceph OSD and clean up all corresponding disks and partitions. Then add a new Ceph OSD to the same data and metadata paths.
A Ceph OSD failed with data or metadata device outage. In this case, you also first remove a failed Ceph OSD and clean up all corresponding disks and partitions. Then add a new Ceph OSD to a newly replaced data device with the same metadata path.
Note
The below procedure also applies to manually created metadata partitions.
Identify the ID of Ceph OSD related to a failed device. For example, use the Ceph CLI in the
rook-ceph-toolsPod:ceph osd metadata
Example of system response:
{ "id": 0, ... "bluestore_bdev_devices": "vdc", ... "devices": "vdc", ... "hostname": "kaas-node-6c5e76f9-c2d2-4b1a-b047-3c299913a4bf", ... "pod_name": "rook-ceph-osd-0-7b8d4d58db-f6czn", ... }, { "id": 1, ... "bluefs_db_devices": "vdf", ... "bluestore_bdev_devices": "vde", ... "devices": "vde,vdf", ... "hostname": "kaas-node-6c5e76f9-c2d2-4b1a-b047-3c299913a4bf", ... "pod_name": "rook-ceph-osd-1-78fbc47dc5-px9n2", ... }, ...
Open the
KaasCephClustercustom resource (CR) for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.In the
nodessection:Find and capture the
metadataDevicepath to reuse it during re-creation of the Ceph OSD.Remove the required device:
Example configuration snippet:
spec: cephClusterSpec: nodes: <machineName>: storageDevices: - name: <deviceName> # remove the entire item from the storageDevices list # fullPath: <deviceByPath> if device is specified using by-path instead of name config: deviceClass: hdd metadataDevice: /dev/bluedb/meta_1
In the example above,
<machineName>is the name of machine that relates to the node on which the device<deviceName>or<deviceByPath>must be replaced.Create a
KaaSCephOperationRequestCR template and save it asreplace-failed-osd-<machineName>-<osdID>-request.yaml:apiVersion: kaas.mirantis.com/v1alpha1 kind: KaaSCephOperationRequest metadata: name: replace-failed-osd-<machineName>-<deviceName> namespace: <managedClusterProjectName> spec: osdRemove: nodes: <machineName>: cleanupByOsdId: - <osdID> kaasCephCluster: name: <kaasCephClusterName> namespace: <managedClusterProjectName>
Substitute the following parameters:
<machineName>and<deviceName>with the machine and device names from the previous step<managedClusterProjectName>with the cluster project name<osdID>with the ID of the affected Ceph OSD<kaasCephClusterName>with theKaaSCephClusterresource name<managedClusterProjectName>with the project name of the related managed cluster
Apply the template to the cluster:
kubectl apply -f replace-failed-osd-<machineName>-<osdID>-request.yaml
Verify that the corresponding request has been created:
kubectl get kaascephoperationrequest -n <managedClusterProjectName>
Verify that the
statussection ofKaaSCephOperationRequestcontains theremoveInfosection:kubectl -n <managedClusterProjectName> get kaascephoperationrequest replace-failed-osd-<machineName>-<osdID> -o yaml
Example of system response:
childNodesMapping: <nodeName>: <machineName> removeInfo: cleanUpMap: <nodeName>: osdMapping: "<osdID>": deviceMapping: <dataDevice>: deviceClass: hdd devicePath: <dataDeviceByPath> devicePurpose: block usedPartition: /dev/ceph-d2d3a759-2c22-4304-b890-a2d87e056bd4/osd-block-ef516477-d2da-492f-8169-a3ebfc3417e2 zapDisk: true <metadataDevice>: deviceClass: hdd devicePath: <metadataDeviceByPath> devicePurpose: db usedPartition: /dev/bluedb/meta_1 uuid: ef516477-d2da-492f-8169-a3ebfc3417e2
Definition of values in angle brackets:
<machineName>- name of the machine on which the device is being replaced, for example,worker-1<nodeName>- underlying node name of the machine, for example,kaas-node-5a74b669-7e53-4535-aabd-5b509ec844af<osdId>- Ceph OSD ID for the device being replaced, for example,1<dataDeviceByPath>-by-pathof the device placed on the node, for example,/dev/disk/by-path/pci-0000:00:1t.9<dataDevice>- name of the device placed on the node, for example,/dev/vde<metadataDevice>- metadata name of the device placed on the node, for example,/dev/vdf<metadataDeviceByPath>- metadataby-pathof the device placed on the node, for example,/dev/disk/by-path/pci-0000:00:12.0
Note
The partitions that are manually created or configured using the
BareMetalHostProfileobject can be removed only manually, or during a complete metadata disk removal, or during theMachineobject removal or re-provisioning.Verify that the
cleanUpMapsection matches the required removal and wait for theApproveWaitingphase to appear instatus:kubectl -n <managedClusterProjectName> get kaascephoperationrequest replace-failed-osd-<machineName>-<osdID> -o yaml
Example of system response:
status: phase: ApproveWaiting
In the
KaaSCephOperationRequestCR, set theapproveflag totrue:kubectl -n <managedClusterProjectName> edit kaascephoperationrequest replace-failed-osd-<machineName>-<osdID>
Configuration snippet:
spec: osdRemove: approve: true
Review the following
statusfields of theKaaSCephOperationRequestCR request processing:status.phase- current state of request processingstatus.messages- description of the current phasestatus.conditions- full history of request processing before the current phasestatus.removeInfo.issuesandstatus.removeInfo.warnings- error and warning messages occurred during request processing, if any
Verify that the
KaaSCephOperationRequesthas been completed. For example:status: phase: Completed # or CompletedWithWarnings if there are non-critical issues
Note
You can spawn Ceph OSD on a raw device, but it must be clean and without any data or partitions. If you want to add a device that was in use, also ensure it is raw and clean. To clean up all data and partitions from a device, refer to official Rook documentation.
If you want to add a Ceph OSD on top of a raw device that already exists on a node or is hot-plugged, add the required device using the following guidelines:
You can add a raw device to a node during node deployment.
If a node supports adding devices without node reboot, you can hot plug a raw device to a node.
If a node does not support adding devices without node reboot, you can hot plug a raw device during node shutdown. In this case, complete the following steps:
Enable maintenance mode on the managed cluster.
Turn off the required node.
Attach the required raw device to the node.
Turn on the required node.
Disable maintenance mode on the managed cluster.
Open the
KaasCephClusterCR for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.In the
nodessection, add the replaced device with the samemetadataDevicepath as on the removed Ceph OSD. For example:spec: cephClusterSpec: nodes: <machineName>: storageDevices: - name: <deviceByID> # Recommended. Add a new device by ID, for example, /dev/disk/by-id/... #fullPath: <deviceByPath> # Add a new device by path, for example, /dev/disk/by-path/... config: deviceClass: hdd metadataDevice: /dev/bluedb/meta_1 # Must match the value of the previously removed OSD
Substitute
<machineName>with the machine name of the node where the new device<deviceByID>or<deviceByPath>must be added.Wait for the replaced disk to apply to the Ceph cluster as a new Ceph OSD.
You can monitor the application state using either the
statussection of theKaaSCephClusterCR or in therook-ceph-toolsPod:kubectl -n rook-ceph exec -it deploy/rook-ceph-tools -- ceph -s
You can apply the below procedure if a Ceph OSD failed with data disk outage
and the metadata partition is not specified in the BareMetalHostProfile
custom resource (CR). This scenario implies that the Ceph cluster
automatically creates a required metadata logical volume on a desired device.
To remove the affected Ceph OSD with a metadata device as a device name, follow the Remove a failed Ceph OSD by ID with a defined metadata device procedure and capture the following details:
While editing
KaasCephClusterin thenodessection, capture themetadataDevicepath to reuse it during re-creation of the Ceph OSD.Example of the
spec.nodessection:spec: cephClusterSpec: nodes: <machineName>: storageDevices: - name: <deviceName> # remove the entire item from the storageDevices list # fullPath: <deviceByPath> if device is specified using by-path instead of name config: deviceClass: hdd metadataDevice: /dev/nvme0n1
In the example above, save the
metadataDevicedevice name/dev/nvme0n1.During verification of
removeInfo, capture theusedPartitionvalue of the metadata device located in thedeviceMapping.<metadataDevice>section.Example of the
removeInfosection:removeInfo: cleanUpMap: <nodeName>: osdMapping: "<osdID>": deviceMapping: <dataDevice>: deviceClass: hdd devicePath: <dataDeviceByPath> devicePurpose: block usedPartition: /dev/ceph-d2d3a759-2c22-4304-b890-a2d87e056bd4/osd-block-ef516477-d2da-492f-8169-a3ebfc3417e2 zapDisk: true <metadataDevice>: deviceClass: hdd devicePath: <metadataDeviceByPath> devicePurpose: db usedPartition: /dev/ceph-b0c70c72-8570-4c9d-93e9-51c3ab4dd9f9/osd-db-ecf64b20-1e07-42ac-a8ee-32ba3c0b7e2f uuid: ef516477-d2da-492f-8169-a3ebfc3417e2
In the example above, capture the following values from the
<metadataDevice>section:ceph-b0c70c72-8570-4c9d-93e9-51c3ab4dd9f9- name of the volume group that contains all metadata partitions on the<metadataDevice>diskosd-db-ecf64b20-1e07-42ac-a8ee-32ba3c0b7e2f- name of the logical volume that relates to a failed Ceph OSD
After you remove the Ceph OSD disk, manually create a separate logical volume for the metadata partition in an existing volume group on the metadata device:
lvcreate -l 100%FREE -n meta_1 <vgName>
Subtitute <vgName> with the name of a volume group captured in the
usedPartiton parameter.
Note
If you removed more than one OSD, replace 100%FREE with the
corresponding partition size. For example:
lvcreate -l <partitionSize> -n meta_1 <vgName>
Substitute <partitionSize> with the corresponding value that matches the
size of other partitions placed on the affected metadata drive. To obtain
<partitionSize>, use the output of the lvs command. For example:
16G.
During execution of the lvcreate command, the system asks you to wipe the found bluestore label on a metadata device. For example:
WARNING: ceph_bluestore signature detected on /dev/ceph-b0c70c72-8570-4c9d-93e9-51c3ab4dd9f9/meta_1 at offset 0. Wipe it? [y/n]:
Using the interactive shell, answer n to keep all metadata partitions
alive. After answering n, the system outputs the following:
Aborted wiping of ceph_bluestore.
1 existing signature left on the device.
Logical volume "meta_1" created.
Note
You can spawn Ceph OSD on a raw device, but it must be clean and without any data or partitions. If you want to add a device that was in use, also ensure it is raw and clean. To clean up all data and partitions from a device, refer to official Rook documentation.
If you want to add a Ceph OSD on top of a raw device that already exists on a node or is hot-plugged, add the required device using the following guidelines:
You can add a raw device to a node during node deployment.
If a node supports adding devices without node reboot, you can hot plug a raw device to a node.
If a node does not support adding devices without node reboot, you can hot plug a raw device during node shutdown. In this case, complete the following steps:
Enable maintenance mode on the managed cluster.
Turn off the required node.
Attach the required raw device to the node.
Turn on the required node.
Disable maintenance mode on the managed cluster.
Open the
KaasCephClusterCR for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.In the
nodessection, add the replaced device with the samemetadataDevicepath as in the previous Ceph OSD:spec: cephClusterSpec: nodes: <machineName>: storageDevices: - fullPath: <deviceByID> # Recommended since MCC 2.25.0 (17.0.0). # Add a new device by-id symlink, for example, /dev/disk/by-id/... #name: <deviceByID> # Add a new device by ID, for example, /dev/disk/by-id/... #fullPath: <deviceByPath> # Add a new device by path, for example, /dev/disk/by-path/... config: deviceClass: hdd metadataDevice: /dev/<vgName>/meta_1
Substitute
<machineName>with the machine name of the node where the new device<deviceByID>or<deviceByPath>must be added. Also specifymetadataDevicewith the path to the logical volume created during the Re-create the metadata partition on the existing metadata disk procedure.Wait for the replaced disk to apply to the Ceph cluster as a new Ceph OSD.
You can monitor the application state using either the
statussection of theKaaSCephClusterCR or in therook-ceph-toolsPod:kubectl -n rook-ceph exec -it deploy/rook-ceph-tools -- ceph -s
Replace a failed metadata device¶
This section describes the scenario when an underlying metadata device fails with all related Ceph OSDs. In this case, the only solution is to remove all Ceph OSDs related to the failed metadata device, then attach a device that will be used as a new metadata device, and re-create all affected Ceph OSDs.
Caution
If you used BareMetalHostProfile to automatically partition
the failed device, you must create a manual partition of the new device
because BareMetalHostProfile does not support hot-load changes and
creates an automatic device partition only during node provisioning.
Save the
KaaSCephClusterspecification of all Ceph OSDs affected by the failed metadata device to re-use this specification during re-creation of Ceph OSDs after disk replacement.Identify Ceph OSD IDs related to the failed metadata device, for example, using Ceph CLI in the
rook-ceph-toolsPod:ceph osd metadata
Example of system response:
{ "id": 11, ... "bluefs_db_devices": "vdc", ... "bluestore_bdev_devices": "vde", ... "devices": "vdc,vde", ... "hostname": "kaas-node-6c5e76f9-c2d2-4b1a-b047-3c299913a4bf", ... }, { "id": 12, ... "bluefs_db_devices": "vdd", ... "bluestore_bdev_devices": "vde", ... "devices": "vdd,vde", ... "hostname": "kaas-node-6c5e76f9-c2d2-4b1a-b047-3c299913a4bf", ... }, { "id": 13, ... "bluefs_db_devices": "vdf", ... "bluestore_bdev_devices": "vde", ... "devices": "vde,vdf", ... "hostname": "kaas-node-6c5e76f9-c2d2-4b1a-b047-3c299913a4bf", ... }, ...
Open the
KaasCephClustercustom resource (CR) for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.In the
nodessection, remove allstorageDevicesitems that relate to the failed metadata device. For example:spec: cephClusterSpec: nodes: <machineName>: storageDevices: - name: <deviceName1> # remove the entire item from the storageDevices list # fullPath: <deviceByPath> if device is specified using symlink instead of name config: deviceClass: hdd metadataDevice: <metadataDevice> - name: <deviceName2> # remove the entire item from the storageDevices list config: deviceClass: hdd metadataDevice: <metadataDevice> - name: <deviceName3> # remove the entire item from the storageDevices list config: deviceClass: hdd metadataDevice: <metadataDevice> ...
In the example above,
<machineName>is the machine name of the node where the metadata device<metadataDevice>must be replaced.Create a
KaaSCephOperationRequestCR template and save it asreplace-failed-meta-<machineName>-<metadataDevice>-request.yaml:apiVersion: kaas.mirantis.com/v1alpha1 kind: KaaSCephOperationRequest metadata: name: replace-failed-meta-<machineName>-<metadataDevice> namespace: <managedClusterProjectName> spec: osdRemove: nodes: <machineName>: cleanupByOsdId: - <osdID-1> - <osdID-2> ... kaasCephCluster: name: <kaasCephClusterName> namespace: <managedClusterProjectName>
Substitute the following parameters:
<machineName>and<metadataDevice>with the machine and device names from the previous step<managedClusterProjectName>with the cluster project name<osdID-*>with IDs of the affected Ceph OSDs<kaasCephClusterName>with theKaaSCephClusterCR name<managedClusterProjectName>with the project name of the related managed cluster
Apply the template to the cluster:
kubectl apply -f replace-failed-meta-<machineName>-<metadataDevice>-request.yaml
Verify that the corresponding request has been created:
kubectl get kaascephoperationrequest -n <managedClusterProjectName>
Verify that the
removeInfosection is present in theKaaSCephOperationRequestCRstatusand that thecleanUpMapsection matches the required removal:kubectl -n <managedClusterProjectName> get kaascephoperationrequest replace-failed-meta-<machineName>-<metadataDevice> -o yaml
Example of system response:
childNodesMapping: <nodeName>: <machineName> removeInfo: cleanUpMap: <nodeName>: osdMapping: "<osdID-1>": deviceMapping: <dataDevice-1>: deviceClass: hdd devicePath: <dataDeviceByPath-1> devicePurpose: block usedPartition: <dataLvPartition-1> zapDisk: true <metadataDevice>: deviceClass: hdd devicePath: <metadataDeviceByPath> devicePurpose: db usedPartition: /dev/ceph-b0c70c72-8570-4c9d-93e9-51c3ab4dd9f9/osd-db-ecf64b20-1e07-42ac-a8ee-32ba3c0b7e2f uuid: ef516477-d2da-492f-8169-a3ebfc3417e2 "<osdID-2>": deviceMapping: <dataDevice-2>: deviceClass: hdd devicePath: <dataDeviceByPath-2> devicePurpose: block usedPartition: <dataLvPartition-2> zapDisk: true <metadataDevice>: deviceClass: hdd devicePath: <metadataDeviceByPath> devicePurpose: db usedPartition: /dev/ceph-b0c70c72-8570-4c9d-93e9-51c3ab4dd9f9/osd-db-ecf64b20-1e07-42ac-a8ee-32ba3c0b7e2f uuid: ef516477-d2da-492f-8169-a3ebfc3417e2 ...
Definition of values in angle brackets:
<machineName>- name of the machine on which the device is being replaced, for example,worker-1<nodeName>- underlying node name of the machine, for example,kaas-node-5a74b669-7e53-4535-aabd-5b509ec844af<osdId>- Ceph OSD ID for the device being replaced, for example,1<dataDeviceByPath>-by-pathof the device placed on the node, for example,/dev/disk/by-path/pci-0000:00:1t.9<dataDevice>- name of the device placed on the node, for example,/dev/vdc<metadataDevice>- metadata name of the device placed on the node, for example,/dev/vde<metadataDeviceByPath>- metadataby-pathof the device placed on the node, for example,/dev/disk/by-path/pci-0000:00:12.0<dataLvPartition>- logical volume partition of the data device
Wait for the
ApproveWaitingphase to appear instatus:kubectl -n <managedClusterProjectName> get kaascephoperationrequest replace-failed-meta-<machineName>-<metadataDevice> -o yaml
Example of system response:
status: phase: ApproveWaiting
In the
KaaSCephOperationRequestCR, set theapproveflag totrue:kubectl -n <managedClusterProjectName> edit kaascephoperationrequest replace-failed-meta-<machineName>-<metadataDevice>
Configuration snippet:
spec: osdRemove: approve: true
Review the following
statusfields of theKaaSCephOperationRequestCR request processing:status.phase- current state of request processingstatus.messages- description of the current phasestatus.conditions- full history of request processing before the current phasestatus.removeInfo.issuesandstatus.removeInfo.warnings- error and warning messages occurred during request processing, if any
Verify that the
KaaSCephOperationRequesthas been completed. For example:status: phase: Completed # or CompletedWithWarnings if there are non-critical issues
Note
This section describes how to create a metadata disk partition on N logical volumes. To create one partition on a metadata disk, refer to Reconfigure a partition of a Ceph OSD metadata device.
Partition the replaced metadata device by N logical volumes (LVs), where N is the number of Ceph OSDs previously located on a failed metadata device.
Calculate the new metadata LV percentage of used volume group capacity using the
100 / Nformula.Log in to the node with the replaced metadata disk.
Create an LVM physical volume atop the replaced metadata device:
pvcreate <metadataDisk>Substitute
<metadataDisk>with the replaced metadata device.Create an LVM volume group atop of the physical volume:
vgcreate bluedb <metadataDisk>
Substitute
<metadataDisk>with the replaced metadata device.Create N LVM logical volumes with the calculated capacity per each volume:
lvcreate -l <X>%VG -n meta_<i> bluedb
Substitute
<X>with the result of the100 / Nformula and<i>with the current number of metadata partitions.
As a result, the replaced metadata device will have N LVM paths, for example,
/dev/bluedb/meta_1.
Note
You can spawn Ceph OSD on a raw device, but it must be clean and without any data or partitions. If you want to add a device that was in use, also ensure it is raw and clean. To clean up all data and partitions from a device, refer to official Rook documentation.
Open the
KaasCephClusterCR for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.In the
nodessection, add the cleaned Ceph OSD device with the replaced LVM paths of the metadata device from previous steps. For example:spec: cephClusterSpec: nodes: <machineName>: storageDevices: - name: <deviceByID-1> # Recommended. Add the new device by ID /dev/disk/by-id/... #fullPath: <deviceByPath-1> # Add a new device by path /dev/disk/by-path/... config: deviceClass: hdd metadataDevice: /dev/<vgName>/<lvName-1> - name: <deviceByID-2> # Recommended. Add the new device by ID /dev/disk/by-id/... #fullPath: <deviceByPath-2> # Add a new device by path /dev/disk/by-path/... config: deviceClass: hdd metadataDevice: /dev/<vgName>/<lvName-2> - name: <deviceByID-3> # Recommended. Add the new device by ID /dev/disk/by-id/... #fullPath: <deviceByPath-3> # Add a new device by path /dev/disk/by-path/... config: deviceClass: hdd metadataDevice: /dev/<vgName>/<lvName-3>
Substitute
<machineName>with the machine name of the node where the metadata device has been replaced.Add all data devices for re-created Ceph OSDs and specify
metadataDevicethat is the path to the previously created logical volume. Substitute<vgName>with a volume group name that contains N logical volumes<lvName-i>.
Wait for the re-created Ceph OSDs to apply to the Ceph cluster.
You can monitor the application state using either the
statussection of theKaaSCephClusterCR or in therook-ceph-toolsPod:kubectl -n rook-ceph exec -it deploy/rook-ceph-tools -- ceph -s
Replace a failed Ceph node¶
After a physical node replacement, you can use the Ceph LCM API to redeploy failed Ceph nodes. The common flow of replacing a failed Ceph node is as follows:
Remove the obsolete Ceph node from the Ceph cluster.
Add a new Ceph node with the same configuration to the Ceph cluster.
Note
Ceph OSD node replacement presupposes usage of a
KaaSCephOperationRequest CR. For workflow overview, spec and phases
description, see High-level workflow of Ceph OSD or node removal.
Open the
KaasCephClusterCR of a managed cluster for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.In the
nodessection, remove the required device:spec: cephClusterSpec: nodes: <machineName>: # remove the entire entry for the node to replace storageDevices: {...} role: [...]
Substitute
<machineName>with the machine name to replace.Save
KaaSCephClusterand close the editor.Create a
KaaSCephOperationRequestCR template and save it asreplace-failed-<machineName>-request.yaml:apiVersion: kaas.mirantis.com/v1alpha1 kind: KaaSCephOperationRequest metadata: name: replace-failed-<machineName>-request namespace: <managedClusterProjectName> spec: osdRemove: nodes: <machineName>: completeCleanUp: true kaasCephCluster: name: <kaasCephClusterName> namespace: <managedClusterProjectName>
Substitute
<kaasCephClusterName>with the correspondingKaaSCephClusterresource from the<managedClusterProjectName>namespace.Apply the template to the cluster:
kubectl apply -f replace-failed-<machineName>-request.yaml
Verify that the corresponding request has been created:
kubectl get kaascephoperationrequest -n <managedClusterProjectName>
Verify that the
removeInfosection appeared in theKaaSCephOperationRequestCRstatus:kubectl -n <managedClusterProjectName> get kaascephoperationrequest replace-failed-<machineName>-request -o yaml
Example of system response:
status: childNodesMapping: <nodeName>: <machineName> osdRemoveStatus: removeInfo: cleanUpMap: <nodeName>: osdMapping: ... <osdId>: deviceMapping: ... <deviceName>: path: <deviceByPath> partition: "/dev/ceph-b-vg_sdb/osd-block-b-lv_sdb" type: "block" class: "hdd" zapDisk: true
If needed, change the following values:
<machineName>- machine name where the replacement occurs, for example,worker-1.<nodeName>- underlying machine node name, for example,kaas-node-5a74b669-7e53-4535-aabd-5b509ec844af.<osdId>- actual Ceph OSD ID for the device being replaced, for example,1.<deviceName>- actual device name placed on the node, for example,sdb.<deviceByPath>- actual deviceby-pathplaced on the node, for example,/dev/disk/by-path/pci-0000:00:1t.9.
Verify that the
cleanUpMapsection matches the required removal and wait for theApproveWaitingphase to appear instatus:kubectl -n <managedClusterProjectName> get kaascephoperationrequest replace-failed-<machineName>-request -o yaml
Example of system response:
status: phase: ApproveWaiting
Edit the
KaaSCephOperationRequestCR and set theapproveflag totrue:kubectl -n <managedClusterProjectName> edit kaascephoperationrequest replace-failed-<machineName>-request
For example:
spec: osdRemove: approve: true
Review the following
statusfields of theKaaSCephOperationRequestCR request processing:status.phase- current state of request processingstatus.messages- description of the current phasestatus.conditions- full history of request processing before the current phasestatus.removeInfo.issuesandstatus.removeInfo.warnings- error and warning messages occurred during request processing, if any
Verify that the
KaaSCephOperationRequesthas been completed. For example:status: phase: Completed # or CompletedWithWarnings if there are non-critical issues
Remove the device cleanup jobs:
kubectl delete jobs -n ceph-lcm-mirantis -l app=miraceph-cleanup-disks
Note
You can spawn Ceph OSD on a raw device, but it must be clean and without any data or partitions. If you want to add a device that was in use, also ensure it is raw and clean. To clean up all data and partitions from a device, refer to official Rook documentation.
Open the
KaasCephClusterCR of a managed cluster for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.In the
nodessection, add a new device:spec: cephClusterSpec: nodes: <machineName>: # add new configuration for replaced Ceph node storageDevices: - fullPath: <deviceByID> # Recommended since MCC 2.25.0 (17.0.0), non-wwn by-id symlink # name: <deviceByID> # Prior MCC 2.25.0, non-wwn by-id symlink # fullPath: <deviceByPath> # if device is supposed to be added with by-path config: deviceClass: hdd ...
Substitute
<machineName>with the machine name of the replaced node and configure it as required.Warning
Since MCC 2.25.0 (17.0.0), Mirantis highly recommends using non-wwn
by-idsymlinks only to specify storage devices in thestorageDeviceslist.For details, see Container Cloud documentation: Addressing storage devices.
Verify that all Ceph daemons from the replaced node have appeared on the Ceph cluster and are
inandup. ThefullClusterInfosection should not contain any issues.kubectl -n <managedClusterProjectName> get kaascephcluster -o yaml
Example of system response:
status: fullClusterInfo: clusterStatus: ceph: health: HEALTH_OK ... daemonStatus: mgr: running: a is active mgr status: Ok mon: running: '3/3 mons running: [a b c] in quorum' status: Ok osd: running: '3/3 running: 3 up, 3 in' status: Ok
Verify the Ceph node on the managed cluster:
kubectl -n rook-ceph get pod -o wide | grep <machineName>
Remove Ceph OSD manually¶
You may need to manually remove a Ceph OSD, for example, in the following cases:
If you have removed a device or node from the
KaaSCephClusterspec.cephClusterSpec.nodesorspec.cephClusterSpec.nodeGroupssection withmanageOsdsset tofalse.If you do not want to rely on Ceph LCM operations and want to manage the Ceph OSDs life cycle manually.
To safely remove one or multiple Ceph OSDs from a Ceph cluster, perform the following procedure for each Ceph OSD one by one.
Warning
The procedure presupposes the Ceph OSD disk or logical volumes partition cleanup.
To remove a Ceph OSD manually:
Edit the
KaaSCephClusterresource on a management cluster:kubectl --kubeconfig <mgmtKubeconfig> -n <managedClusterProjectName> edit kaascephcluster
Substitute
<mgmtKubeconfig>with the management clusterkubeconfigand<managedClusterProjectName>with the project name of the managed cluster.In the
spec.cephClusterSpec.nodessection, remove the requiredstorageDevicesitem of the corresponding node spec. If after removalstorageDevicesbecomes empty and the node spec has no roles specified, also remove the node spec.Obtain
kubeconfigof the managed cluster and provide it as an environment variable:export KUBECONFIG=<pathToManagedKubeconfig>
Verify that all Ceph OSDs are
upandin, the Ceph cluster is healthy, and no rebalance or recovery is in progress:kubectl -n rook-ceph exec -it $(kubectl -n rook-ceph get pod -l \ app=rook-ceph-tools -o jsonpath='{.items[0].metadata.name}') -- ceph -s
Example of system response:
cluster: id: 8cff5307-e15e-4f3d-96d5-39d3b90423e4 health: HEALTH_OK ... osd: 4 osds: 4 up (since 10h), 4 in (since 10h)
Stop the
rook-ceph/rook-ceph-operatordeployment to avoid premature reorchestration of the Ceph cluster:kubectl -n rook-ceph scale deploy rook-ceph-operator --replicas 0
Enter the
ceph-toolspod:kubectl -n rook-ceph exec -it $(kubectl -n rook-ceph get pod -l \ app=rook-ceph-tools -o jsonpath='{.items[0].metadata.name}') bash
Mark the required Ceph OSD as
out:ceph osd out osd.<ID>
Note
In the command above and in the steps below, substitute
<ID>with the number of the Ceph OSD to remove.Wait until data backfilling to other OSDs is complete:
ceph -sOnce all of the PGs are
active+clean, backfilling is complete and it is safe to remove the disk.Note
For additional information on PGs backfilling, run ceph pg dump_stuck.
Exit from the
ceph-toolspod:exitScale the
rook-ceph/rook-ceph-osd-<ID>deployment to0replicas:kubectl -n rook-ceph scale deploy rook-ceph-osd-<ID> --replicas 0
Enter the
ceph-toolspod:kubectl -n rook-ceph exec -it $(kubectl -n rook-ceph get pod -l \ app=rook-ceph-tools -o jsonpath='{.items[0].metadata.name}') bash
Verify that the number of Ceph OSDs that are
upandinhas decreased by one daemon:ceph -sExample of system response:
osd: 4 osds: 3 up (since 1h), 3 in (since 5s)
Remove the Ceph OSD from the Ceph cluster:
ceph osd purge <ID> --yes-i-really-mean-it
Delete the Ceph OSD
authentry, if present. Otherwise, skip this step.ceph auth del osd.<ID>
If you have removed the last Ceph OSD on the node and want to remove this node from the Ceph cluster, remove the CRUSH map entry:
ceph osd crush remove <nodeName>
Substitute
<nodeName>with the name of the node where the removed Ceph OSD was placed.Verify that the failure domain within Ceph OSDs has been removed from the CRUSH map:
ceph osd tree
If you have removed the node, it will be removed from the CRUSH map.
Exit from the
ceph-toolspod:exitClean up the disk used by the removed Ceph OSD. For details, see official Rook documentation.
Warning
If you are using multiple Ceph OSDs per device or metadata device, make sure that you can clean up the entire disk. Otherwise, instead clean up only the logical volume partitions for the volume group by running lvremove <lvpartion_uuid> any Ceph OSD pod that belongs to the same host as the removed Ceph OSD.
Delete the
rook-ceph/rook-ceph-osd-<ID>deployment previously scaled to0replicas:kubectl -n rook-ceph delete deploy rook-ceph-osd-<ID>
Substitute
<ID>with the number of the removed Ceph OSD.Scale the
rook-ceph/rook-ceph-operatordeployment to1replica and wait for the orchestration to complete:kubectl -n rook-ceph scale deploy rook-ceph-operator --replicas 1 kubectl -n rook-ceph get pod -w
Once done, Ceph OSD removal is complete.
Migrate Ceph cluster to address storage devices using by-id¶
The by-id identifier is the only persistent device identifier for a Ceph
cluster that remains stable after the cluster upgrade or any other maintenance.
Therefore, Mirantis recommends using device by-id symlinks rather than
device names or by-path symlinks.
Container Cloud uses the device by-id identifier as the default method
of addressing the underlying devices of Ceph OSDs. Thus, you should migrate
all existing Ceph clusters, which are still utilizing the device names or
device by-path symlinks, to the by-id format.
This section explains how to configure the KaaSCephCluster specification
to use the by-id symlinks instead of disk names and by-path
identifiers as the default method of addressing storage devices.
Note
Mirantis recommends avoiding the use of wwn symlinks as
by-id identifiers due to their lack of persistence expressed in
inconsistent discovery during node boot.
Besides migrating to by-id, consider using the fullPath field for the
by-id symlinks configuration, instead of the name field in the
spec.cephClusterSpec.nodes.storageDevices section. This approach allows for
clear understanding of field namings and their use cases.
Note
MOSK enables you to use fullPath for the
by-id symlinks since MCC 2.25.0 (Cluster release 17.0.0). For earlier
product versions, use the name field instead.
Migrate the Ceph nodes section to by-id identifiers¶
Available since MCC 2.25.0 (Cluster release 17.0.0)
Make sure that your managed cluster is not currently running an upgrade or any other maintenance process.
Obtain the list of all
KaasCephClusterstorage devices that use disk names or diskby-pathas identifiers of Ceph node storage devices:kubectl -n <managedClusterProject> get kcc -o yaml
Substitute
<managedClusterProject>with the corresponding managed cluster namespace.Output example:
spec: cephClusterSpec: nodes: ... managed-worker-1: storageDevices: - config: deviceClass: hdd name: sdc - config: deviceClass: hdd fullPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:2 managed-worker-2: storageDevices: - config: deviceClass: hdd name: /dev/disk/by-id/wwn-0x26d546263bd312b8 - config: deviceClass: hdd name: /dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_2e52abb48862dsdc managed-worker-3: storageDevices: - config: deviceClass: nvme name: nvme3n1 - config: deviceClass: hdd fullPath: /dev/disk/by-id/scsi-SATA_HGST_HUS724040AL_PN1334PEHN18ZS
Verify the items from the
storageDevicessections to be moved to theby-idsymlinks. The list of the items to migrate includes:A disk name in the
namefield. For example,sdc,nvme3n1, and so on.A disk
/dev/disk/by-pathsymlink in thefullPathfield. For example,/dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:2.A disk
/dev/disk/by-idsymlink in thenamefield.Note
This condition applies since MCC 2.25.0 (Cluster release 17.0.0).
A disk
/dev/disk/by-id/wwnsymlink, which is programmatically calculated at boot. For example,/dev/disk/by-id/wwn-0x26d546263bd312b8.
For the example above, we have to migrate both items of
managed-worker-1, both items ofmanaged-worker-2, and the first item ofmanaged-worker-3. The second item ofmanaged-worker-3has already been configured in the required format, therefore, we are leaving it as is.To migrate all affected
storageDevicesitems toby-idsymlinks, open theKaaSCephClustercustom resource for editing:kubectl -n <managedClusterProject> edit kccFor each affected node from the
spec.cephClusterSpec.nodessection, obtain a correspondingstatus.providerStatus.hardware.storagesection from theMachinecustom resource:kubectl -n <managedClusterProject> get machine <machineName> -o yaml
Substitute
<managedClusterProject>with the corresponding cluster namespace and<machineName>with the machine name.Output example for
managed-worker-1:status: providerStatus: hardware: storage: - byID: /dev/disk/by-id/wwn-0x05ad99618d66a21f byIDs: - /dev/disk/by-id/scsi-0QEMU_QEMU_HARDDISK_05ad99618d66a21f - /dev/disk/by-id/scsi-305ad99618d66a21f - /dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_05ad99618d66a21f - /dev/disk/by-id/wwn-0x05ad99618d66a21f byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:0 byPaths: - /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:0 name: /dev/sda serialNumber: 05ad99618d66a21f size: 61 type: hdd - byID: /dev/disk/by-id/wwn-0x26d546263bd312b8 byIDs: - /dev/disk/by-id/scsi-0QEMU_QEMU_HARDDISK_26d546263bd312b8 - /dev/disk/by-id/scsi-326d546263bd312b8 - /dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_26d546263bd312b8 - /dev/disk/by-id/wwn-0x26d546263bd312b8 byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:2 byPaths: - /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:2 name: /dev/sdb serialNumber: 26d546263bd312b8 size: 32 type: hdd - byID: /dev/disk/by-id/wwn-0x2e52abb48862dbdc byIDs: - /dev/disk/by-id/lvm-pv-uuid-MncrcO-6cel-0QsB-IKaY-e8UK-6gDy-k2hOtf - /dev/disk/by-id/scsi-0QEMU_QEMU_HARDDISK_2e52abb48862dbdc - /dev/disk/by-id/scsi-32e52abb48862dbdc - /dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_2e52abb48862dbdc - /dev/disk/by-id/wwn-0x2e52abb48862dbdc byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:1 byPaths: - /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:1 name: /dev/sdc serialNumber: 2e52abb48862dbdc size: 61 type: hdd
For each affected
storageDevicesitem from the consideredMachine, obtain a correctby-idsymlink fromstatus.providerStatus.hardware.storage.byIDs. Suchby-idsymlink must containstatus.providerStatus.hardware.storage.serialNumberand must not containwwn.For
managed-worker-1, according to the example output above, we can use the followingby-idsymlinks:Replace the first item of
storageDevicesthat containsname: sdcwithfullPath: /dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_2e52abb48862dbdc;Replace the second item of
storageDevicesthat containsfullPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:2withfullPath: /dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_26d546263bd312b8.
Replace all affected
storageDevicesitems inKaaSCephClusterwith the obtained ones.Note
Prior to MCC 2.25.0 (Cluster release 17.0.0), place the
by-idsymlinks in thenamefield instead of thefullPathfield.The resulting example of the storage device identifier migration:
spec: cephClusterSpec: nodes: ... managed-worker-1: storageDevices: - config: deviceClass: hdd fullPath: /dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_2e52abb48862dbdc - config: deviceClass: hdd fullPath: /dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_26d546263bd312b8 managed-worker-2: storageDevices: - config: deviceClass: hdd fullPath: /dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_031d9054c9b48f79 - config: deviceClass: hdd fullPath: /dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_2e52abb48862dsdc managed-worker-3: storageDevices: - config: deviceClass: nvme fullPath: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB3T8HMLA-00007_S46FNY0R394543 - config: deviceClass: hdd fullPath: /dev/disk/by-id/scsi-SATA_HGST_HUS724040AL_PN1334PEHN18ZS
Save and quit editing the
KaaSCephClustercustom resource.
After migration, the re-orchestration occurs. The whole procedure should not result in any real changes to the Ceph cluster state in Ceph OSDs.
Migrate the Ceph nodeGroups section to by-id identifiers¶
Available since MCC 2.25.0 (Cluster release 17.0.0)
Besides the nodes section, your cluster may contain the nodeGroups
section specified with disk names instead of by-id symlinks. Despite
of inplace replacement of the nodes storage device identifiers,
nodeGroups requires another approach because of the repeatable spec
section for different nodes.
In the case of migrating nodeGroups storage devices, the deviceLabels
section should be used to label different disks with the same labels and use
these labels in node groups after. For the deviceLabels section
specification, refer to Ceph advanced configuration: extraOpts.
The following procedure describes how to keep the nodeGroups section
but use unique by-id identifiers instead of disk names.
To migrate the Ceph nodeGroups section to by-id identifiers:
Make sure that your managed cluster is not currently running an upgrade or any other maintenance process.
Obtain the list of all
KaasCephClusterstorage devices that use disk names or diskby-pathas identifiers of Ceph node group storage devices:kubectl -n <managedClusterProject> get kcc -o yaml
Substitute
<managedClusterProject>with the corresponding managed cluster namespace.Output example of the
KaaSCephClusternodeGroupssection with disk names used as identifiers:spec: cephClusterSpec: nodeGroups: ... rack-1: nodes: - node-1 - node-2 spec: crush: rack: "rack-1" storageDevices: - name: nvme0n1 config: deviceClass: nvme - name: nvme1n1 config: deviceClass: nvme - name: nvme2n1 config: deviceClass: nvme rack-2: nodes: - node-3 - node-4 spec: crush: rack: "rack-2" storageDevices: - name: nvme0n1 config: deviceClass: nvme - name: nvme1n1 config: deviceClass: nvme - name: nvme2n1 config: deviceClass: nvme rack-3: nodes: - node-5 - node-6 spec: crush: rack: "rack-3" storageDevices: - name: nvme0n1 config: deviceClass: nvme - name: nvme1n1 config: deviceClass: nvme - name: nvme2n1 config: deviceClass: nvme
Verify the items from the
storageDevicessections to be moved toby-idsymlinks. The list of the items to migrate includes:A disk name in the
namefield. For example,sdc,nvme3n1, and so on.A disk
/dev/disk/by-pathsymlink in thefullPathfield. For example,/dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:2.A disk
/dev/disk/by-idsymlink in thenamefield.Note
This condition applies since MCC 2.25.0 (Cluster release 17.0.0).
A disk
/dev/disk/by-id/wwnsymlink, which is programmatically calculated at boot. For example,/dev/disk/by-id/wwn-0x26d546263bd312b8.
All
storageDevicesections in the example above contain disk names in thenamefield. Therefore, you need to replace them withby-idsymlinks.Open the
KaaSCephClustercustom resource for editing to start migration of all affectedstorageDevicesitems toby-idsymlinks:kubectl -n <managedClusterProject> edit kccWithin each impacted Ceph node group in the
nodeGroupssection, add disk labels to thedeviceLabelssections for every affected storage device linked with the nodes listed innodesof that specific node group. Verify that these disk labels are equal toby-idsymlinks of corresponding disks.For example, if the node group
rack-1contains two nodesnode-1andnode-2andspeccontains three items withname, you need to obtain properby-idsymlinks for disk names from both nodes and write it down with the same disk labels. The following example contains the labels forby-idsymlinks ofnvme0n1,nvme1n1, andnvme2n1disks fromnode-1andnode-2correspondingly:spec: cephClusterSpec: extraOpts: deviceLabels: node-1: nvme-1: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB3T8HMLA-00007_S46FNY0R394543 nvme-2: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB3T8HMLA-00007_S46FNY0R372150 nvme-3: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB3T8HMLA-00007_S46FNY0R183266 node-2: nvme-1: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB4040ALR-00007_S46FNY0R900128 nvme-2: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB4040ALR-00007_S46FNY0R805840 nvme-3: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB4040ALR-00007_S46FNY0R848469
Note
Keep device labels repeatable for all nodes from the node group. This allows for specifying unified
specfor differentby-idsymlinks of different nodes.Example of the full
deviceLabelssection for thenodeGroupssection:spec: cephClusterSpec: extraOpts: deviceLabels: node-1: nvme-1: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB3T8HMLA-00007_S46FNY0R394543 nvme-2: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB3T8HMLA-00007_S46FNY0R372150 nvme-3: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB3T8HMLA-00007_S46FNY0R183266 node-2: nvme-1: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB4040ALR-00007_S46FNY0R900128 nvme-2: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB4040ALR-00007_S46FNY0R805840 nvme-3: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB4040ALR-00007_S46FNY0R848469 node-3: nvme-1: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB00T2B0A-00007_S46FNY0R900128 nvme-2: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB00T2B0A-00007_S46FNY0R805840 nvme-3: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB00T2B0A-00007_S46FNY0R848469 node-4: nvme-1: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB00Z4SA0-00007_S46FNY0R286212 nvme-2: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB00Z4SA0-00007_S46FNY0R350024 nvme-3: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB00Z4SA0-00007_S46FNY0R300756 node-5: nvme-1: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB8UK0QBD-00007_S46FNY0R577024 nvme-2: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB8UK0QBD-00007_S46FNY0R718411 nvme-3: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB8UK0QBD-00007_S46FNY0R831424 node-6: nvme-1: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB01DAU34-00007_S46FNY0R908440 nvme-2: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB01DAU34-00007_S46FNY0R945405 nvme-3: /dev/disk/by-id/nvme-SAMSUNG_MZ1LB01DAU34-00007_S46FNY0R224911
For each affected node group in the
nodeGroupssection, replace the field with the insufficient disk identifier to thedevLabelfield with the disk label from thedeviceLabelssection.For the example above, the updated
nodeGroupssection looks as follows:spec: cephClusterSpec: nodeGroups: ... rack-1: nodes: - node-1 - node-2 spec: crush: rack: "rack-1" storageDevices: - devLabel: nvme-1 config: deviceClass: nvme - devLabel: nvme-2 config: deviceClass: nvme - devLabel: nvme-3 config: deviceClass: nvme rack-2: nodes: - node-3 - node-4 spec: crush: rack: "rack-2" storageDevices: - devLabel: nvme-1 config: deviceClass: nvme - devLabel: nvme-2 config: deviceClass: nvme - devLabel: nvme-3 config: deviceClass: nvme rack-3: nodes: - node-5 - node-6 spec: crush: rack: "rack-3" storageDevices: - devLabel: nvme-1 config: deviceClass: nvme - devLabel: nvme-2 config: deviceClass: nvme - devLabel: nvme-3 config: deviceClass: nvme
Save and quit editing the
KaaSCephClustercustom resource.
After migration, the re-orchestration occurs. The whole procedure should not result in any real changes to the Ceph cluster state in Ceph OSDs.
Obtain a by-id symlink of a storage device¶
You can start using a storage device only after a corresponding Machine
becomes ready and accessible. Thus, KaaSCephCluster can be created only
after all machines receive the status.providerStatus.hardware.storage
configuration containing all required device by-id symlinks.
To obtain a device by-id symlink:
Verify that the
MachineisReady:kubectl -n <managedClusterProject> get machine <machineName> -o jsonpath='{.status.phase}{"\n"}'
Substitute
<managedClusterProject>with the cluster namespace and<machineName>with the machine name.Obtain storage details for the
Machine:kubectl -n <managedClusterProject> get machine <machineName> -o yaml
Output example:
status: providerStatus: hardware: storage: - byID: /dev/disk/by-id/wwn-0x05ad99618d66a21f byIDs: - /dev/disk/by-id/scsi-0QEMU_QEMU_HARDDISK_05ad99618d66a21f - /dev/disk/by-id/scsi-305ad99618d66a21f - /dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_05ad99618d66a21f - /dev/disk/by-id/wwn-0x05ad99618d66a21f byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:0 byPaths: - /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:0 name: /dev/sda serialNumber: 05ad99618d66a21f size: 61 type: hdd - byID: /dev/disk/by-id/wwn-0x26d546263bd312b8 byIDs: - /dev/disk/by-id/scsi-0QEMU_QEMU_HARDDISK_26d546263bd312b8 - /dev/disk/by-id/scsi-326d546263bd312b8 - /dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_26d546263bd312b8 - /dev/disk/by-id/wwn-0x26d546263bd312b8 byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:2 byPaths: - /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:2 name: /dev/sdb serialNumber: 26d546263bd312b8 size: 32 type: hdd - byID: /dev/disk/by-id/wwn-0x2e52abb48862dbdc byIDs: - /dev/disk/by-id/lvm-pv-uuid-MncrcO-6cel-0QsB-IKaY-e8UK-6gDy-k2hOtf - /dev/disk/by-id/scsi-0QEMU_QEMU_HARDDISK_2e52abb48862dbdc - /dev/disk/by-id/scsi-32e52abb48862dbdc - /dev/disk/by-id/scsi-SQEMU_QEMU_HARDDISK_2e52abb48862dbdc - /dev/disk/by-id/wwn-0x2e52abb48862dbdc byPath: /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:1 byPaths: - /dev/disk/by-path/pci-0000:00:05.0-scsi-0:0:0:1 name: /dev/sdc serialNumber: 2e52abb48862dbdc size: 61 type: hdd
Obtain the item from the
byIDslist from thestatus.providerStatus.hardware.storagesection that containsserialNumberand does not containwwnas a bus ID.In the example above, for the disk with the
/dev/sdcname, you can use the/dev/disk/by-id/scsi-0QEMU_QEMU_HARDDISK_2e52abb48862dbdcsymlink as a persistent identifier of the storage device because it contains the2e52abb48862dbdcserial number and does not containwwn.Note
Do not rely on the
byIDfield only. This field may contain a/dev/disk/by-id/wwnsymlink that cannot be considered a persistent identifier of a storage device.
Increase Ceph cluster storage size¶
This section describes how to increase the overall storage size for all Ceph
pools of the same device class: hdd, ssd, or nvme.
The procedure presupposes adding a new Ceph OSD. The overall storage size for
the required device class automatically increases once the Ceph OSD becomes
available in the Ceph cluster.
To increase the overall storage size for a device class:
Identify the current storage size for the required device class:
kubectl --kubeconfig <managedClusterKubeconfig> -n rook-ceph exec -it deploy/rook-ceph-tools -- ceph df
Substitute
<managedClusterKubeconfig>with a managed clusterkubeconfig.Example of system response:
--- RAW STORAGE --- CLASS SIZE AVAIL USED RAW USED %RAW USED hdd 128 GiB 101 GiB 23 GiB 27 GiB 21.40 TOTAL 128 GiB 101 GiB 23 GiB 27 GiB 21.40 --- POOLS --- POOL ID PGS STORED OBJECTS USED %USED MAX AVAIL device_health_metrics 1 1 0 B 0 0 B 0 30 GiB kubernetes-hdd 2 32 12 GiB 3.13k 23 GiB 20.57 45 GiB
Identify the number of Ceph OSDs with the required device class:
kubectl --kubeconfig <managedClusterKubeconfig> -n rook-ceph exec -it deploy/rook-ceph-tools -- ceph osd df <deviceClass>
Substitute the following parameters:
<managedClusterKubeconfig>with a managed clusterkubeconfig<deviceClass>with the required device class:hdd,ssd, ornvme
Example of system response for the
hdddevice class:ID CLASS WEIGHT REWEIGHT SIZE RAW USE DATA OMAP META AVAIL %USE VAR PGS STATUS 1 hdd 0.03119 1.00000 32 GiB 5.8 GiB 4.8 GiB 1.5 MiB 1023 MiB 26 GiB 18.22 0.85 14 up 3 hdd 0.03119 1.00000 32 GiB 6.9 GiB 5.9 GiB 1.1 MiB 1023 MiB 25 GiB 21.64 1.01 17 up 0 hdd 0.03119 0.84999 32 GiB 6.8 GiB 5.8 GiB 1013 KiB 1023 MiB 25 GiB 21.24 0.99 16 up 2 hdd 0.03119 1.00000 32 GiB 7.9 GiB 6.9 GiB 1.2 MiB 1023 MiB 24 GiB 24.55 1.15 20 up TOTAL 128 GiB 27 GiB 23 GiB 4.8 MiB 4.0 GiB 101 GiB 21.41 MIN/MAX VAR: 0.85/1.15 STDDEV: 2.29
Follow Add a Ceph OSD on a managed cluster to add a new device with a supported device class:
hdd,ssd, ornvme.Wait for the new Ceph OSD pod to start
Running:kubectl --kubeconfig <managedClusterKubeconfig> -n rook-ceph get pod -l app=rook-ceph-osd
Substitute
<managedClusterKubeconfig>with a managed clusterkubeconfig.Verify that the new Ceph OSD has rebalanced and Ceph health is
HEALTH_OK:kubectl --kubeconfig <managedClusterKubeconfig> -n rook-ceph exec -it deploy/rook-ceph-tools -- ceph -s
Substitute
<managedClusterKubeconfig>with a managed cluster kubeconfig.Verify that the new Ceph has been OSD added to the list of device class OSDs:
kubectl --kubeconfig <managedClusterKubeconfig> -n rook-ceph exec -it deploy/rook-ceph-tools -- ceph osd df <deviceClass>
Substitute the following parameters:
<managedClusterKubeconfig>with a managed clusterkubeconfig<deviceClass>with the required device class:hdd,ssd, ornvme
Example of system response for the
hdddevice class after adding a new Ceph OSD:ID CLASS WEIGHT REWEIGHT SIZE RAW USE DATA OMAP META AVAIL %USE VAR PGS STATUS 1 hdd 0.03119 1.00000 32 GiB 4.5 GiB 3.5 GiB 1.5 MiB 1023 MiB 28 GiB 13.93 0.78 10 up 3 hdd 0.03119 1.00000 32 GiB 5.5 GiB 4.5 GiB 1.1 MiB 1023 MiB 26 GiB 17.22 0.96 13 up 0 hdd 0.03119 0.84999 32 GiB 6.5 GiB 5.5 GiB 1013 KiB 1023 MiB 25 GiB 20.32 1.14 15 up 2 hdd 0.03119 1.00000 32 GiB 7.5 GiB 6.5 GiB 1.2 MiB 1023 MiB 24 GiB 23.43 1.31 19 up 4 hdd 0.03119 1.00000 32 GiB 4.6 GiB 3.6 GiB 0 B 1 GiB 27 GiB 14.45 0.81 10 up TOTAL 160 GiB 29 GiB 24 GiB 4.8 MiB 5.0 GiB 131 GiB 17.87 MIN/MAX VAR: 0.78/1.31 STDDEV: 3.62
Verify the total storage capacity increased for the entire device class:
kubectl --kubeconfig <managedClusterKubeconfig> -n rook-ceph exec -it deploy/rook-ceph-tools -- ceph df
Substitute
<managedClusterKubeconfig>with a managed clusterkubeconfig.Example of system response:
--- RAW STORAGE --- CLASS SIZE AVAIL USED RAW USED %RAW USED hdd 160 GiB 131 GiB 24 GiB 29 GiB 17.97 TOTAL 160 GiB 131 GiB 24 GiB 29 GiB 17.97 --- POOLS --- POOL ID PGS STORED OBJECTS USED %USED MAX AVAIL device_health_metrics 1 1 0 B 0 0 B 0 38 GiB kubernetes-hdd 2 32 12 GiB 3.18k 24 GiB 17.17 57 GiB
Move a Ceph Monitor daemon to another node¶
This document describes how to migrate a Ceph Monitor daemon from one node to another without changing the general number of Ceph Monitors in the cluster. In the Ceph Controller concept, migration of a Ceph Monitor means manually removing it from one node and adding it to another.
Consider the following exemplary placement scheme of Ceph Monitors in the
nodes spec of the KaaSCephCluster CR:
nodes:
node-1:
roles:
- mon
- mgr
node-2:
roles:
- mgr
Using the example above, if you want to move the Ceph Monitor from node-1
to node-2 without changing the number of Ceph Monitors, the roles table
of the nodes spec must result as follows:
nodes:
node-1:
roles:
- mgr
node-2:
roles:
- mgr
- mon
However, due to the Rook limitation related to Kubernetes architecture, once
you move the Ceph Monitor through the KaaSCephCluster CR, changes will not
apply automatically. This is caused by the following Rook behavior:
Rook creates Ceph Monitor resources as deployments with
nodeSelector, which binds Ceph Monitor pods to a requested node.Rook does not recreate new Ceph Monitors with the new node placement if the current
monquorum works.
Therefore, to move a Ceph Monitor to another node, you must also manually apply the new Ceph Monitors placement to the Ceph cluster as described below.
To move a Ceph Monitor to another node:
Open the
KaasCephClusterCR of a managed cluster:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with the corresponding value.In the
nodesspec of theKaaSCephClusterCR, change themonroles placement without changing the total number ofmonroles. For details, see the example above. Note the nodes on which themonroles have been removed.Wait until the corresponding
MiraCephresource is updated with the newnodesspec:kubectl --kubeconfig <kubeconfig> -n ceph-lcm-mirantis get miraceph -o yaml
Substitute
<kubeconfig>with the Container Cloud clusterkubeconfigthat hosts the required Ceph cluster.In the
MiraCephresource, determine which node has been changed in thenodesspec. Save thenamevalue of the node where themonrole has been removed for further usage.kubectl -n <managedClusterProjectName> get machine -o jsonpath='{range .items[*]}{.metadata.name .status.nodeRef.name}{"\n"}{end}'
Substitute
<managedClusterProjectName>with the corresponding value.If you perform a managed cluster update, follow additional steps:
Verify that the following conditions are met before proceeding to the next step:
There are at least 2 running and available Ceph Monitors so that the Ceph cluster is accessible during the Ceph Monitor migration:
kubectl -n rook-ceph get pod -l app=rook-ceph-mon kubectl -n rook-ceph exec -it deploy/rook-ceph-tools -- ceph -s
The
MiraCephobject on the managed cluster has the required node with themonrole added in thenodessection ofspec:kubectl -n ceph-lcm-mirantis get miraceph -o yaml
The Ceph
NodeWorkloadLockfor the required node is created:kubectl --kubeconfig child-kubeconfig get nodeworkloadlock -o jsonpath='{range .items[?(@.spec.nodeName == "<desiredNodeName>")]}{@.metadata.name}{"\n"}{end}' | grep ceph
Scale the
ceph-maintenance-controllerdeployment to0replicas:kubectl -n ceph-lcm-mirantis scale deploy ceph-maintenance-controller --replicas 0
Manually edit the managed cluster node labels: remove the
ceph_role_monlabel from the obsolete node and add this label to the new node:kubectl label node <obsoleteNodeName> ceph_role_mon- kubectl label node <newNodeName> ceph_role_mon=true
Verify that the
rook-ceph-operatordeployment is scaled to0replica:kubectl -n rook-ceph scale deploy rook-ceph-operator --replicas 0
Obtain the
rook-ceph-mondeployment name placed on the obsolete node using the previously obtained node name:kubectl -n rook-ceph get deploy -l app=rook-ceph-mon -o jsonpath="{.items[?(@.spec.template.spec.nodeSelector['kubernetes\.io/hostname'] == '<nodeName>')].metadata.name}"
Substitute
<nodeName>with the name of the node where you removed themonrole.Back up the
rook-ceph-mondeployment placed on the obsolete node:kubectl -n rook-ceph get deploy <rook-ceph-mon-name> -o yaml > <rook-ceph-mon-name>-backup.yaml
Remove the
rook-ceph-mondeployment placed on the obsolete node:kubectl -n rook-ceph delete deploy <rook-ceph-mon-name>
If you perform a managed cluster update, follow additional steps:
Enter the
ceph-toolspod:kubectl -n rook-ceph exec -it deploy/rook-ceph-tools -- bash
Remove the Ceph Monitor from the Ceph monmap by letter:
ceph mon rm <monLetter>
Substitute
<monLetter>with the old Ceph Monitor letter. For example,mon-bhas the letterb.Verify that the Ceph cluster does not have any information about the the removed Ceph Monitor:
ceph mon dump ceph -s
Exit the
ceph-toolspod.Scale up the
rook-ceph-operatordeployment to1replica:kubectl -n rook-ceph scale deploy rook-ceph-operator --replicas 1
Wait for the missing Ceph Monitor failover process to start:
kubectl -n rook-ceph logs -l app=rook-ceph-operator -f
Example of log extract:
2024-03-01 12:33:08.741215 W | op-mon: mon b NOT found in ceph mon map, failover 2024-03-01 12:33:08.741244 I | op-mon: marking mon "b" out of quorum ... 2024-03-01 12:33:08.766822 I | op-mon: Failing over monitor "b" 2024-03-01 12:33:08.766881 I | op-mon: starting new mon...
Select one of the following options:
Wait approximately 10 minutes until
rook-ceph-operatorperforms a failover of thePendingmonpod. Inspect the logs during the failover process:kubectl -n rook-ceph logs -l app=rook-ceph-operator -f
Example of log extract:
2021-03-15 17:48:23.471978 W | op-mon: mon "a" not found in quorum, waiting for timeout (554 seconds left) before failover
Note
If the failover process fails:
Scale down the
rook-ceph-operatordeployment to0replicas.Apply the backed-up
rook-ceph-mondeployment.Scale back the
rook-ceph-operatordeployment to1replica.
Scale the
rook-ceph-operatordeployment to0replicas:kubectl -n rook-ceph scale deploy rook-ceph-operator --replicas 0
Scale the
ceph-maintenance-controllerdeployment to3replicas:kubectl -n ceph-lcm-mirantis scale deploy ceph-maintenance-controller --replicas 3
Once done, Rook removes the obsolete Ceph Monitor from the node and creates
a new one on the specified node with a new letter. For example, if the a,
b, and c Ceph Monitors were in quorum and mon-c was obsolete, Rook
removes mon-c and creates mon-d. In this case, the new quorum includes
the a, b, and d Ceph Monitors.
Migrate a Ceph Monitor before machine replacement¶
Available since MCC 2.25.0 (Cluster release 17.0.0)
This document describes how to migrate a Ceph Monitor to another machine on baremetal-based clusters before node replacement as described in Delete a cluster machine using web UI.
Warning
Remove the Ceph Monitor role before the machine removal.
Make sure that the Ceph cluster always has an odd number of Ceph Monitors.
The procedure of a Ceph Monitor migration assumes that you temporarily move the Ceph Manager/Monitor to a worker machine. After a node replacement, we recommend migrating the Ceph Manager/Monitor to the new manager machine.
To migrate a Ceph Monitor to another machine:
Move the Ceph Manager/Monitor daemon from the affected machine to one of the worker machines as described in Move a Ceph Monitor daemon to another node.
Delete the affected machine as described in Delete a cluster machine.
Add a new manager machine without the Monitor and Manager role as described in Create a machine using CLI.
Warning
The addition of a new machine with the Monitor and Manager role breaks the odd number quorum of Ceph Monitors.
Move the previously migrated Ceph Manager/Monitor daemon to the new manager machine as described in Move a Ceph Monitor daemon to another node.
Enable Ceph RGW Object Storage¶
Ceph Controller enables you to deploy RADOS Gateway (RGW) Object Storage instances and automatically manage its resources such as users and buckets. Ceph Object Storage has an integration with OpenStack Object Storage (Swift) in MOSK.
To enable the RGW Object Storage:
Open the
KaasCephClusterCR of a managed cluster for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with a corresponding value.Using the following table, update the
cephClusterSpec.objectStorage.rgwsection specification as required:Caution
Since MCC 2.24.0 (Cluster releases 15.0.1 and 14.0.1), explicitly specify the
deviceClassparameter fordataPoolandmetadataPool.Warning
Since Container Cloud 2.6.0, the
spec.rgwsection is deprecated and its parameters are moved underobjectStorage.rgw. If you continue usingspec.rgw, it is automatically translated intoobjectStorage.rgwduring the Container Cloud update to 2.6.0.We strongly recommend changing
spec.rgwtoobjectStorage.rgwin allKaaSCephClusterCRs beforespec.rgwbecomes unsupported and is deleted.RADOS Gateway parameters¶ Parameter
Description
nameCeph Object Storage instance name.
dataPoolMutually exclusive with the
zoneparameter. Object storage data pool spec that should only containreplicatedorerasureCodedandfailureDomainparameters. ThefailureDomainparameter may be set toosdorhost, defining the failure domain across which the data will be spread. FordataPool, Mirantis recommends using anerasureCodedpool. For details, see Rook documentation: Erasure coding. For example:cephClusterSpec: objectStorage: rgw: dataPool: erasureCoded: codingChunks: 1 dataChunks: 2
metadataPoolMutually exclusive with the
zoneparameter. Object storage metadata pool spec that should only containreplicatedandfailureDomainparameters. ThefailureDomainparameter may be set toosdorhost, defining the failure domain across which the data will be spread. Can use onlyreplicatedsettings. For example:cephClusterSpec: objectStorage: rgw: metadataPool: replicated: size: 3 failureDomain: host
where
replicated.sizeis the number of full copies of data on multiple nodes.Warning
When using the non-recommended Ceph pools
replicated.sizeof less than3, Ceph OSD removal cannot be performed. The minimal replica size equals a rounded up half of the specifiedreplicated.size.For example, if
replicated.sizeis2, the minimal replica size is1, and ifreplicated.sizeis3, then the minimal replica size is2. The replica size of1allows Ceph having PGs with only one Ceph OSD in theactingstate, which may cause aPG_TOO_DEGRADEDhealth warning that blocks Ceph OSD removal. Mirantis recommends settingreplicated.sizeto3for each Ceph pool.gatewayThe gateway settings corresponding to the
rgwdaemon settings. Includes the following parameters:port- the port on which the Ceph RGW service will be listening on HTTP.securePort- the port on which the Ceph RGW service will be listening on HTTPS.instances- the number of pods in the Ceph RGW ReplicaSet. IfallNodesis set totrue, a DaemonSet is created instead.Note
Mirantis recommends using 2 instances for Ceph Object Storage.
allNodes- defines whether to start the Ceph RGW pods as a DaemonSet on all nodes. Theinstancesparameter is ignored ifallNodesis set totrue.
For example:
cephClusterSpec: objectStorage: rgw: gateway: allNodes: false instances: 1 port: 80 securePort: 8443
preservePoolsOnDeleteDefines whether to delete the data and metadata pools in the
rgwsection if the object storage is deleted. Set this parameter totrueif you need to store data even if the object storage is deleted. However, Mirantis recommends setting this parameter tofalse.objectUsersandbucketsOptional. To create new Ceph RGW resources, such as buckets or users, specify the following keys. Ceph Controller will automatically create the specified object storage users and buckets in the Ceph cluster.
objectUsers- a list of user specifications to create for object storage. Contains the following fields:name- a user name to create.displayName- the Ceph user name to display.capabilities- user capabilities:user- admin capabilities to read/write Ceph Object Store users.bucket- admin capabilities to read/write Ceph Object Store buckets.metadata- admin capabilities to read/write Ceph Object Store metadata.usage- admin capabilities to read/write Ceph Object Store usage.zone- admin capabilities to read/write Ceph Object Store zones.
The available options are
*,read,write,read, write. For details, see Ceph documentation: Add/remove admin capabilities.quotas- user quotas:maxBuckets- the maximum bucket limit for the Ceph user. Integer, for example,10.maxSize- the maximum size limit of all objects across all the buckets of a user. String size, for example,10G.maxObjects- the maximum number of objects across all buckets of a user. Integer, for example,10.
For example:
objectUsers: - capabilities: bucket: '*' metadata: read user: read displayName: test-user name: test-user quotas: maxBuckets: 10 maxSize: 10G
users- a list of strings that contain user names to create for object storage.Note
This field is deprecated. Use
objectUsersinstead. Ifusersis specified, it will be automatically transformed to theobjectUserssection.buckets- a list of strings that contain bucket names to create for object storage.
zoneOptional. Mutually exclusive with
metadataPoolanddataPool. Defines the Ceph Multisite zone where the object storage must be placed. Includes thenameparameter that must be set to one of thezonesitems. For details, see Enable multisite for Ceph RGW Object Storage.For example:
cephClusterSpec: objectStorage: multisite: zones: - name: master-zone ... rgw: zone: name: master-zone
SSLCertOptional. Custom TLS certificate parameters used to access the Ceph RGW endpoint. If not specified, a self-signed certificate will be generated.
For example:
cephClusterSpec: objectStorage: rgw: SSLCert: cacert: | -----BEGIN CERTIFICATE----- ca-certificate here -----END CERTIFICATE----- tlsCert: | -----BEGIN CERTIFICATE----- private TLS certificate here -----END CERTIFICATE----- tlsKey: | -----BEGIN RSA PRIVATE KEY----- private TLS key here -----END RSA PRIVATE KEY-----
SSLCertInRefOptional. Available since {{ product_name_abbr }} 25.1. Flag to determine that a TLS certificate for accessing the Ceph RGW endpoint is used but not exposed in
spec. For example:cephClusterSpec: objectStorage: rgw: SSLCertInRef: true
The operator must manually provide TLS configuration using the
rgw-ssl-certificatesecret in therook-cephnamespace of the managed cluster. The secret object must have the following structure:data: cacert: <base64encodedCaCertificate> cert: <base64encodedCertificate>
When removing an already existing
SSLCertblock, no additional actions are required, because this block uses the samergw-ssl-certificatesecret in therook-cephnamespace.When adding a new secret directly without exposing it in
spec, the following rules apply:cert- base64 representation of a file with the server TLS key, server TLS cert, and cacert.cacert- base64 representation of a cacert only.
For example:
cephClusterSpec: objectStorage: rgw: name: rgw-store dataPool: deviceClass: hdd erasureCoded: codingChunks: 1 dataChunks: 2 failureDomain: host metadataPool: deviceClass: hdd failureDomain: host replicated: size: 3 gateway: allNodes: false instances: 1 port: 80 securePort: 8443 preservePoolsOnDelete: false
Enable multisite for Ceph RGW Object Storage¶
TechPreview
The Ceph multisite feature allows object storage to replicate its data over multiple Ceph clusters. Using multisite, such object storage is independent and isolated from another object storage in the cluster. Only the multi-zone multisite setup is currently supported. For more details, see Ceph documentation: Multisite.
Enable the multisite RGW Object Storage¶
Select from the following options:
If you do not have a Container cloud cluster yet, open
kaascephcluster.yaml.templatefor editing.If the Container cloud cluster is already deployed, open the
KaasCephClusterCR of a managed cluster for editing:kubectl edit kaascephcluster -n <managedClusterProjectName>
Substitute
<managedClusterProjectName>with a corresponding value.
Using the following table, update the
cephClusterSpec.objectStorage.multisitesection specification as required:Multisite parameters¶ Parameter
Description
realmsTechnical PreviewList of realms to use, represents the realm namespaces. Includes the following parameters:
name- the realm name.pullEndpoint- optional, required only when the master zone is in a different storage cluster. The endpoint, access key, and system key of the system user from the realm to pull from. Includes the following parameters:endpoint- the endpoint of the master zone in the master zone group.accessKey- the access key of the system user from the realm to pull from.secretKey- the system key of the system user from the realm to pull from.
zoneGroupsTechnical PreviewThe list of zone groups for realms. Includes the following parameters:
name- the zone group name.realmName- the realm namespace name to which the zone group belongs to.
zonesTechnical PreviewThe list of zones used within one zone group. Includes the following parameters:
name- the zone name.metadataPool- the settings used to create the Object Storage metadata pools. Must use replication. For details, see Pool parameters.dataPool- the settings to create the Object Storage data pool. Can use replication or erasure coding. For details, see Pool parameters.zoneGroupName- the zone group name.endpointsForZone- available since {{ product_name_abbr }} 24.2. The list of all endpoints in the zone group. If you use ingress proxy for RGW, the list of endpoints must contain that FQDN/IP address to access RGW. By default, if no ingress proxy is used, the list of endpoints is set to the IP address of the RGW external service. Endpoints must follow the HTTP URL format.
Caution
The multisite configuration requires master and secondary zones to be reachable from each other.
Select from the following options:
If you do not need to replicate data from a different storage cluster, and the current cluster represents the master zone, modify the current
objectStoragesection to use the multisite mode:Configure the
zoneRADOS Gateway (RGW) parameter by setting it to the RGW Object Storage name.Note
Leave
dataPoolandmetadataPoolempty. These parameters are ignored because thezoneblock in the multisite configuration specifies the pools parameters. Other RGW parameters do not require changes.For example:
objectStorage: rgw: dataPool: {} gateway: allNodes: false instances: 2 port: 80 securePort: 8443 healthCheck: {} metadataPool: {} name: openstack-store preservePoolsOnDelete: false zone: name: openstack-store
Create the
multiSitesection where the names of realm, zone group, and zone must match the current RGW name.Since MCC 2.27.0 (Cluster release 17.2.0), specify the
endpointsForZoneparameter according to your configuration:If you use ingress proxy, which is defined in the
spec.cephClusterSpec.ingresssection, add the FQDN endpoint.If you do not use any ingress proxy and access the RGW API using the default RGW external service, add the IP address of the external service or leave this parameter empty.
The following example illustrates a complete
objectStoragesection:objectStorage: multiSite: realms: - name: openstack-store zoneGroups: - name: openstack-store realmName: openstack-store zones: - name: openstack-store zoneGroupName: openstack-store endpointsForZone: http://10.11.0.75:8080 metadataPool: failureDomain: host replicated: size: 3 dataPool: erasureCoded: codingChunks: 1 dataChunks: 2 failureDomain: host rgw: dataPool: {} gateway: allNodes: false instances: 2 port: 80 securePort: 8443 healthCheck: {} metadataPool: {} name: openstack-store preservePoolsOnDelete: false zone: name: openstack-store
If you use a different storage cluster, and its object storage data must be replicated, specify the realm and zone group names along with the
pullEndpointparameter. Additionally, specify the endpoint, access key, and system keys of the system user of the realm from which you need to replicate data. For details, see the step 2 of this procedure.To obtain the endpoint of the cluster zone that must be replicated, run the following command by specifying the zone group name of the required master zone on the master zone side:
radosgw-admin zonegroup get --rgw-zonegroup=<ZONE_GROUP_NAME> | jq -r '.endpoints'
The endpoint is located in the
endpointsfield.To obtain the access key and the secret key of the system user, run the following command on the required Ceph cluster:
radosgw-admin user list
To obtain the system user name, which has your RGW
ObjectStoragename as prefix:radosgw-admin user info --uid="<USER_NAME>" | jq -r '.keys'
For example:
objectStorage: multiSite: realms: - name: openstack-store pullEndpoint: endpoint: http://10.11.0.75:8080 accessKey: DRND5J2SVC9O6FQGEJJF secretKey: qpjIjY4lRFOWh5IAnbrgL5O6RTA1rigvmsqRGSJk zoneGroups: - name: openstack-store realmName: openstack-store zones: - name: openstack-store-backup zoneGroupName: openstack-store metadataPool: failureDomain: host replicated: size: 3 dataPool: erasureCoded: codingChunks: 1 dataChunks: 2 failureDomain: host
Note
Mirantis recommends using the same
metadataPoolanddataPoolsettings as you use in the master zone.
Configure the
zoneRGW parameter and leavedataPoolandmetadataPoolempty. These parameters are ignored because thezonesection in the multisite configuration specifies the pools parameters.Also, you can split the RGW daemon on daemons serving clients and daemons running synchronization. To enable this option, specify
splitDaemonForMultisiteTrafficSyncin thegatewaysection.For example:
objectStorage: multiSite: realms: - name: openstack-store pullEndpoint: endpoint: http://10.11.0.75:8080 accessKey: DRND5J2SVC9O6FQGEJJF secretKey: qpjIjY4lRFOWh5IAnbrgL5O6RTA1rigvmsqRGSJk zoneGroups: - name: openstack-store realmName: openstack-store zones: - name: openstack-store-backup zoneGroupName: openstack-store metadataPool: failureDomain: host replicated: size: 3 dataPool: erasureCoded: codingChunks: 1 dataChunks: 2 failureDomain: host rgw: dataPool: {} gateway: allNodes: false instances: 2 splitDaemonForMultisiteTrafficSync: true port: 80 securePort: 8443 healthCheck: {} metadataPool: {} name: openstack-store-backup preservePoolsOnDelete: false zone: name: openstack-store-backup
On the
ceph-toolspod, verify the multisite status:radosgw-admin sync status
Once done, ceph-operator will create the required resources and Rook will
handle the multisite configuration. For details, see: Rook documentation:
Object Multisite.
Configure and clean up a multisite configuration¶
Warning
Rook does not handle multisite configuration changes and cleanup.
Therefore, once you enable multisite for Ceph RGW Object Storage, perform
these operations manually in the ceph-tools pod. For details, see
Rook documentation: Multisite cleanup.
If automatic update of zone group hostnames is disabled, manually specify all
required hostnames and update the zone group. In the ceph-tools pod, run
the following script:
/usr/local/bin/zonegroup_hostnames_update.sh --rgw-zonegroup <ZONEGROUP_NAME> --hostnames fqdn1[,fqdn2]
If the multisite setup is completely cleaned up, manually execute the following
steps on the ceph-tools pod:
Remove the
.rgw.rootpool:ceph osd pool rm .rgw.root .rgw.root --yes-i-really-really-mean-it
Some other RGW pools may also require a removal after cleanup.
Remove the related RGW crush rules:
ceph osd crush rule ls | grep rgw | xargs -I% ceph osd crush rule rm %
Manage Ceph RBD or CephFS clients and RGW users¶
Available since 2.23.1 (Cluster release 12.7.0)
The section describes how to create, access, and remove Ceph RADOS Block Device (RBD) or Ceph File System (CephFS) clients and RADOS Gateway (RGW) users.
Manage Ceph RBD or CephFS clients¶
Available since 2.23.1 (Cluster release 12.7.0)
The KaaSCephCluster resource allows managing custom Ceph RADOS Block
Device (RBD) or Ceph File System (CephFS) clients. This section describes
how to create, access, and remove Ceph RBD or CephFS clients.
For all supported parameters of Ceph clients, refer to Clients parameters.
Edit the
KaaSCephClusterresource by adding a new Ceph client to thespecsection:kubectl -n <managedClusterProjectName> edit kaascephcluster
Substitute
<managedClusterProject>with the corresponding Container Cloud project where the managed cluster was created.Example of adding an RBD client to the
kubernetes-ssdpool:spec: cephClusterSpec: clients: - name: rbd-client caps: mon: allow r, allow command "osd blacklist" osd: profile rbd pool=kubernetes-ssd
Example of adding a CephFS client to the
cephfs-1Ceph File System :spec: cephClusterSpec: clients: - name: cephfs-1-client caps: mds: allow rwp mon: allow r, allow command "osd blacklist" osd: allow rw tag cephfs data=cephfs-1 metadata=*
For details about
caps, refer to Ceph documentation: Authorization (capabilities).Note
Ceph supports only providing of client access to the whole Ceph File System with all data pools in it.
Wait for created clients to become ready in the
KaaSCephClusterstatus:kubectl -n <managedClusterProject> get kaascephcluster -o yaml
Example output:
status: fullClusterInfo: blockStorageStatus: clientsStatus: rbd-client: present: true status: Ready cephfs-1-client: present: true status: Ready
Using the
KaaSCephClusterstatus, obtainsecretInfowith the Ceph client credentials :kubectl -n <managedClusterProject> get kaascephcluster -o yaml
Example output:
status: miraCephSecretsInfo: secretInfo: clientSecrets: - name: rbd-client secretName: rook-ceph-client-rbd-client secretNamespace: rook-ceph - name: cephfs-1-client secretName: rook-ceph-client-cephfs-1-client secretNamespace: rook-ceph
Use
secretNameandsecretNamespaceto access the Ceph client credentials from a managed cluster:kubectl --kubeconfig <managedClusterKubeconfig> -n <secretNamespace> get secret <secretName> -o jsonpath='{.data.<clientName>}' | base64 -d; echo
Substitute the following parameters:
<managedClusterKubeconfig>with a managed clusterkubeconfig<secretNamespace>withsecretNamespacefrom the previous step<secretName>withsecretNamefrom the previous step<clientName>with the Ceph RBD or CephFS client name set inspec.cephClusterSpec.clientstheKaaSCephClusterresource, for example,rbd-client
Example output:
AQAGHDNjxWYXJhAAjafCn3EtC6KgzgI1x4XDlg==
Using the obtained credentials, create two configuration files on the required workloads to connect them with Ceph pools or file systems:
/etc/ceph/ceph.conf:[default] mon_host = <mon1IP>:6789,<mon2IP>:6789,...,<monNIP>:6789
where
mon_hostare the comma-separated IP addresses with6789ports of the current Ceph Monitors. For example,10.10.0.145:6789,10.10.0.153:6789,10.10.0.235:6789./etc/ceph/ceph.client.<clientName>.keyring:[client.<clientName>] key = <cephClientCredentials>
<clientName>is a client name set inspec.cephClusterSpec.clientstheKaaSCephClusterresource, for example,rbd-client<cephClientCredentials>are the client credentials obtained in the previous steps. For example,AQAGHDNjxWYXJhAAjafCn3EtC6KgzgI1x4XDlg==
If the client
capsparameters containmon: allow r, verify the client access using the following command:ceph -n client.<clientName> -s
Edit the
KaaSCephClusterresource by removing the Ceph client fromspec.cephClusterSpec.clients:kubectl -n <managedClusterProject> edit kaascephcluster
Wait for the client to be removed from the
KaaSCephClusterstatus instatus.fullClusterInfo.blockStorageStatus.clientsStatus:kubectl -n <managedClusterProject> get kaascephcluster -o yaml
Manage Ceph Object Storage users¶
Available since 2.23.1 (Cluster release 12.7.0)
The KaaSCephCluster resource allows managing custom Ceph Object Storage
users. This section describes how to create, access, and remove Ceph Object
Storage users.
For all supported parameters of Ceph Object Storage users, refer to RADOS Gateway parameters.
Edit the
KaaSCephClusterresource by adding a new Ceph Object Storage user to thespecsection:kubectl -n <managedClusterProject> edit kaascephcluster
Substitute
<managedClusterProject>with the corresponding Container Cloud project where the managed cluster was created.Example of adding the Ceph Object Storage user
user-a:Caution
For user
name, apply the UUID format with no capital letters.spec: cephClusterSpec: objectStorage: rgw: objectUsers: - capabilities: bucket: '*' metadata: read user: read displayName: user-a name: userA quotas: maxBuckets: 10 maxSize: 10G
Wait for the created user to become ready in the
KaaSCephClusterstatus:kubectl -n <managedClusterProject> get kaascephcluster -o yaml
Example output:
status: fullClusterInfo: objectStorageStatus: objectStoreUsers: user-a: present: true phase: Ready
Using the
KaaSCephClusterstatus, obtainsecretInfowith the Ceph user credentials :kubectl -n <managedClusterProject> get kaascephcluster -o yaml
Example output:
status: miraCephSecretsInfo: secretInfo: rgwUserSecrets: - name: user-a secretName: rook-ceph-object-user-<objstoreName>-<username> secretNamespace: rook-ceph
Substitute
<objstoreName>with a Ceph Object Storage name and<username>with a Ceph Object Storage user name.Use
secretNameandsecretNamespaceto access the Ceph Object Storage user credentials from a managed cluster. The secret contains Amazon S3 access and secret keys.To obtain the user S3 access key:
kubectl --kubeconfig <managedClusterKubeconfig> -n <secretNamespace> get secret <secretName> -o jsonpath='{.data.AccessKey}' | base64 -d; echo
Substitute the following parameters in the commands above and below:
<managedClusterKubeconfig>with a managed clusterkubeconfig<secretNamespace>withsecretNamespacefrom the previous step<secretName>withsecretNamefrom the previous step
Example output:
D49G060HQ86U5COBTJ13To obtain the user S3 secret key:
kubectl --kubeconfig <managedClusterKubeconfig> -n <secretNamespace> get secret <secretName> -o jsonpath='{.data.SecretKey}' | base64 -d; echo
Example output:
bpuYqIieKvzxl6nzN0sd7L06H40kZGXNStD4UNda
Configure the S3 client with the access and secret keys of the created user. You can access the S3 client using various tools such as s3cmd or awscli.
Edit the
KaaSCephClusterresource by removing the required Ceph Object Storage user fromspec.cephClusterSpec.objectStorage.rgw.objectUsers:kubectl -n <managedClusterProject> edit kaascephcluster
Wait for the removed user to be removed from the
KaaSCephClusterstatus instatus.fullClusterInfo.objectStorageStatus.objectStoreUsers:kubectl -n <managedClusterProject> get kaascephcluster -o yaml
Verify Ceph¶
This section describes how to verify the components of a Ceph cluster after deployment. For troubleshooting, verify Ceph Controller and Rook logs as described in Verify Ceph Controller and Rook.
Verify the Ceph core services¶
To confirm that all Ceph components including mon, mgr, osd, and
rgw have joined your cluster properly, analyze the logs for each pod and
verify the Ceph status:
kubectl -n rook-ceph exec -it $(kubectl -n rook-ceph get pod -l "app=rook-ceph-tools" -o jsonpath='{.items[0].metadata.name}') bash
ceph -s
Example of a positive system response:
cluster:
id: 4336ab3b-2025-4c7b-b9a9-3999944853c8
health: HEALTH_OK
services:
mon: 3 daemons, quorum a,b,c (age 20m)
mgr: a(active, since 19m)
osd: 6 osds: 6 up (since 16m), 6 in (since 16m)
rgw: 1 daemon active (miraobjstore.a)
data:
pools: 12 pools, 216 pgs
objects: 201 objects, 3.9 KiB
usage: 6.1 GiB used, 174 GiB / 180 GiB avail
pgs: 216 active+clean
Verify rook-discover¶
To ensure that rook-discover is running properly, verify if the
local-device configmap has been created for each Ceph node specified
in the cluster configuration:
Obtain the list of local devices:
kubectl get configmap -n rook-ceph | grep local-device
Example of a system response:
local-device-01 1 30m local-device-02 1 29m local-device-03 1 30m
Verify that each device from the list contains information about available devices for the Ceph node deployment:
kubectl describe configmap local-device-01 -n rook-ceph
Example of a positive system response:
Name: local-device-01 Namespace: rook-ceph Labels: app=rook-discover rook.io/node=01 Annotations: <none> Data ==== devices: ---- [{"name":"vdd","parent":"","hasChildren":false,"devLinks":"/dev/disk/by-id/virtio-41d72dac-c0ff-4f24-b /dev/disk/by-path/virtio-pci-0000:00:09.0","size":32212254720,"uuid":"27e9cf64-85f4-48e7-8862-faa7270202ed","serial":"41d72dac-c0ff-4f24-b","type":"disk","rotational":true,"readOnly":false,"Partitions":null,"filesystem":"","vendor":"","model":"","wwn":"","wwnVendorExtension":"","empty":true,"cephVolumeData":"{\"path\":\"/dev/vdd\",\"available\":true,\"rejected_reasons\":[],\"sys_api\":{\"size\":32212254720.0,\"scheduler_mode\":\"none\",\"rotational\":\"1\",\"vendor\":\"0x1af4\",\"human_readable_size\":\"30.00 GB\",\"sectors\":0,\"sas_device_handle\":\"\",\"rev\":\"\",\"sas_address\":\"\",\"locked\":0,\"sectorsize\":\"512\",\"removable\":\"0\",\"path\":\"/dev/vdd\",\"support_discard\":\"0\",\"model\":\"\",\"ro\":\"0\",\"nr_requests\":\"128\",\"partitions\":{}},\"lvs\":[]}","label":""},{"name":"vdb","parent":"","hasChildren":false,"devLinks":"/dev/disk/by-path/virtio-pci-0000:00:07.0","size":67108864,"uuid":"988692e5-94ac-4c9a-bc48-7b057dd94fa4","serial":"","type":"disk","rotational":true,"readOnly":false,"Partitions":null,"filesystem":"","vendor":"","model":"","wwn":"","wwnVendorExtension":"","empty":true,"cephVolumeData":"{\"path\":\"/dev/vdb\",\"available\":false,\"rejected_reasons\":[\"Insufficient space (\\u003c5GB)\"],\"sys_api\":{\"size\":67108864.0,\"scheduler_mode\":\"none\",\"rotational\":\"1\",\"vendor\":\"0x1af4\",\"human_readable_size\":\"64.00 MB\",\"sectors\":0,\"sas_device_handle\":\"\",\"rev\":\"\",\"sas_address\":\"\",\"locked\":0,\"sectorsize\":\"512\",\"removable\":\"0\",\"path\":\"/dev/vdb\",\"support_discard\":\"0\",\"model\":\"\",\"ro\":\"0\",\"nr_requests\":\"128\",\"partitions\":{}},\"lvs\":[]}","label":""},{"name":"vdc","parent":"","hasChildren":false,"devLinks":"/dev/disk/by-id/virtio-e8fdba13-e24b-41f0-9 /dev/disk/by-path/virtio-pci-0000:00:08.0","size":32212254720,"uuid":"190a50e7-bc79-43a9-a6e6-81b173cd2e0c","serial":"e8fdba13-e24b-41f0-9","type":"disk","rotational":true,"readOnly":false,"Partitions":null,"filesystem":"","vendor":"","model":"","wwn":"","wwnVendorExtension":"","empty":true,"cephVolumeData":"{\"path\":\"/dev/vdc\",\"available\":true,\"rejected_reasons\":[],\"sys_api\":{\"size\":32212254720.0,\"scheduler_mode\":\"none\",\"rotational\":\"1\",\"vendor\":\"0x1af4\",\"human_readable_size\":\"30.00 GB\",\"sectors\":0,\"sas_device_handle\":\"\",\"rev\":\"\",\"sas_address\":\"\",\"locked\":0,\"sectorsize\":\"512\",\"removable\":\"0\",\"path\":\"/dev/vdc\",\"support_discard\":\"0\",\"model\":\"\",\"ro\":\"0\",\"nr_requests\":\"128\",\"partitions\":{}},\"lvs\":[]}","label":""}]
Verify Ceph cluster state through CLI¶
Verifying Ceph cluster state is an entry point for issues investigation.
This section describes how to verify Ceph state using the KaaSCephCluster,
MiraCeph, and MiraCephLog resources.
Note
Before MOSK 25.1, use MiraCephLog
instead of MiraCephHealth.
To verify the state of a Ceph cluster, Ceph Controller provides special
sections in KaaSCephCluster.status. The resource contains information about
the state of the Ceph cluster components, their health, and potentially
problematic components.
To verify the Ceph cluster state from a managed cluster:
Obtain
kubeconfigof a managed cluster and provide it as an environment variable:export KUBECONFIG=<pathToManagedKubeconfig>
Obtain the
MiraCephresource in YAML format:kubectl -n ceph-lcm-mirantis get miraceph -o yaml
Information from
MiraCeph.statusis passed to themiraCephInfosection of theKaaSCephClusterCR. For details, see KaaSCephCluster.status miraCephInfo specification.Obtain the
MiraCephHealthresource in YAML format:kubectl -n ceph-lcm-mirantis get miracephhealth -o yaml
Information from
MiraCephHealthis passed to thefullClusterInfoandshortClusterInfosections of theKaaSCephClusterCR. For details, see KaaSCephCluster.status shortClusterInfo specification and KaaSCephCluster.status fullClusterInfo specification.Note
Before MOSK 25.1, use
MiraCephLoginstead ofMiraCephHealthas the resource name and in the command above.
To verify the Ceph cluster state from a management cluster:
Obtain the
KaaSCephClusterresource in the YAML format:kubectl -n <projectName> get kaascephcluster -o yaml
Substitute
<projectName>with the project name of the managed cluster.Verify the state of the required component using KaaSCephCluster.status description.
KaaSCephCluster.status allows you to learn the current health of a Ceph
cluster and identify potentially problematic components. This section describes
KaaSCephCluster.status and its fields. To view KaaSCephCluster.status,
perform the steps described in Verify Ceph cluster state through CLI.
KaaSCephCluster.status miraCephSecretsInfo specification Since MCC 2.23.1 (Cluster release 12.7.0)
Field |
Description |
|---|---|
|
Available since MCC 2.25.0 (Cluster release 17.0.0).
Describes the current state of |
|
Describes the current phase of Ceph spec reconciliation and spec
validation result. The |
|
Reresents a short version of |
|
Contains a complete Ceph cluster information including cluster, Ceph
resources, and daemons health. It helps to reveal the potentially
problematic components. For |
|
Contains information about secrets of the managed cluster that are used
in the Ceph cluster, such as keyrings, Ceph clients, RADOS Gateway user
credentials, and so on. For |
The following tables describe all sections of KaaSCephCluster.status.
Field |
Description |
|---|---|
|
Contains the current phase of handling of the applied Ceph cluster spec.
Can equal to |
|
Contains a detailed description of the current phase or an error message
if the phase is |
|
Contains the validation:
result: Succeed or Failed
messages: ["error", "messages", "list"]
|
Field |
Description |
|---|---|
|
Current Ceph cluster collector status:
|
|
|
|
|
|
List of error or warning messages found when gathering the facts about the Ceph cluster. |
Field |
Description |
|---|---|
|
General information from Rook about the Ceph cluster health and current
state. The clusterStatus:
state: <rook ceph cluster common status>
phase: <rook ceph cluster spec reconcile phase>
message: <rook ceph cluster phase details>
conditions: <history of rook ceph cluster
reconcile steps>
ceph: <ceph cluster health>
storage:
deviceClasses: <list of used device classes
in ceph cluster>
version:
image: <ceph image used in ceph cluster>
version: <ceph version of ceph cluster>
|
|
Status of the Rook Ceph Operator pod that is |
|
Map of statuses for each Ceph cluster daemon type. Indicates the
expected and actual number of Ceph daemons on the cluster. Available
daemon types are: daemonsStatus:
<daemonType>:
status: <daemons status>
running: <number of running daemons with
details>
For example: daemonsStatus:
mgr:
running: a is active mgr ([] standBy)
status: Ok
mon:
running: '3/3 mons running: [a c d] in quorum'
status: Ok
osd:
running: '4/4 running: 4 up, 4 in'
status: Ok
rgw:
running: 2/2 running
([openstack.store.a openstack.store.b])
status: Ok
|
|
State of the Ceph cluster block storage resources. Includes the following fields:
|
|
State of the Ceph cluster object storage resources. Includes the following fields:
|
|
Verbose details of the Ceph cluster state.
|
|
Contains information, similar to the cephCSIPluginDaemonsStatus:
<csiPlugin>:
running: <number of running daemons with details>
status: <csi plugin status>
For example: cephCSIPluginDaemonsStatus:
csi-rbdplugin:
running: 1/3 running
status: Some csi-rbdplugin daemons are not ready
csi-cephfsplugin:
running: 3/3 running
status: Ok
|
Field |
Description |
|---|---|
|
Current state of the secret collector on the Ceph cluster:
|
|
|
|
|
|
List of secrets for Ceph clients and RADOS Gateway users:
For example: lastSecretCheck: "2022-09-05T07:05:35Z"
lastSecretUpdate: "2022-09-05T06:02:00Z"
secretInfo:
clientSecrets:
- name: client.admin
secretName: rook-ceph-admin-keyring
secretNamespace: rook-ceph
state: Ready
|
|
List of error or warning messages, if any, found when collecting information about the Ceph cluster. |
View Ceph cluster summary through the Container Cloud web UI¶
Warning
Mirantis highly recommends verifying a Ceph cluster using the CLI instead of the web UI. For details, see Verify Ceph cluster state through CLI.
The web UI capabilities for adding and managing a Ceph cluster are limited and lack flexibility in defining Ceph cluster specifications. For example, if an error occurs while adding a Ceph cluster using the web UI, usually you can address it only through the CLI.
The web UI functionality for managing Ceph cluster is going to be deprecated in one of the following releases.
Verifying Ceph cluster state is an entry point for issues investigation. Through the Ceph Clusters page of the Container Cloud web UI, you can view a detailed summary on all Ceph clusters deployed, including the cluster name and ID, health status, number of Ceph OSDs, and so on.
To view Ceph cluster summary:
Log in to the Container Cloud web UI with the
m:kaas:namespace@operatororm:kaas:namespace@writerpermissions.Switch to the required project using the Switch Project action icon located on top of the main left-side navigation panel.
In the Clusters tab, click the required cluster name. The page with cluster details opens.
In the Ceph Clusters tab, verify the overall cluster health and rebalancing statuses.
Available since MCC 2.25.0 (Cluster release 17.0.0). Click Cluster Details:
The Machines tab contains the list of deployed Ceph machines with the following details:
Status - deployment status
Role - role assigned to a machine, manager or monitor
Storage devices - number of storage devices assigned to a machine
UP OSDs and IN OSDs - number of
upandinCeph OSDs belonging to a machine
Note
To obtain details about a specific machine used for Ceph deployment, in the Clusters > <clusterName> > Machines tab, click the required machine name containing the storage label.
The OSDs tab contains the list of Ceph OSDs comprising the Ceph cluster with the following details:
OSD - Ceph OSD ID
Storage Device ID - storage device ID assigned to a Ceph OSD
Type - type of storage device assigned to a Ceph OSD
Partition - partition name where Ceph OSD is located
Machine - machine name where Ceph OSD is located
UP/DOWN - status of a Ceph OSD in a cluster
IN/OUT - service state of a Ceph OSD in a cluster
Verify Ceph Controller and Rook¶
The starting point for Ceph troubleshooting is the ceph-controller and
rook-operator logs. Once you locate the component that causes issues,
verify the logs of the related pod. This section describes how to verify the
Ceph Controller and Rook objects of a Ceph cluster.
To verify Ceph Controller and Rook:
Verify the Ceph cluster status:
Verify that the status of each pod in the
ceph-lcm-mirantisandrook-cephnamespaces isRunning:For
ceph-lcm-mirantis:kubectl get pod -n ceph-lcm-mirantis
For
rook-ceph:kubectl get pod -n rook-ceph
Verify Ceph Controller. Ceph Controller prepares the configuration that Rook uses to deploy the Ceph cluster, managed using the
KaasCephClusterresource. If Rook cannot finish the deployment, verify the Rook Operator logs as described in the step 4.List the pods:
kubectl -n ceph-lcm-mirantis get pods
Verify the logs of the required pod:
kubectl -n ceph-lcm-mirantis logs <ceph-controller-pod-name>
Verify the configuration:
kubectl get kaascephcluster -n <managedClusterProjectName> -o yaml
On the managed cluster, verify the
MiraCephsubresource:kubectl get miraceph -n ceph-lcm-mirantis -o yaml
Verify the Rook Operator logs. Rook deploys a Ceph cluster based on custom resources created by the Ceph Controller, such as pools, clients,
cephcluster, and so on. Rook logs contain details about components orchestration. For details about the Ceph cluster status and to get access to CLI tools, connect to theceph-toolspod as described in the step 5.Verify the Rook Operator logs:
kubectl -n rook-ceph logs -l app=rook-ceph-operator
Verify the
CephClusterconfiguration:Note
The Ceph Controller manages the
CephClusterCR . Open theCephClusterCR only for verification and do not modify it manually.kubectl get cephcluster -n rook-ceph -o yaml
Verify the
ceph-toolspod:Execute the
ceph-toolspod:kubectl --kubeconfig <pathToManagedClusterKubeconfig> -n rook-ceph exec -it $(kubectl --kubeconfig <pathToManagedClusterKubeconfig> -n rook-ceph get pod -l app=rook-ceph-tools -o jsonpath='{.items[0].metadata.name}') bash
Verify that CLI commands can run on the
ceph-toolspod:ceph -s
Verify hardware:
Through the
ceph-toolspod, obtain the required device in your cluster:ceph osd tree
Enter all Ceph OSD pods in the
rook-cephnamespace one by one:kubectl exec -it -n rook-ceph <osd-pod-name> bash
Verify that the
ceph-volumetool is available on all pods running on the target node:ceph-volume lvm list
Verify data access. Ceph volumes can be consumed directly by Kubernetes workloads and internally, for example, by OpenStack services. To verify the Kubernetes storage:
Verify the available storage classes. The storage classes that are automatically managed by Ceph Controller use the
rook-ceph.rbd.csi.ceph.comprovisioner.kubectl get storageclass
Example of system response:
NAME PROVISIONER RECLAIMPOLICY VOLUMEBINDINGMODE ALLOWVOLUMEEXPANSION AGE kubernetes-ssd (default) rook-ceph.rbd.csi.ceph.com Delete Immediate false 55m stacklight-alertmanager-data kubernetes.io/no-provisioner Delete WaitForFirstConsumer false 55m stacklight-elasticsearch-data kubernetes.io/no-provisioner Delete WaitForFirstConsumer false 55m stacklight-postgresql-db kubernetes.io/no-provisioner Delete WaitForFirstConsumer false 55m stacklight-prometheus-data kubernetes.io/no-provisioner Delete WaitForFirstConsumer false 55m
Verify that volumes are properly connected to the Pod:
Obtain the list of volumes in all namespaces or use a particular one:
kubectl get persistentvolumeclaims -A
Example of system response:
NAMESPACE NAME STATUS VOLUME CAPACITY ACCESS MODES STORAGECLASS AGE rook-ceph app-test Bound pv-test 1Gi RWO kubernetes-ssd 11m
For each volume, verify the connection. For example:
kubectl describe pvc app-test -n rook-ceph
Example of a positive system response:
Name: app-test Namespace: kaas StorageClass: rook-ceph Status: Bound Volume: pv-test Labels: <none> Annotations: pv.kubernetes.io/bind-completed: yes pv.kubernetes.io/bound-by-controller: yes volume.beta.kubernetes.io/storage-provisioner: rook-ceph.rbd.csi.ceph.com Finalizers: [kubernetes.io/pvc-protection] Capacity: 1Gi Access Modes: RWO VolumeMode: Filesystem Events: <none>
In case of connection issues, inspect the Pod description for the volume information:
kubectl describe pod <crashloopbackoff-pod-name>
Example of system response:
... Events: FirstSeen LastSeen Count From SubObjectPath Type Reason Message --------- -------- ----- ---- ------------- -------- ------ ------- 1h 1h 3 default-scheduler Warning FailedScheduling PersistentVolumeClaim is not bound: "app-test" (repeated 2 times) 1h 35s 36 kubelet, 172.17.8.101 Warning FailedMount Unable to mount volumes for pod "wordpress-mysql-918363043-50pjr_default(08d14e75-bd99-11e7-bc4c-001c428b9fc8)": timeout expired waiting for volumes to attach/mount for pod "default"/"wordpress-mysql-918363043-50pjr". list of unattached/unmounted volumes=[mysql-persistent-storage] 1h 35s 36 kubelet, 172.17.8.101 Warning FailedSync Error syncing pod
Verify that the CSI provisioner plugins started properly and are in the
Runningstatus:Obtain the list of CSI provisioner plugins:
kubectl -n rook-ceph get pod -l app=csi-rbdplugin-provisioner
Verify the logs of the required CSI provisioner:
kubectl logs -n rook-ceph <csi-provisioner-plugin-name> csi-provisioner
Enable Ceph tolerations and resources management¶
This section describes how to configure Ceph Controller to manage Ceph nodes resources.
Enable Ceph tolerations and resources management¶
Warning
This document does not provide any specific recommendations on requests and limits for Ceph resources. The document stands for a native Ceph resources configuration for any cluster with MOSK.
You can configure Ceph Controller to manage Ceph resources by specifying their requirements and constraints. To configure the resources consumption for the Ceph nodes, consider the following options that are based on different Helm release configuration values:
Configuring tolerations for taint nodes for the Ceph Monitor, Ceph Manager, and Ceph OSD daemons. For details, see Taints and Tolerations.
Configuring nodes resources requests or limits for the Ceph daemons and for each Ceph OSD device class such as HDD, SSD, or NVMe. For details, see Managing Resources for Containers.
To enable Ceph tolerations and resources management:
To avoid Ceph cluster health issues during daemons configuration changing, set Ceph
noout,nobackfill,norebalance, andnorecoverflags through theceph-toolspod before editing Ceph tolerations and resources:kubectl -n rook-ceph exec -it $(kubectl -n rook-ceph get pod -l \ "app=rook-ceph-tools" -o jsonpath='{.items[0].metadata.name}') bash ceph osd set noout ceph osd set nobackfill ceph osd set norebalance ceph osd set norecover exit
Note
Skip this step if you are only configuring the PG rebalance timeout and replicas count parameters.
Edit the
KaaSCephClusterresource of a managed cluster:kubectl -n <managedClusterProjectName> edit kaascephcluster
Substitute
<managedClusterProjectName>with the project name of the required managed cluster.Specify the parameters in the
hyperconvergesection as required. Thehyperconvergesection includes the following parameters:Ceph tolerations and resources parameters¶ Parameter
Description
Example values
tolerationsSpecifies tolerations for taint nodes for the defined daemon type. Each daemon type key contains the following parameters:
cephClusterSpec: hyperconverge: tolerations: <daemonType>: rules: - key: "" operator: "" value: "" effect: "" tolerationSeconds: 0
Possible values for
<daemonType>areosd,mon,mgr, andrgw. The following values are also supported:all- specifies general toleration rules for all daemons if no separate daemon rule is specified.mds- specifies the CephFS Metadata Server daemons.
hyperconverge: tolerations: mon: rules: - effect: NoSchedule key: node-role.kubernetes.io/controlplane operator: Exists mgr: rules: - effect: NoSchedule key: node-role.kubernetes.io/controlplane operator: Exists osd: rules: - effect: NoSchedule key: node-role.kubernetes.io/controlplane operator: Exists rgw: rules: - effect: NoSchedule key: node-role.kubernetes.io/controlplane operator: Exists
resourcesSpecifies resources requests or limits. The parameter is a map with the daemon type as a key and the following structure as a value:
hyperconverge: resources: <daemonType>: requests: <kubernetes valid spec of daemon resource requests> limits: <kubernetes valid spec of daemon resource limits>
Possible values for
<daemonType>aremon,mgr,osd,osd-hdd,osd-ssd,osd-nvme,prepareosd,rgw, andmds. Theosd-hdd,osd-ssd, andosd-nvmeresource requirements handle only the Ceph OSDs with a corresponding device class.hyperconverge: resources: mon: requests: memory: 1Gi cpu: 2 limits: memory: 2Gi cpu: 3 mgr: requests: memory: 1Gi cpu: 2 limits: memory: 2Gi cpu: 3 osd: requests: memory: 1Gi cpu: 2 limits: memory: 2Gi cpu: 3 osd-hdd: requests: memory: 1Gi cpu: 2 limits: memory: 2Gi cpu: 3 osd-ssd: requests: memory: 1Gi cpu: 2 limits: memory: 2Gi cpu: 3 osd-nvme: requests: memory: 1Gi cpu: 2 limits: memory: 2Gi cpu: 3
For the Ceph node specific resources settings, specify the
resourcessection in the correspondingnodesspec ofKaaSCephCluster:spec: cephClusterSpec: nodes: <nodeName>: resources: requests: <kubernetes valid spec of daemon resource requests> limits: <kubernetes valid spec of daemon resource limits>
Substitute
<nodeName>with the node requested for specific resources. For example:spec: cephClusterSpec: nodes: <nodeName>: resources: requests: memory: 1Gi cpu: 2 limits: memory: 2Gi cpu: 3
For the RADOS Gateway instances specific resources settings, specify the
resourcessection in thergwspec ofKaaSCephCluster:spec: cephClusterSpec: objectStorage: rgw: gateway: resources: requests: <kubernetes valid spec of daemon resource requests> limits: <kubernetes valid spec of daemon resource limits>
For example:
spec: cephClusterSpec: objectStorage: rgw: gateway: resources: requests: memory: 1Gi cpu: 2 limits: memory: 2Gi cpu: 3
Save the reconfigured
KaaSCephClusterresource and wait forceph-controllerto apply the updated Ceph configuration. It will recreate Ceph Monitors, Ceph Managers, or Ceph OSDs according to the specifiedhyperconvergeconfiguration.If you have specified any
osdtolerations, additionally specify tolerations for therookinstances:Open the
Clusterresource of the required Ceph cluster on a management cluster:kubectl -n <ClusterProjectName> edit cluster
Substitute
<ClusterProjectName>with the project name of the required cluster.Specify the parameters in the
ceph-controllersection ofspec.providerSpec.value.helmReleases:Specify the
hyperconverge.tolerations.rookparameter as required:hyperconverge: tolerations: rook: | <yamlFormattedKubernetesTolerations>
In
<yamlFormattedKubernetesTolerations>, specify YAML-formatted tolerations fromcephClusterSpec.hyperconverge.tolerations.osd.rulesof theKaaSCephClusterspec. For example:hyperconverge: tolerations: rook: | - effect: NoSchedule key: node-role.kubernetes.io/controlplane operator: Exists
In
controllers.cephRequest.parameters.pgRebalanceTimeoutMin, specify the PG rebalance timeout for requests. The default is30minutes. For example:controllers: cephRequest: parameters: pgRebalanceTimeoutMin: 35
In
controllers.cephController.replicas,controllers.cephRequest.replicas, andcontrollers.cephStatus.replicas, specify the replicas count. The default is3replicas. For example:controllers: cephController: replicas: 1 cephRequest: replicas: 1 cephStatus: replicas: 1
Save the reconfigured
Clusterresource and wait for theceph-controllerHelm release update. It will recreate Ceph CSI and discover pods according to the specifiedhyperconverge.tolerations.rookconfiguration.
Specify tolerations for different Rook resources using the following chart-based options:
hyperconverge.tolerations.rook- general toleration rules for each Rook service if no exact rules specifiedhyperconverge.tolerations.csiplugin- for tolerations of theceph-csiplugins DaemonSetshyperconverge.tolerations.csiprovisioner- for theceph-csiprovisioner deployment tolerationshyperconverge.nodeAffinity.csiprovisioner- provides theceph-csiprovisioner node affinity with avaluesection
After a successful Ceph reconfiguration, unset the flags set in step 1 through the
ceph-toolspod:kubectl -n rook-ceph exec -it $(kubectl -n rook-ceph get pod -l \ "app=rook-ceph-tools" -o jsonpath='{.items[0].metadata.name}') bash ceph osd unset ceph osd unset noout ceph osd unset nobackfill ceph osd unset norebalance ceph osd unset norecover exit
Note
Skip this step if you have only configured the PG rebalance timeout and replicas count parameters.
Once done, proceed to Verify Ceph tolerations and resources management.
Verify Ceph tolerations and resources management¶
After you enable Ceph resources management as described in Enable Ceph tolerations and resources management, perform the steps below to verify that the configured tolerations, requests, or limits have been successfully specified in the Ceph cluster.
To verify Ceph tolerations and resources management:
To verify that the required tolerations are specified in the Ceph cluster, inspect the output of the following commands:
kubectl -n rook-ceph get $(kubectl -n rook-ceph get cephcluster -o name) -o jsonpath='{.spec.placement.mon.tolerations}' kubectl -n rook-ceph get $(kubectl -n rook-ceph get cephcluster -o name) -o jsonpath='{.spec.placement.mgr.tolerations}' kubectl -n rook-ceph get $(kubectl -n rook-ceph get cephcluster -o name) -o jsonpath='{.spec.placement.osd.tolerations}'
To verify RADOS Gateway tolerations:
kubectl -n rook-ceph get $(kubectl -n rook-ceph get cephobjectstore -o name) -o jsonpath='{.spec.gateway.placement.tolerations}'
To verify that the required resources requests or limits are specified for the Ceph
mon,mgr, orosddaemons, inspect the output of the following command:kubectl -n rook-ceph get $(kubectl -n rook-ceph get cephcluster -o name) -o jsonpath='{.spec.resources}'
To verify that the required resources requests and limits are specified for the RADOS Gateway daemons, inspect the output of the following command:
kubectl -n rook-ceph get $(kubectl -n rook-ceph get cephobjectstore -o name) -o jsonpath='{.spec.gateway.resources}'
To verify that the required resources requests or limits are specified for the Ceph OSDs
hdd,ssd, ornvmedevice classes, perform the following steps:Identify which Ceph OSDs belong to the
<deviceClass>device class in question:kubectl -n rook-ceph exec -it $(kubectl -n rook-ceph get pod -l app=rook-ceph-tools -o name) -- ceph osd crush class ls-osd <deviceClass>
For each
<osdID>obtained in the previous step, run the following command. Compare the output with the desired result.kubectl -n rook-ceph get deploy rook-ceph-osd-<osdID> -o jsonpath='{.spec.template.spec.containers[].resources}'
Enable Ceph multinetwork¶
Ceph allows establishing multiple IP networks and subnet masks for clusters
with configured L3 network rules. In MOSK, you can configure
multinetwork through the network section of the KaaSCephCluster CR.
Ceph Controller uses this section to specify the Ceph networks for external
access and internal daemon communication. The parameters in the network
section use the CIDR notation, for example, 10.0.0.0/24.
Before enabling multiple networks for a Ceph cluster, consider the following requirements:
Do not confuse the IP addresses you define with the public-facing IP addresses the network clients may use to access the services.
If you define more than one IP address and subnet mask for the public or cluster network, ensure that the subnets within the network can route to each other.
Include each IP address or subnet in the
networksection to IP tables and open ports for them as necessary.The pods of the Ceph OSD and RadosGW daemons use cross-pods health checkers to verify that the entire Ceph cluster is healthy. Therefore, each CIDR must be accessible inside Ceph pods.
Avoid using the
0.0.0.0/0CIDR in thenetworksection. With a zero range inpublicNetand/orclusterNet, the Ceph daemons behavior is unpredictable.
To enable multinetwork for Ceph:
Select from the following options:
If the Ceph cluster is not deployed on a managed cluster yet, edit the deployment
KaaSCephClusterYAML template.If the Ceph cluster is already deployed on a managed cluster, open
KaaSCephClusterfor editing:kubectl -n <managedClusterProjectName> edit kaascephcluster
Substitute
<managedClusterProjectName>with a corresponding value.
In the
clusterNetand/orpublicNetparameters of thecephClusterSpec.networksection, define a comma-separated array of CIDRs. For example:network: publicNet: 10.12.0.0/24,10.13.0.0/24 clusterNet: 10.10.0.0/24,10.11.0.0/24
Select from the following options:
If you are creating a managed cluster, save the updated
KaaSCephClustertemplate to the corresponding file and proceed with the managed cluster creation.If you are configuring
KaaSCephClusterof an existing managed cluster, exiting the text editor will apply the changes.
Once done, the specified network CIDRs will be passed to the Ceph daemons pods
through the rook-config-override ConfigMap.
Enable Ceph RBD mirroring¶
This section describes how to configure and use RADOS Block Device (RBD)
mirroring for Ceph pools using the rbdMirror section in the
KaaSCephCluster CR. The feature may be useful if, for example, you have
two interconnected managed clusters. Once you enable RBD mirroring, the
images in the specified pools will be replicated and if a cluster becomes
unreachable, the second one will provide users with instant access to all
images. For details, see Ceph Documentation: RBD Mirroring.
Note
Ceph Controller only supports bidirectional mirroring.
To enable Ceph RBD monitoring, follow the procedure below and use the following
rbdMirror parameters description:
Parameter |
Description |
|---|---|
|
Count of |
|
Optional. List of mirroring peers of an external cluster to connect to.
Only a single peer is supported. The
|
To enable Ceph RBD mirroring:
In
KaaSCephClusterCRs of both Ceph clusters where you want to enable mirroring, specify positivedaemonsCountin thespec.cephClusterSpec.rbdMirrorsection:spec: cephClusterSpec: rbdMirror: daemonsCount: 1
On both Ceph clusters where you want to enable mirroring, wait for the Ceph RBD Mirror daemons to start running:
kubectl -n rook-ceph get pod -l app=rook-ceph-rbd-mirror
In
KaaSCephClusterof both Ceph clusters where you want to enable mirroring, specify thespec.cephClusterSpec.pools.mirroring.modeparameter for allpoolsthat must be mirrored.Mirroring mode recommendations
Mirantis recommends using the
poolmode for mirroring. For thepoolmode, explicitly enable journaling for each image.To use the
imagemirroring mode, explicitly enable mirroring as described in the step 8.
spec: cephClusterSpec: pools: - name: image-hdd ... mirroring: mode: pool - name: volumes-hdd ... mirroring: mode: pool
Obtain the name of an external site to mirror with. On pools with mirroring enabled, the name is typically
ceph fsid:kubectl -n rook-ceph exec -it $(kubectl -n rook-ceph get pod -l \ "app=rook-ceph-tools" -o name rbd mirror pool info <mirroringPoolName> # or ceph fsid
Substitute
<mirroringPoolName>with the name of a pool to be mirrored.On an external site to mirror with, create a new bootstrap peer token. Execute the following command within the
ceph-toolspod CLI:kubectl -n rook-ceph exec -it $(kubectl -n rook-ceph get pod -l \ "app=rook-ceph-tools" -o jsonpath='{.items[0].metadata.name}') bash rbd mirror pool peer bootstrap create <mirroringPoolName> --site-name <siteName>
Substitute
<mirroringPoolName>with the name of a pool to be mirrored. In<siteName>, assign a label for the external Ceph cluster that will be used along with mirroring.For details, see Ceph documentation: Bootstrap peers.
In
KaaSCephClusteron the cluster that should mirror pools, specifyspec.cephClusterSpec.rbdMirror.peerswith the obtained peer and pools to mirror:spec: cephClusterSpec: rbdMirror: ... peers: - site: <siteName> token: <bootstrapPeer> pools: [<mirroringPoolName1>, <mirroringPoolName2>, ...]
Substitute
<siteName>with the label assigned to the external Ceph cluster,<bootstrapPeer>with the token obtained in the previous step, and<mirroringPoolName>with names of pools that have themirroring.modeparameter defined.For example:
spec: cephClusterSpec: rbdMirror: ... peers: - site: cluster-b token: <base64-string> pools: - images-hdd - volumes-hdd - special-pool-ssd
Verify that mirroring is enabled and each pool with
spec.cephClusterSpec.pools.mirroring.modedefined has an external peer site:kubectl -n rook-ceph exec -it $(kubectl -n rook-ceph get pod -l \ "app=rook-ceph-tools" -o jsonpath='{.items[0].metadata.name}') bash rbd mirror pool info <mirroringPoolName>
Substitute
<mirroringPoolName>with the name of a pool with mirroring enabled.If you have set the
imagemirroring mode in thepoolssection, explicitly enable mirroring for each image withrbdwithin the pool:Note
Execute the following command within the
ceph-toolspod withcephandrbdCLI.rbd mirror image enable <poolName>/<imageName> <imageMirroringMode>
Substitute
<poolName>with the name of a pool with theimagemirroring mode,<imageName>with the name of an image stored in the specified pool. Substitute<imageMirroringMode>with one of:journal- for mirroring to use the RBD journaling image feature to replicate the image contents. If the RBD journaling image feature is not yet enabled on the image, it will be enabled automatically.snapshot- for mirroring to use RBD image mirror-snapshots to replicate the image contents. Once enabled, an initial mirror-snapshot will automatically be created. To create additional RBD image mirror-snapshots, use the rbd command.
For details, see Ceph Documentation: Enable image mirroring.
Specify placement of Ceph cluster daemons¶
If you need to configure the placement of Rook daemons on nodes, you can add
extra values in the Cluster providerSpec section of the
ceph-controller Helm release.
The procedures in this section describe how to specify the placement of
rook-ceph-operator, rook-discover, and csi-rbdplugin.
To specify rook-ceph-operator placement:
On the management cluster, edit the
Clusterresource of the target managed cluster:kubectl -n <managedClusterProjectName> edit cluster
Add the following parameters to the
ceph-controllerHelm release values:spec: providerSpec: value: helmReleases: - name: ceph-controller values: rookOperatorPlacement: affinity: <rookOperatorAffinity> nodeSelector: <rookOperatorNodeSelector> tolerations: <rookOperatorTolerations>
<rookOperatorAffinity>is a key-value mapping that contains a valid Kubernetesaffinityspecification<rookOperatorNodeSelector>is a key-value mapping that contains a valid KubernetesnodeSelectorspecification<rookOperatorTolerations>is a list that contains valid Kubernetestolerationitems
Wait for some time and verify on a managed cluster that the changes have applied:
kubectl -n rook-ceph get deploy rook-ceph-operator -o yaml
To specify rook-discover and csi-rbdplugin placement simultaneously:
On the management cluster, edit the desired
Clusterresource:kubectl -n <managedClusterProjectName> edit cluster
Add the following parameters to the
ceph-controllerHelm release values:spec: providerSpec: value: helmReleases: - name: ceph-controller values: rookExtraConfig: extraDaemonsetLabels: <labelSelector>
Substitute
<labelSelector>with a valid Kubernetes label selector expression to place therook-discoverandcsi-rbdpluginDaemonSet pods.Wait for some time and verify on a managed cluster that the changes have applied:
kubectl -n rook-ceph get ds rook-discover -o yaml kubectl -n rook-ceph get ds csi-rbdplugin -o yaml
To specify rook-discover and csi-rbdplugin placement separately:
On the management cluster, edit the desired
Clusterresource:kubectl -n <managedClusterProjectName> edit cluster
If required, add the following parameters to the
ceph-controllerHelm release values:spec: providerSpec: value: helmReleases: - name: ceph-controller values: hyperconverge: nodeAffinity: csiplugin: <labelSelector1> rookDiscover: <labelSelector2>
Substitute
<labelSelectorX>with a valid Kubernetes label selector expression to place therook-discoverandcsi-rbdpluginDaemonSet pods. For example,"role=storage-node; discover=true".Wait for some time and verify on the managed cluster that the changes have applied:
kubectl -n rook-ceph get ds rook-discover -o yaml kubectl -n rook-ceph get ds csi-rbdplugin -o yaml
Migrate Ceph pools from one failure domain to another¶
The document describes how to change the failure domain of an already deployed Ceph cluster.
Note
This document focuses on changing the failure domain from a smaller
to wider one, for example, from host to rack. Using the same
instruction, you can move the failure domain from a wider to smaller scale.
Caution
Data movement implies the Ceph cluster rebalancing that may impact cluster performance, depending on the cluster size.
High-level overview of the procedure includes the following steps:
Set correct labels on the nodes.
Create the new bucket hierarchy.
Move nodes to new buckets.
Modify the CRUSH rules.
Add the manual changes to the
KaaSCephClusterspec.Scale the Ceph controllers.
Prerequisites¶
Verify that the Ceph cluster has enough space for multiple copies of data to migrate. Mirantis highly recommends that the Ceph cluster has a minimum of 25% of free space for the procedure to succeed.
Note
The migration procedure implies data movement and optional modification of CRUSH rules that cause a large amount of data (depending on the cluster size) to be first copied to a new location in the Ceph cluster before data removal.
Create a backup of the current
KaaSCephClusterobject from the managed namespace of the management cluster:kubectl -n <managedClusterProject> get kaascephcluster -o yaml > kcc-backup.yaml
Substitute
<managedClusterProject>with the corresponding managed cluster namespace of the management cluster.In the
rook-ceph-toolspod on a managed cluster, obtain a backup of the CRUSH map:ceph osd getcrushmap -o /tmp/crush-map-orig crushtool -d /tmp/crush-map-orig -o /tmp/crush-map-orig.txt
Migrate Ceph pools¶
This procedure contains an example of moving failure domains of all pools from
host to rack. Using the same instruction, you can migrate pools from
other types of failure domains, migrate pools separately, and so on.
To migrate Ceph pools from one failure domain to another:
Set the required CRUSH topology in the
KaaSCephClusterobject for each defined node. For details on thecrushparameter, see Node parameters.Setting the CRUSH topology to each node causes the Ceph Controller to set proper Kubernetes labels on the nodes.
Example of adding the
rackCRUSH topology key for each node in thenodessectionspec: cephClusterSpec: nodes: machine1: crush: rack: rack-1 machine2: crush: rack: rack-1 machine3: crush: rack: rack-2 machine4: crush: rack: rack-2 machine5: crush: rack: rack-3 machine6: crush: rack: rack-3
On the managed cluster, verify that the required buckets and bucket types are present in the Ceph hierarchy:
Enter the
ceph-toolspod:kubectl -n rook-ceph exec -it deploy/rook-ceph-tools -- bash
Verify that the required bucket type is present by default:
ceph osd getcrushmap -o /tmp/crush-map crushtool -d /tmp/crush-map -o /tmp/crush-map.txt cat /tmp/crush-map.txt # Look for the section named → “# types”
Example of system response:
# types type 0 osd type 1 host type 2 chassis type 3 rack type 4 row type 5 pdu type 6 pod type 7 room type 8 datacenter type 9 zone type 10 region type 11 root
Verify that the buckets with the required bucket type are present:
cat /tmp/crush-map.txt # Look for the section named → “# buckets”
Example of system response of an existing
rackbucket:# buckets rack rack-1 { id -15 id -16 class hdd # weight 0.00000 alg straw2 hash 0 }
If the required buckets are not created, create new ones with the required bucket type:
ceph osd crush add-bucket <bucketName> <bucketType> root=default
For example:
ceph osd crush add-bucket rack-1 rack root=default ceph osd crush add-bucket rack-2 rack root=default ceph osd crush add-bucket rack-3 rack root=default
Exit the
ceph-toolspod.
Optional. Order buckets as required:
On the managed cluster, add the first Ceph CRUSH smaller bucket to its respective wider bucket:
kubectl -n rook-ceph exec -it deploy/rook-ceph-tools -- bash ceph osd crush move <smallerBucketName> <bucketType>=<widerBucketName>
Substitute the following parameters:
<smallerBucketName>with the name of the smaller bucket, for example host name<bucketType>with the required bucket type, for examplerack<widerBucketName>with the name of the wider bucket, for example rack name
For example:
ceph osd crush move kaas-node-1 rack=rack-1 root=default
Warning
Mirantis highly recommends moving one bucket at a time.
For more details, refer to official Ceph documentation: CRUHS Maps: Moving a bucket.
After the bucket is moved to the new location in the CRUSH hierarchy, verify that no data rebalancing occurs:
ceph -sCaution
Wait for rebalancing to complete before proceeding to the next step.
Add the remaining Ceph CRUSH smaller buckets to their respective wider buckets one by one.
Scale the Ceph Controller and Rook Operator deployments to
0replicas:kubectl -n ceph-lcm-mirantis scale deploy --all --replicas 0 kubectl -n rook-ceph scale deploy rook-ceph-operator --replicas 0
On the managed cluster, manually modify the CRUSH rules for Ceph pools to enable data placement on a new failure domain:
Enter the
ceph-toolspod:kubectl -n rook-ceph exec -it deploy/rook-ceph-tools -- bash
List the CRUSH rules and erasure code profiles for the pools:
ceph osd pool ls detail
Example output
pool 1 'mirablock-k8s-block-hdd' replicated size 2 min_size 1 crush_rule 9 object_hash rjenkins pg_num 32 pgp_num 32 autoscale_mode on last_change 1193 lfor 0/0/85 flags hashpspool,selfmanaged_snaps stripe_width 0 application rbd read_balance_score 1.31 pool 2 '.mgr' replicated size 2 min_size 1 crush_rule 1 object_hash rjenkins pg_num 1 pgp_num 1 autoscale_mode on last_change 70 flags hashpspool stripe_width 0 pg_num_max 32 pg_num_min 1 application mgr read_balance_score 6.06 pool 3 'openstack-store.rgw.otp' replicated size 2 min_size 1 crush_rule 11 object_hash rjenkins pg_num 8 pgp_num 8 autoscale_mode on last_change 1197 flags hashpspool stripe_width 0 pg_num_min 8 application rook-ceph-rgw read_balance_score 2.27 pool 4 'openstack-store.rgw.meta' replicated size 2 min_size 1 crush_rule 12 object_hash rjenkins pg_num 8 pgp_num 8 autoscale_mode on last_change 1197 flags hashpspool stripe_width 0 pg_num_min 8 application rook-ceph-rgw read_balance_score 1.50 pool 5 'openstack-store.rgw.log' replicated size 2 min_size 1 crush_rule 10 object_hash rjenkins pg_num 8 pgp_num 8 autoscale_mode on last_change 1197 flags hashpspool stripe_width 0 pg_num_min 8 application rook-ceph-rgw read_balance_score 3.00 pool 6 'openstack-store.rgw.buckets.non-ec' replicated size 2 min_size 1 crush_rule 13 object_hash rjenkins pg_num 8 pgp_num 8 autoscale_mode on last_change 1197 flags hashpspool stripe_width 0 pg_num_min 8 application rook-ceph-rgw read_balance_score 1.50 pool 7 'openstack-store.rgw.buckets.index' replicated size 2 min_size 1 crush_rule 15 object_hash rjenkins pg_num 8 pgp_num 8 autoscale_mode on last_change 1197 flags hashpspool stripe_width 0 pg_num_min 8 application rook-ceph-rgw read_balance_score 2.25 pool 8 '.rgw.root' replicated size 2 min_size 1 crush_rule 14 object_hash rjenkins pg_num 8 pgp_num 8 autoscale_mode on last_change 1197 flags hashpspool stripe_width 0 pg_num_min 8 application rook-ceph-rgw read_balance_score 3.75 pool 9 'openstack-store.rgw.control' replicated size 2 min_size 1 crush_rule 16 object_hash rjenkins pg_num 8 pgp_num 8 autoscale_mode on last_change 1197 flags hashpspool stripe_width 0 pg_num_min 8 application rook-ceph-rgw read_balance_score 3.00 pool 10 'other-hdd' replicated size 2 min_size 1 crush_rule 19 object_hash rjenkins pg_num 32 pgp_num 32 autoscale_mode on last_change 1179 lfor 0/0/85 flags hashpspool,selfmanaged_snaps stripe_width 0 application rbd read_balance_score 1.69 pool 11 'openstack-store.rgw.buckets.data' erasure profile openstack-store.rgw.buckets.data_ecprofile size 3 min_size 2 crush_rule 18 object_hash rjenkins pg_num 32 pgp_num 32 autoscale_mode on last_change 1198 lfor 0/0/86 flags hashpspool,ec_overwrites stripe_width 8192 application rook-ceph-rgw pool 12 'vms-hdd' replicated size 2 min_size 1 crush_rule 21 object_hash rjenkins pg_num 256 pgp_num 256 autoscale_mode on last_change 1182 lfor 0/0/95 flags hashpspool,selfmanaged_snaps stripe_width 0 target_size_ratio 0.4 application rbd read_balance_score 1.24 pool 13 'volumes-hdd' replicated size 2 min_size 1 crush_rule 23 object_hash rjenkins pg_num 64 pgp_num 64 autoscale_mode on last_change 1185 lfor 0/0/89 flags hashpspool,selfmanaged_snaps stripe_width 0 target_size_ratio 0.2 application rbd read_balance_score 1.31 pool 14 'backup-hdd' replicated size 2 min_size 1 crush_rule 25 object_hash rjenkins pg_num 32 pgp_num 32 autoscale_mode on last_change 1188 lfor 0/0/90 flags hashpspool,selfmanaged_snaps stripe_width 0 target_size_ratio 0.1 application rbd read_balance_score 2.06 pool 15 'images-hdd' replicated size 2 min_size 1 crush_rule 27 object_hash rjenkins pg_num 32 pgp_num 32 autoscale_mode on last_change 1191 lfor 0/0/90 flags hashpspool,selfmanaged_snaps stripe_width 0 target_size_ratio 0.1 application rbd read_balance_score 1.50
For each replicated Ceph pool:
Obtain the current CRUSH rule name:
ceph osd crush rule dump <oldCrushRuleName>
Create a new CRUSH rule with the required bucket type using the same root, device class, and new bucket type:
ceph osd crush rule create-replicated <newCrushRuleName> <root> <bucketType> <deviceClass>
For example:
ceph osd crush rule create-replicated images-hdd-rack default rack hdd
For more details, refer to official Ceph documentation: CRUSH Maps: Creating a rule for a replicated pool.
Apply a new crush rule to the Ceph pool:
ceph osd pool set <poolName> crush_rule <newCrushRuleName>
For example:
ceph osd pool set images-hdd crush_rule images-hdd-rack
Wait for data to be rebalanced after moving the Ceph pool under the new failure domain (bucket type) by monitoring Ceph health:
ceph -sCaution
Update the following Ceph pool only after data rebalancing completes for the current Ceph pool.
Verify that the old CRUSH rule is not used anymore:
ceph osd pool ls detail
The rule ID is located in the CRUSH map and must match the rule ID in the output of ceph osd dump.
Remove the old unused CRUSH rule and rename the new one to the original name:
ceph osd crush rule rm <oldCrushRuleName> ceph osd crush rule rename <newCrushRuleName> <oldCrushRuleName>
For each erasure-coded Ceph pool:
Note
Erasure-coded pools require different number of buckets to store data. Instead of the number of replicas in replicated pools, erasure-coded pools require the
coding chunks + data chunksnumber of buckets existing in the Ceph cluster. For example, if an erasure-coded pool has 2 coding chunks and 2 data chunks configured, then the pool requires 4 different buckets, for example, 4 racks, to store data.Obtain the current parameters of the erasure-coded profile:
ceph osd erasure-code-profile get <ecProfile>
In the profile, add the new bucket type as the failure domain using the
crush-failure-domainparameter:ceph osd erasure-code-profile set <ecProfile> k=<int> m=<int> crush-failure-domain=<bucketType> crush-device-class=<deviceClass>
Create a new CRUSH rule in the profile:
ceph osd crush rule create-erasure <newEcCrushRuleName> <ecProfile>
Apply the new CRUSH rule to the pool:
ceph osd pool set <poolName> crush_rule <newEcCrushRuleName>
Wait for data to be rebalanced after moving the Ceph pool under the new failure domain (bucket type) by monitoring Ceph health:
ceph -sCaution
Update the following Ceph pool only after data rebalancing completes for the current Ceph pool.
Verify that the old CRUSH rule is not used anymore:
ceph osd pool ls detail
The rule ID is located in the CRUSH map and must match the rule ID in the output of ceph osd dump.
Remove the old unused CRUSH rule and rename the new one to the original name:
ceph osd crush rule rm <oldCrushRuleName> ceph osd crush rule rename <newCrushRuleName> <oldCrushRuleName>
Note
New erasure-coded profiles cannot be renamed, so they will not be removed automatically during pools cleanup. Remove them manually, if needed.
Exit the
ceph-toolspod.
In the management cluster, update the
KaaSCephClusterobject by setting thefailureDomain: rackparameter for each pool. The configuration from the Rook perspective must match the manually created configuration. For example:spec: cephClusterSpec: pools: - name: images ... failureDomain: rack - name: volumes ... failureDomain: rack ... objectStorage: rgw: dataPool: failureDomain: rack ... metadataPool: failureDomain: rack ...
Monitor the Ceph cluster health and wait until rebalancing is completed:
ceph -sExample of a successful system response:
HEALTH_OK
Scale back the Ceph Controller and Rook Operator deployments:
kubectl -n ceph-lcm-mirantis scale deploy --all --replicas 3 kubectl -n rook-ceph scale deploy rook-ceph-operator --replicas 1
Enable periodic Ceph performance testing¶
TechPreview
Warning
Performance testing affects the overall Ceph cluster performance. Do not run it unless you are sure that user load will not be affected.
This section describes how to configure periodic Ceph performance testing using Kubernetes batch or cron jobs that execute a fio process in a separate container with a connection to the Ceph cluster. The test results can then be stored in a persistent volume attached to the container.
Ceph performance testing is managed by the KaaSCephOperationRequest CR that
creates separate CephPerfTestRequest requests to handle the test run. Once
you configure the perfTest section of the KaaSCephOperationRequest
spec, it propagates to CephPerfTestRequest on the managed cluster in the
ceph-lcm-mirantis namespace. You can create a performance test for a single
or scheduled runs.
Create a Ceph performance test request¶
TechPreview
Warning
Performance testing affects the overall Ceph cluster performance. Do not run it unless you are sure that user load will not be affected.
This section describes how to create a Ceph performance test request through
the KaaSCephOperationRequest CR.
To create a Ceph performance test request:
Create an RBD image with the required parameters. For example, run the following command in
ceph-tools-containerto allow execution of theperftestexample below on a managed cluster:kubectl exec -ti -n rook-ceph <ceph-tools-pod> -- bash rbd create <pool_name>/<image_name> --size 10G
Substitute
<ceph-tools-pod>with theceph-toolsPod ID,<pool_name>and<image_name>with pool and image names, and specify the size. In the example below,mirablock-k8s-block-hddis used as pool name andtestsas image name:kubectl exec -ti -n rook-ceph rook-ceph-tools-94985cd9f-tjl29 -- bash rbd create mirablock-k8s-block-hdd/tests --size 10G
Create a YAML template for the
KaaSCephOperationRequestCR. For details, see KaaSCephOperationRequest CR perftest specification.kubectl apply -f <example_file_name>.yaml
Example of the
KaaSCephOperationRequestresourceapiVersion: kaas.mirantis.com/v1alpha1 kind: KaaSCephOperationRequest metadata: name: test-perf-req namespace: managed-ns spec: kaasCephCluster: name: ceph-kaas-managed namespace: managed-ns perfTest: parameters: - --ioengine=rbd - --pool=mirablock-k8s-block-hdd - --rbdname=tests - --name=single_perftest - --rw=randrw:16k - --rwmixread=40 - --bs=4k - --size=500M - --iodepth=32 - --numjobs=8 - --group_reporting - --direct=1 - --fsync=32 - --buffered=0 - --exitall
Review the
KaaSCephOperationRequeststatus. For details, see KaaSCephOperationRequest perftest status.kubectl get kaascephoperationrequest test-managed-req -n managed-ns
Example of system response:
NAME KAASCEPHCLUSTER CLUSTER AGE PHASE MESSAGE test-perf-req ceph-kaas-managed kaas-managed 20m Completed
Review the
CephPerfTestRequeststatus on the managed cluster.kubectl get cephperftestrequest -n ceph-lcm-mirantis
Example of system response:
NAME AGE PHASE START TIME DURATION JOB STATUS SCHEDULE test-perf-req 55m Finished 2022-06-17T09:29:57Z 5m53s Completed
Review the performance test result by inspecting logs for the corresponding job Pod on the managed cluster:
kubectl --kubeconfig <managedKubeconfig> -n rook-ceph logs -l app=ceph-perftest,perftest=<name>
Substitute
<managedKubeconfig>with the managed clusterkubeconfigand<name>with theKaaSCephOperationRequestmetadata.name, for example,test-perf-req.Optional. Remove the
KaaSCephOperationRequest. Removal ofKaaSCephOperationRequestalso removes theCephPerfTestRequestCR propagated to the managed cluster.
KaaSCephOperationRequest CR perftest specification¶
TechPreview
This section describes the KaaSCephOperationRequest CR specification used
to automatically create a CephPerfTestRequest request. For the procedure
workflow, see Enable periodic Ceph performance testing.
Parameter |
Description |
|---|---|
|
Describes the definition for the |
|
Defines spec:
kaasCephCluster:
name: ceph-kaas-mgmt
namespace: default
|
|
Defines the cluster on which the spec:
k8sCluster:
name: kaas-mgmt
namespace: default
If you omit this parameter, |
Parameter |
Description |
|---|---|
|
A list of command arguments for a performance test execution. For all available parameters, see fio documentation. Note Performance test results will be saved on a PVC if the test
run parameters contain an argument to save to a file. Otherwise, test
results will be saved only as Pod logs. For example, for the default
|
|
Optional. Entrypoint command to run performance test in the container. If the performance image is updated, you may also update the command. By default, equals the image entry point. |
|
Container image to use for jobs. By default, |
|
Configuration of the performance test runs as periodic jobs. Leave empty if a single run is required. For details, see Ceph performance periodic parameters. |
|
Option that enables saving of the performance test results on a PVC. Contains the following fields:
|
Parameter |
Description |
|---|---|
|
Required. Schedule in base cron format. For example, |
|
Pause CronJob scheduling to prevent performance test execution. Only for future scheduling. |
|
Number of runs to keep in history. Supported only by keeping old run Pods with their outputs. |
Example of KaaSCephOperationRequest
apiVersion: kaas.mirantis.com/v1alpha1
kind: KaaSCephOperationRequest
metadata:
name: test-managed-req
namespace: managed-ns
spec:
kaasCephCluster:
name: ceph-cluster-managed-cluster
namespace: managed-ns
perfTest:
parameters:
- --ioengine=rbd
- --pool=mirablock-k8s-block-hdd
- --rbdname=tests
- --name=single_perftest
- --rw=randrw:16k
- --rwmixread=40
- --bs=4k
- --size=500M
- --iodepth=32
- --numjobs=8
- --group_reporting
- --direct=1
- --fsync=32
- --buffered=0
- --exitall
KaaSCephOperationRequest perftest status¶
TechPreview
This section describes the status.perfTestStatus fields of the
KaaSCephOperationRequest CR that you can use to check the status of a Ceph
performance test request.
Note
Performance test results will be saved on PVC if the test run
parameters contain the saveResultOnPvc option. Otherwise, test
results will be saved only as Pod logs. For details, see
Ceph performance test parameters.
Parameter |
Description |
|---|---|
|
Describes the status of the current |
Parameter |
Description |
|---|---|
|
Describes the current request phase:
|
|
The last start time of the performance test execution. |
|
The duration of the last successful performance test. |
|
The execution status of the last performance test. |
|
Issues or warnings found during the performance test run. |
|
Location of the performance test result. Contains the following fields:
|
|
History of statuses and timings for cron jobs:
|
Example of status.perfTestStatus
status:
perfTestStatus:
lastDurationTime: 5m53s
lastJobStatus: Completed
lastStartTime: "2022-06-17T09:29:57Z"
phase: Finished
results:
storedOnPvc:
pvcName: test-perf-req-cephperftest
pvcNamespace: rook-ceph
phase: Completed
See also
StackLight operations¶
The section covers the StackLight management aspects.
Configure StackLight¶
This section describes how to configure StackLight in your Mirantis OpenStack on Kubernetes deployment and includes the description of StackLight parameters and their verification.
StackLight configuration procedure¶
This section describes the StackLight configuration workflow.
To configure StackLight:
Download your management cluster
kubeconfig:Log in to the Container Cloud web UI with the
m:kaas:namespace@operatororm:kaas:namespace@writerpermissions.Switch to the required project using the Switch Project action icon located on top of the main left-side navigation panel.
Expand the menu of the tab with your user name.
Click Download kubeconfig to download
kubeconfigof your management cluster.Log in to any local machine with
kubectlinstalled.Copy the downloaded
kubeconfigto this machine.
Run one of the following commands:
For a management cluster:
kubectl --kubeconfig <mgmtClusterKubeconfigPath> edit -n default cluster <mgmtClusterName>
For a managed cluster:
kubectl --kubeconfig <mgmtClusterKubeconfigPath> edit -n <managedClusterProjectName> cluster <managedClusterName>
In the following section of the opened manifest, configure the required StackLight parameters as described in StackLight configuration parameters.
spec: providerSpec: value: helmReleases: - name: stacklight values:
StackLight configuration parameters¶
This section describes the StackLight configuration keys that you can specify
in the values section to change StackLight settings as required. Prior to
making any changes to StackLight configuration, perform the steps described in
StackLight configuration procedure.
After changing StackLight configuration, verify the changes as described in
Verify StackLight after configuration.
Important
Some parameters are marked as mandatory. Failure to specify values for such parameters causes the Admission Controller to reject cluster creation.
This section describes the OpenStack-related StackLight configuration keys. For MOSK cluster configuration keys, see MOSK cluster configuration parameters.
Key |
Description |
Example values |
|---|---|---|
|
Enables Gnocchi monitoring. Set to |
|
Key |
Description |
Example values |
|---|---|---|
|
Enables Ironic monitoring. Set to |
|
Key |
Description |
Example values |
|---|---|---|
|
Enables OpenStack monitoring. Set to |
|
|
Defines the namespace within which the OpenStack virtualized control
plane is installed. Set to |
|
Key |
Description |
Example values |
|---|---|---|
|
Defines the RabbitMQ credentials to use if credentials discovery is disabled or some required parameters were not found during the discovery. |
credentialsConfig:
username: "stacklight"
password: "stacklight"
host: "rabbitmq.openstack.svc"
queue: "notifications"
vhost: "openstack"
|
|
Enables the credentials discovery to obtain the username and password from the secret object. |
credentialsDiscovery:
enabled: true
namespace: openstack
secretName: os-rabbitmq-user-credentials
|
Key |
Description |
Example values |
|---|---|---|
|
External FQDN used to communicate with OpenStack services for
certificates monitoring. The option is deprecated, use
|
|
|
External FQDN used to communicate with OpenStack services. Used for
certificates monitoring. Set to |
|
|
Defines whether to verify the trust chain of the OpenStack endpoint SSL certificates during monitoring. |
insecure:
internal: true
external: false
|
Key |
Description |
Example values |
|---|---|---|
|
Specifies the OpenStack credentials to use if the credentials discovery is disabled or some required parameters were not found during the discovery. |
credentialsConfig:
identityEndpoint: "" # "http://keystone-api.openstack.svc:5000/v3"
domain: "" # "default"
password: "" # "workshop"
project: "" # "admin"
region: "" # "RegionOne"
username: "" # "admin"
|
|
Enables the credentials discovery to obtain all required parameters from the secret object. |
credentialsDiscovery:
enabled: true
namespace: openstack
secretName: keystone-keystone-admin
|
|
Specifies the interval of metrics gathering from the OpenStack API. Set
to |
|
|
Enables or disables the server certificate chain and host name
verification. Set to |
|
|
Enables or disables HTTP probes for public endpoints from the OpenStack
service catalog. Set to |
|
Key |
Description |
Example values |
|---|---|---|
|
Enables Tungsten Fabric monitoring. Since MOSK 23.1, the parameter is set
to Before MOSK 23.1, the parameter is set to
|
|
|
Available since MOSK 23.3. Defines the timeout of
the |
tungstenFabricMonitoring:
exportersTimeout: "5s"
|
|
Available since MOSK 24.1. Enables or disables monitoring of the Tungsten Fabric analytics services. In MOSK 24.1, defaults to Since MOSK 24.2, the default value is set automatically based on the real state of the Tungsten Fabric analytics services (enabled or disabled) in the Tungsten Fabric cluster. |
|
This section describes the MOSK cluster StackLight configuration keys. For OpenStack cluster configuration keys, see OpenStack cluster configuration parameters.
Key |
Description |
Example values |
|---|---|---|
|
Defines custom alerts. Also, modifies or disables existing alert configurations. For the list of predefined alerts, see StackLight alerts. While adding or modifying alerts, follow the Alerting rules. |
customAlerts:
# To add a new alert:
- alert: ExampleAlert
annotations:
description: Alert description
summary: Alert summary
expr: example_metric > 0
for: 5m
labels:
severity: warning
# To modify an existing alert expression:
- alert: AlertmanagerFailedReload
expr: alertmanager_config_last_reload_successful == 5
# To disable an existing alert:
- alert: TargetDown
enabled: false
An optional field |
Key |
Description |
Example values |
|---|---|---|
|
Enables or disables Alerta. Using the Alerta web UI, you can view the
most recent or watched alerts, group, and filter alerts. Set to |
|
On the managed clusters with limited Internet access, proxy is required for StackLight components that use HTTP and HTTPS and are disabled by default but need external access if enabled, for example, for the Salesforce integration and Alertmanager notifications external rules.
Key |
Description |
Example values |
|---|---|---|
|
Provides a generic template for notifications receiver configurations. For a list of supported receivers, see Prometheus Alertmanager documentation: Receiver. |
For example, to enable notifications to OpsGenie: alertmanagerSimpleConfig:
genericReceivers:
- name: HTTP-opsgenie
enabled: true # optional
opsgenie_configs:
- api_url: "https://example.app.eu.opsgenie.com/"
api_key: "secret-key"
send_resolved: true
|
|
Provides a template for notifications route configuration. For details, see Prometheus Alertmanager documentation: Route. |
genericRoutes:
- receiver: HTTP-opsgenie
enabled: true # optional
matchers:
severity=~"major|critical"
continue: true
|
|
Disables or enables alert inhibition rules. If enabled, Alertmanager decreases alert noise by suppressing dependent alerts notifications to provide a clearer view on the cloud status and simplify troubleshooting. Enabled by default. For details, see Alert dependencies. For details on inhibition rules, see Prometheus documentation. |
|
Key |
Description |
Example values |
|---|---|---|
|
Enables or disables Alertmanager integration with email. Set to
|
|
|
Defines the notification parameters for Alertmanager integration with email. For details, see Prometheus Alertmanager documentation: Email configuration. |
email:
enabled: false
send_resolved: true
to: "to@test.com"
from: "from@test.com"
smarthost: smtp.gmail.com:587
auth_username: "from@test.com"
auth_password: password
auth_identity: "from@test.com"
require_tls: true
|
|
Defines the route for Alertmanager integration with email. For details, see Prometheus Alertmanager documentation: Route. |
route:
matchers: []
routes: []
|
On the managed clusters with limited Internet access, proxy is required for StackLight components that use HTTP and HTTPS and are disabled by default but need external access if enabled. The Microsoft Teams integration depends on the Internet access through HTTPS.
Key |
Description |
Example values |
|---|---|---|
|
Enables or disables Alertmanager integration with Microsoft Teams.
Requires a set up Microsoft Teams channel and a channel connector. Set
to |
|
|
Defines the URL of an Incoming Webhook connector of a Microsoft Teams channel. For details about channel connectors, see Microsoft documentation. |
|
|
Defines the notifications route for Alertmanager integration with MS Teams. For details, see Prometheus Alertmanager documentation: Route. |
route:
matchers: []
routes: []
|
On the managed clusters with limited Internet access, proxy is required for StackLight components that use HTTP and HTTPS and are disabled by default but need external access if enabled. The Salesforce integration depends on the Internet access through HTTPS.
Key |
Description |
Example values |
|---|---|---|
|
Unique cluster identifier
The |
Do not modify |
|
Enables or disables Alertmanager integration with Salesforce using the
|
|
|
Defines the Salesforce parameters and credentials for integration with Alertmanager. |
auth:
url: "<SF instance URL>"
username: "<SF account email address>"
password: "<SF password>"
environment_id: "<Cloud identifier>"
organization_id: "<Organization identifier>"
sandbox_enabled: "<Set to true or false>"
|
|
Defines the notifications route for Alertmanager integration with Salesforce. For details, see Prometheus Alertmanager documentation: Route. |
route:
matchers:
- severity="critical"
routes: []
Note By default, only |
|
Enables or disables feed update in Salesforce. To save API calls, this
parameter is set to |
|
|
Enables or disables links to the Prometheus web UI in alerts sent to
Salesforce. To simplify troubleshooting, set to |
|
Caution
Prior to configuring the integration with ServiceNow, perform the following prerequisite steps using the ServiceNow documentation of the required version.
In a new or existing Incident table, add the Alert ID field as described in Add fields to a table. To avoid alerts duplication, select Unique.
Create an Access Control List (ACL) with read/write permissions for the Incident table as described in Securing table records.
Key |
Description |
Example values |
|---|---|---|
|
Enables or disables Alertmanager integration with ServiceNow. Set to
|
|
|
Defines the ServiceNow parameters and credentials for integration with Alertmanager:
|
serviceNow:
enabled: true
incident_table: "incident"
api_version: "v1"
alert_id_field: "u_alert_id"
auth:
instance: "https://dev00001.service-now.com"
username: "testuser"
password: "testpassword"
|
On the managed clusters with limited Internet access, proxy is required for StackLight components that use HTTP and HTTPS and are disabled by default but need external access if enabled. The Slack integration depends on the Internet access through HTTPS.
Key |
Description |
Example values |
|---|---|---|
|
Enables or disables Alertmanager integration with Slack. For
details, see Prometheus Alertmanager documentation: Slack configuration.
Set to |
|
|
Defines the Slack webhook URL. |
|
|
Defines the Slack channel or user to send notifications to. |
|
|
Defines the notifications route for Alertmanager integration with Slack. For details, see Prometheus Alertmanager documentation: Route. |
route:
matchers: []
routes: []
|
Key |
Description |
Example values |
|---|---|---|
|
Enables or disables the |
|
For internal StackLight use only
Key |
Description |
Example values |
|---|---|---|
|
Specifies the size limit of the incoming data length in bytes for the
Telemeter client. Defaults to |
|
Key |
Description |
Example values |
|---|---|---|
|
Specifies the approximate expected cluster size. Set to Caution Since Container Cloud 2.28.0 (Cluster releases 17.3.0 and
16.3.0), The values differ by the OpenSearch and Prometheus resource limits:
|
|
Key |
Description |
Example values |
|---|---|---|
|
Removed in Container Cloud 2.27.0 (Cluster releases 17.2.0 and 16.2.0). Disables Grafana Image Renderer. For example, for resource-limited environments. Enabled by default. |
|
|
Defines the home dashboard. Set to |
|
Key |
Description |
Example values |
|---|---|---|
|
Enables or disables StackLight multiserver mode. For details, see
StackLight database modes in Container Cloud Reference Architecture:
StackLight deployment architecture.
On managed clusters, set to |
|
Available since MCC 2.25.1 (Cluster releases 17.0.1 and 16.0.1)
Key |
Description |
Example values |
|---|---|---|
|
Enables or disables the Kubernetes Network Policy resource that allows
controlling network connections to and from Pods deployed in the
For the list of network policy rules, refer to StackLight rules for Kubernetes network policies. Customization of network policies is not supported. |
|
Key |
Description |
Example values |
|---|---|---|
|
Kubernetes tolerations to add to all StackLight components. |
default:
- key: "com.docker.ucp.manager"
operator: "Exists"
effect: "NoSchedule"
|
|
Defines Kubernetes tolerations (overrides the default ones) for any StackLight component. |
component:
# elasticsearch:
opensearch:
- key: "com.docker.ucp.manager"
operator: "Exists"
effect: "NoSchedule"
postgresql:
- key: "node-role.kubernetes.io/master"
operator: "Exists"
effect: "NoSchedule"
|
Available since MCC 2.25.0 (Cluster releases 17.0.0 and 16.0.0)
Key |
Description |
Example values |
|---|---|---|
|
Limits the number of namespaces for Pods log collection. Enabled by default with the following list of monitored Kubernetes namespaces:
Kubernetes namespaces monitored by default
|
|
|
Adds extra namespaces to collect Kubernetes Pod logs from. Requires
|
logging:
namespaceFiltering:
logs:
enabled: true
extraNamespaces:
- custom-ns-1
|
|
Limits the number of namespaces for Kubernetes events collection.
Disabled by default due to sysdig scanner present on some
MOSK clusters and due to cluster-scoped objects
producing events by default to the |
|
|
Adds extra namespaces to collect Kubernetes events from. Requires
|
logging:
namespaceFiltering:
events:
enabled: true
extraNamespaces:
- custom-ns-1
|
Key |
Description |
Example values |
|---|---|---|
|
Defines the log verbosity level for all StackLight components if not
defined using |
|
|
Defines (overrides the |
component:
kubeStateMetrics: ""
prometheusAlertManager: ""
prometheusBlackboxExporter: ""
prometheusNodeExporter: ""
prometheusServer: ""
alerta: ""
alertmanagerWebhookServicenow: ""
elasticsearchCurator: ""
postgresql: ""
prometheusEsExporter: ""
sfNotifier: ""
sfReporter: ""
fluentd: ""
# fluentdElasticsearch ""
fluentdLogs: ""
telemeterClient: ""
telemeterServer: ""
tfControllerExporter: ""
tfVrouterExporter: ""
telegrafDs: ""
telegrafS: ""
# elasticsearch: ""
opensearch: ""
# kibana: ""
grafana: ""
opensearchDashboards: ""
metricbeat: ""
prometheusMsTeams: ""
|
Key |
Description |
Example values |
|---|---|---|
|
Enables or disables the StackLight logging stack. For details about the
logging components, see Container Cloud Reference Architecture: StackLight
Deployment architecture.
Set to |
|
|
Removed in Container Cloud 2.26.0 (Cluster releases 17.1.0 and 16.1.0).
Sets the least important level of log messages to send to OpenSearch.
Requires The default logging level is Note The |
|
|
Allows configuring OpenSearch queries for the data present in OpenSearch. Prometheus Elasticsearch Exporter then queries the OpenSearch database and exposes such metrics in the Prometheus format. For details, see Create logs-based metrics. Includes the following parameters:
|
For usage example, see Create logs-based metrics. |
|
Removed in Container Cloud 2.26.0 (Cluster releases 17.1.0 and 16.1.0). Specifies the retention time per index. Includes the following parameters:
The allowed values include integers (days) and numbers with suffixes: y, m, w, d, h, including capital letters. |
logging:
retentionTime:
logstash: 3
events: "2w"
notifications: "1M"
|
Available since MCC 2.25.0 (Cluster releases 17.0.0 and 16.0.0)
Key |
Description |
Example values |
|---|---|---|
|
Enforces 32 GB of heap size, unless the defined memory limit allows using
50 GB of heap. Requires |
logging:
enforceOopsCompression: true
|
Available since MCC 2.23.0 (Cluster release 11.7.0)
Key |
Description |
Example values |
|---|---|---|
|
Specifies external Elasticsearch, OpenSearch, and syslog destinations
as |
logging:
externalOutputs:
elasticsearch:
# disabled: false
type: elasticsearch
level: info
plugin_log_level: info
tag_exclude: '{fluentd-logs,systemd}'
host: elasticsearch-host
port: 9200
logstash_date_format: '%Y.%m.%d'
logstash_format: true
logstash_prefix: logstash
...
buffer:
# disabled: false
chunk_limit_size: 16m
flush_interval: 15s
flush_mode: interval
overflow_action: block
...
opensearch:
disabled: true
type: opensearch
...
|
Available since MCC 2.23.0 (Cluster release 11.7.0)
Key |
Description |
Example values |
|---|---|---|
|
Specifies authentication secret mounts for external log destinations.
Requires
|
Secret mount configuration: logging:
externalOutputSecretMounts:
- secretName: elasticsearch-certs
mountPath: /tmp/elasticsearch-certs
defaultMode: 420
- secretName: opensearch-certs
mountPath: /tmp/opensearch-certs
Elasticsearch configuration for the above secret mount: logging:
externalOutputs:
elasticsearch:
...
ca_file: /tmp/elasticsearch-certs/ca.pem
client_cert: /tmp/elasticsearch-certs/client.pem
client_key: /tmp/elasticsearch-certs/client.key
client_key_pass: password
|
Deprecated since MCC 2.23.0 (Cluster release 11.7.0)
Note
Since Container Cloud 2.23.0 (Cluster release 11.7.0),
logging.syslog is deprecated for the sake of
logging.externalOutputs. For details, see
Logging to external outputs.
Key |
Description |
Example values |
|---|---|---|
|
Enables or disables remote logging to syslog. Disabled by default.
Requires |
|
|
Specifies the remote syslog host. |
|
|
Removed in Container Cloud 2.26.0 (Cluster releases 17.1.0 and 16.1.0). Specifies logging level for the syslog output. |
|
|
Specifies the remote syslog port. |
|
|
Defines the packet size in bytes for the syslog logging output. Set to
|
|
|
Specifies the remote syslog protocol. Set to |
|
|
Optional. Disabled by default. Enables or disables TLS. Use TLS only for the TCP protocol. TLS will not be enabled if you set a protocol other than TCP. |
|
|
Optional. Configures TLS verification. |
|
|
Defines how to pass the certificate.
|
certificate:
secret: ""
hostPath: "/etc/ssl/certs/ca-bundle.pem"
|
tag_exclude (string)Since MCC 2.23.0 (11.7.0)
|
Optional. Overrides
How to obtain tags for logsSelect from the following options:
The values for |
|
tag_include (string)Since MCC 2.23.0 (11.7.0)
|
Optional. Is overridden by |
|
Key |
Description |
Example values |
|---|---|---|
|
Enables or disables Ceph monitoring on managed clusters. Set to |
|
Key |
Description |
Example values |
|---|---|---|
|
Enables or disables HTTP endpoints monitoring. If enabled, the
monitoring tool performs the probes against the defined endpoints every
15 seconds. Set to |
|
|
Defines the directory path with external endpoints certificates on host. |
|
|
Defines the list of HTTP endpoints to monitor. The endpoints must successfully respond to a liveness probe. For success, a request to a specific endpoint must result in a 2xx HTTP response code. |
domains:
- https://prometheus.io/health
- http://example.com:8080/status
- http://example.net:8080/pulse
|
Key |
Description |
Example values |
|---|---|---|
|
Enables or disables monitoring of bare metal Ironic. To enable, specify the Ironic API URL. |
|
|
Defines whether to skip the chain and host verification. Set to
|
|
Key |
Description |
Example values |
|---|---|---|
|
Enables or disables Mirantis Kubernetes Engine (MKE) monitoring.
Set to |
|
|
Defines the dockerd data root directory of persistent Docker state. For details, see Docker documentation: Daemon CLI (dockerd). |
|
Key |
Description |
Example values |
|---|---|---|
|
Enables or disables StackLight to monitor and alert on the expiration
date of the TLS certificate of an HTTPS endpoint. If enabled, the
monitoring tool performs the probes against the defined endpoints every
hour. Set to |
|
|
Defines the list of HTTPS endpoints to monitor the certificates from. |
domains:
- https://prometheus.io
- https://example.com:8080
|
Key |
Description |
Example values |
|---|---|---|
|
On the clusters that run large-scale workloads, workload monitoring generates a big amount of resource-consuming metrics. To prevent generation of excessive metrics, you can disable workload monitoring in the StackLight metrics and monitor only the infrastructure. The |
metricFilter:
enabled: true
action: keep
namespaces:
- kaas
- kube-system
- stacklight
|
Key |
Description |
Example values |
|---|---|---|
|
Defines the |
default:
role: stacklight
|
|
Defines the |
component:
alerta:
role: stacklight
component: alerta
# kibana:
# role: stacklight
# component: kibana
opensearchDashboards:
role: stacklight
component: opensearchdashboards
|
Key |
Description |
Example values |
|---|---|---|
|
Removed in Container Cloud 2.26.0 (Cluster releases 17.1.0 and 16.1.0). Specifies the retention time per index. Includes the following parameters:
The allowed values include integers (days) and numbers with suffixes: y, m, w, d, h, including capital letters. By default, values set in |
elasticsearch:
retentionTime:
logstash: 3
events: "2w"
notifications: "1M"
|
|
Removed in Container Cloud 2.26.0 (Cluster releases 17.1.0 and 16.1.0). Defines the OpenSearch (Elasticsearch) Note Due to the known issue 27732-2,
a custom setting for this parameter is dismissed during cluster deployment
and changes to one day (default). Refer to the known issue description
for the affected |
|
|
Specifies the OpenSearch (Elasticsearch) PVC(s) size. The number of PVCs depends on the StackLight database mode. For HA, three PVCs will be created, each of the size specified in this parameter. For non-HA, one PVC of the specified size. Important You cannot modify this parameter after cluster creation. Note Due to the known issue 27732-1,
that is fixed in Container Cloud 2.22.0 (Cluster releases 11.6.0 and 12.7.0),
the OpenSearch PVC size configuration is dismissed during a cluster
deployment. Refer to the known issue description for affected
|
elasticsearch:
persistentVolumeClaimSize: 30Gi
|
|
Available since Container Cloud 2.26.0 (Cluster releases 17.1.0 and 16.1.0). Optional. Specifies the number of gigabytes that is exclusively available for the OpenSearch data.
If not set (by default), the number of gigabytes from
This parameter is useful in the following cases:
|
elasticsearch:
persistentVolumeUsableStorageSizeGB: 160
|
Key |
Description |
Example values |
|---|---|---|
|
Additional configuration for |
logging:
dashboardsExtraConfig:
opensearch.requestTimeout: 60000
|
Key |
Description |
Example values |
|---|---|---|
|
Additional configuration for Since Container Cloud 2.29.0 and MOSK 25.1, by default,
StackLight manages watermarks efficiently (low/high/flood: 150/100/50 GB).
If |
logging:
extraConfig:
cluster.max_shards_per_node: 5000
|
Key |
Description |
Example values |
|---|---|---|
|
Defines the minimum amount of time for Prometheus to wait before
resending an alert to Alertmanager. Passed to the
|
|
|
Available since Container Cloud 2.26.0 (Cluster releases 17.1.0 and 16.1.0). Defines the list of labels to be injected to firing alerts while they are sent to Alertmanager. Empty by default. The following labels are reserved for internal purposes and cannot
be overridden: Caution When new labels are injected, Prometheus sends alert updates with a new set of labels, which can potentially cause Alertmanager to have duplicated alerts for a short period of time if the cluster currently has firing alerts. |
alertsCommonLabels:
region: west
environment: prod
|
|
Specifies the Prometheus PVC(s) size. The number of PVCs depends on the StackLight database mode. For HA, three PVCs will be created, each of the size specified in this parameter. For non-HA, one PVC of the specified size. Important You cannot modify this parameter after cluster creation. |
prometheusServer:
persistentVolumeClaimSize: 16Gi
|
|
Available since Container Cloud 2.24.0 (Cluster release 14.0.0).
Defines the number of concurrent queries limit. Passed to the
|
|
|
Defines the Prometheus database retention size. Passed to the
|
|
|
Defines the Prometheus database retention period. Passed to the
|
|
Key |
Description |
Example values |
|---|---|---|
|
Specifies a set of custom Blackbox Exporter modules. For details, see
Blackbox Exporter configuration: module.
The |
customModules:
http_post_2xx:
prober: http
timeout: 5s
http:
method: POST
headers:
Content-Type: application/json
body: '{}'
|
|
Specifies the offset to subtract from timeout in seconds
( |
|
Key |
Description |
Example values |
|---|---|---|
|
Defines custom Prometheus recording rules. Overriding of existing recording rules is not supported. |
customRecordingRules:
- name: ExampleRule.http_requests_total
rules:
- expr: sum by(job) (rate(http_requests_total[5m]))
record: job:http_requests:rate5m
- expr: avg_over_time(job:http_requests:rate5m[1w])
record: job:http_requests:rate5m:avg_over_time_1w
|
Key |
Description |
Example values |
|---|---|---|
|
Defines custom Prometheus scrape configurations. For details, see Prometheus documentation: scrape_config. The names of default StackLight scrape configurations, which you can view in the Status -> Targets tab of the Prometheus web UI, are reserved for internal usage and any overrides will be discarded. Therefore, provide unique names to avoid overrides. |
customScrapeConfigs:
custom-grafana:
scrape_interval: 10s
scrape_timeout: 5s
kubernetes_sd_configs:
- role: endpoints
relabel_configs:
- source_labels:
- __meta_kubernetes_service_label_app
- __meta_kubernetes_endpoint_port_name
regex: grafana;service
action: keep
- source_labels:
- __meta_kubernetes_pod_name
target_label: pod
|
Available since Container Cloud 2.24.0 (Cluster release 14.0.0)
Key |
Description |
Example values |
|---|---|---|
|
Configuration for managing Prometheus metrics filtering. When enabled (default), only actively used and explicitly white-listed metrics get scraped by Prometheus. |
prometheusServer:
metricsFiltering:
enabled: true
|
|
List of extra metrics to whitelist, which are dropped by default. Contains the following parameters:
|
prometheusServer:
metricsFiltering:
enabled: true
extraMetricsInclude:
cadvisor:
- container_memory_failcnt
- container_network_transmit_errors_total
calico:
- felix_route_table_per_iface_sync_seconds_sum
- felix_bpf_dataplane_endpoints
_group-go-collector-metrics:
- go_gc_heap_goal_bytes
- go_gc_heap_objects_objects
|
Key |
Description |
Example values |
|---|---|---|
|
Excludes monitoring of RegExp-specified network devices. The number of network interface-related metrics is significant and may cause extended Prometheus RAM usage in big clusters. Therefore, Prometheus Node Exporter only collects information of a basic set of interfaces (both host and container) and excludes the following monitoring interfaces:
To enable information collecting for the interfaces above, edit the list of blacklisted devices as needed. |
nodeExporter:
netDeviceExclude: "^(veth.+|cali.+|o-hm0|tap.+|qg-.+|qr-.+|ha-.+|br-.+|ovs-system|docker0)$"
|
|
Enables Node Exporter collectors. For a list of available collectors, see Node Exporter Collectors. The following collectors are enabled by default in StackLight:
|
extraCollectorsEnabled:
- bcache
- bonding
- softnet
|
Note
Prometheus Relay is set up as an endpoint in the Prometheus datasource in Grafana. Therefore, all requests from Grafana are sent to Prometheus through Prometheus Relay. If Prometheus Relay reports request timeouts or exceeds the response size limits, you can configure the parameters below. In this case, Prometheus Relay resource limits may also require tuning.
Key |
Description |
Example values |
|---|---|---|
|
Specifies the client timeout in seconds. If empty, defaults to a value
determined by the cluster size: Note The cluster size parameters are available since Container Cloud 2.24.0 (Cluster release 14.0.0). |
|
|
Specifies the response size limit in bytes. If empty, defaults to a
value determined by the cluster size: Note The cluster size parameters are available since Container Cloud 2.24.0 (Cluster release 14.0.0). |
|
Allows sending of metrics from Prometheus to a custom monitoring endpoint. For details, see Prometheus Documentation: remote_write.
Key |
Description |
Example values |
|---|---|---|
|
Skip this step if your remote server does not have authorization.
Defines additional mounts for Note To create more than one file for the same remote write
endpoint, for example, to configure TLS connections,
use a single secret object with multiple keys in the ...
data:
cert_file: aWx1dnRlc3Rz
key_file: dGVzdHVzZXI=
...
|
remoteWriteSecretMounts:
- secretName: prom-secret-files
mountPath: /etc/config/remote_write
|
|
Defines the configuration of a custom remote_write endpoint for sending Prometheus samples. Note If the remote server uses authorization, first create
secret(s) in the |
remoteWrites:
- url: http://remote_url/push
authorization:
credentials_file: /etc/config/remote_write/key_file
|
Key |
Description |
Example values |
|---|---|---|
|
Provides the capability to override the default resource requests or limits for any StackLight component for the predefined cluster sizes. Caution Since Container Cloud 2.28.0 (Cluster releases 17.3.0 and
16.3.0),
StackLight components for resource limits customizationNote The below list has the
alerta: alerta/alerta
alertmanager: prometheus-alertmanager/prometheus-alertmanager
alertmanagerWebhookServicenow: alertmanager-webhook-servicenow/alertmanager-webhook-servicenow
blackboxExporter: prometheus-blackbox-exporter/blackbox-exporter
elasticsearch: opensearch-master/opensearch # Deprecated
elasticsearchCurator: elasticsearch-curator/elasticsearch-curator
elasticsearchExporter: elasticsearch-exporter/elasticsearch-exporter
fluentdElasticsearch: fluentd-logs/fluentd-logs # Deprecated
fluentdLogs: fluentd-logs/fluentd-logs
fluentdNotifications: fluentd-notifications/fluentd
grafana: grafana/grafana
grafanaRenderer: grafana/grafana-renderer # Removed in MCC 2.27.0 (17.2.0 and 16.2.0)
iamProxy: iam-proxy/iam-proxy # Deprecated
iamProxyAlerta: iam-proxy-alerta/iam-proxy
iamProxyAlertmanager: iam-proxy-alertmanager/iam-proxy
iamProxyGrafana: iam-proxy-grafana/iam-proxy
iamProxyKibana: iam-proxy-kibana/iam-proxy # Deprecated
iamProxyOpenSearchDashboards: iam-proxy-kibana/iam-proxy
iamProxyPrometheus: iam-proxy-prometheus/iam-proxy
kibana: opensearch-dashboards/opensearch-dashboards # Deprecated
kubeStateMetrics: prometheus-kube-state-metrics/prometheus-kube-state-metrics
libvirtExporter: prometheus-libvirt-exporter/prometheus-libvirt-exporter
metricCollector: metric-collector/metric-collector
metricbeat: metricbeat/metricbeat
nodeExporter: prometheus-node-exporter/prometheus-node-exporter
opensearch: opensearch-master/opensearch
opensearchDashboards: opensearch-dashboards/opensearch-dashboards
patroniExporter: patroni/patroni-patroni-exporter
pgsqlExporter: patroni/patroni-pgsql-exporter
postgresql: patroni/patroni
prometheusEsExporter: prometheus-es-exporter/prometheus-es-exporter
prometheusMsTeams: prometheus-msteams/prometheus-msteams
prometheusRelay: prometheus-relay/prometheus-relay
prometheusServer: prometheus-server/prometheus-server
sfNotifier: sf-notifier/sf-notifier
sfReporter: sf-reporter/sf-reporter
stacklightHelmControllerController: stacklight-helm-controller/controller
telegrafDockerSwarm: telegraf-docker-swarm/telegraf-docker-swarm
telegrafDs: telegraf-ds-smart/telegraf-ds-smart # Deprecated
telegrafDsSmart: telegraf-ds-smart/telegraf-ds-smart
telegrafOpenstack: telegraf-openstack/telegraf-openstack # replaced with osdpl-exporter in 24.1
telegrafS: telegraf-docker-swarm/telegraf-docker-swarm # Deprecated
telemeterClient: telemeter-client/telemeter-client
telemeterServer: telemeter-server/telemeter-server
telemeterServerAuthServer: telemeter-server/telemeter-server-authorization-server
tfControllerExporter: prometheus-tf-controller-exporter/prometheus-tungstenfabric-exporter
tfVrouterExporter: prometheus-tf-vrouter-exporter/prometheus-tungstenfabric-exporter
|
resourcesPerClusterSize:
# elasticsearch:
opensearch:
small:
limits:
cpu: "1000m"
memory: "4Gi"
medium:
limits:
cpu: "2000m"
memory: "8Gi"
requests:
cpu: "1000m"
memory: "4Gi"
large:
limits:
cpu: "4000m"
memory: "16Gi"
|
|
Provides the capability to override the containers resource requests or limits for any StackLight component.
StackLight components for resource limits customizationNote The below list has the
alerta: alerta/alerta
alertmanager: prometheus-alertmanager/prometheus-alertmanager
alertmanagerWebhookServicenow: alertmanager-webhook-servicenow/alertmanager-webhook-servicenow
blackboxExporter: prometheus-blackbox-exporter/blackbox-exporter
elasticsearch: opensearch-master/opensearch # Deprecated
elasticsearchCurator: elasticsearch-curator/elasticsearch-curator
elasticsearchExporter: elasticsearch-exporter/elasticsearch-exporter
fluentdElasticsearch: fluentd-logs/fluentd-logs # Deprecated
fluentdLogs: fluentd-logs/fluentd-logs
fluentdNotifications: fluentd-notifications/fluentd
grafana: grafana/grafana
grafanaRenderer: grafana/grafana-renderer # Removed in MCC 2.27.0 (17.2.0 and 16.2.0)
iamProxy: iam-proxy/iam-proxy # Deprecated
iamProxyAlerta: iam-proxy-alerta/iam-proxy
iamProxyAlertmanager: iam-proxy-alertmanager/iam-proxy
iamProxyGrafana: iam-proxy-grafana/iam-proxy
iamProxyKibana: iam-proxy-kibana/iam-proxy # Deprecated
iamProxyOpenSearchDashboards: iam-proxy-kibana/iam-proxy
iamProxyPrometheus: iam-proxy-prometheus/iam-proxy
kibana: opensearch-dashboards/opensearch-dashboards # Deprecated
kubeStateMetrics: prometheus-kube-state-metrics/prometheus-kube-state-metrics
libvirtExporter: prometheus-libvirt-exporter/prometheus-libvirt-exporter
metricCollector: metric-collector/metric-collector
metricbeat: metricbeat/metricbeat
nodeExporter: prometheus-node-exporter/prometheus-node-exporter
opensearch: opensearch-master/opensearch
opensearchDashboards: opensearch-dashboards/opensearch-dashboards
patroniExporter: patroni/patroni-patroni-exporter
pgsqlExporter: patroni/patroni-pgsql-exporter
postgresql: patroni/patroni
prometheusEsExporter: prometheus-es-exporter/prometheus-es-exporter
prometheusMsTeams: prometheus-msteams/prometheus-msteams
prometheusRelay: prometheus-relay/prometheus-relay
prometheusServer: prometheus-server/prometheus-server
sfNotifier: sf-notifier/sf-notifier
sfReporter: sf-reporter/sf-reporter
stacklightHelmControllerController: stacklight-helm-controller/controller
telegrafDockerSwarm: telegraf-docker-swarm/telegraf-docker-swarm
telegrafDs: telegraf-ds-smart/telegraf-ds-smart # Deprecated
telegrafDsSmart: telegraf-ds-smart/telegraf-ds-smart
telegrafOpenstack: telegraf-openstack/telegraf-openstack # replaced with osdpl-exporter in 24.1
telegrafS: telegraf-docker-swarm/telegraf-docker-swarm # Deprecated
telemeterClient: telemeter-client/telemeter-client
telemeterServer: telemeter-server/telemeter-server
telemeterServerAuthServer: telemeter-server/telemeter-server-authorization-server
tfControllerExporter: prometheus-tf-controller-exporter/prometheus-tungstenfabric-exporter
tfVrouterExporter: prometheus-tf-vrouter-exporter/prometheus-tungstenfabric-exporter
|
resources:
alerta:
requests:
cpu: "50m"
memory: "200Mi"
limits:
memory: "500Mi"
Using the example above, each pod in the Note The logging mechanism performance depends on the cluster log
load. If the cluster components send an excessive amount of logs, the
default resource requests and limits for resources:
# fluentdElasticsearch:
fluentdLogs:
requests:
memory: "500Mi"
limits:
memory: "1500Mi"
|
On the managed clusters with limited Internet access, proxy is required for StackLight components that use HTTP and HTTPS and are disabled by default but need external access if enabled. The Salesforce reporter depends on the Internet access through HTTPS.
Key |
Description |
Example values |
|---|---|---|
|
Unique cluster identifier
The |
Do not modify |
|
Enables or disables reporting of Prometheus metrics to Salesforce. For details, see Container Cloud Reference Architecture: StackLight Deployment architecture. Disabled by default. |
|
|
Salesforce parameters and credentials for the metrics reporting integration. |
Note Modify this parameter if salesForceAuth:
url: "<SF instance URL>"
username: "<SF account email address>"
password: "<SF password>"
environment_id: "<Cloud identifier>"
organization_id: "<Organization identifier>"
sandbox_enabled: "<Set to true or false>"
|
|
Defines the Kubernetes cron job for sending metrics to Salesforce. By default, reports are sent at midnight server time. |
cronjob:
schedule: "0 0 * * *"
concurrencyPolicy: "Allow"
failedJobsHistoryLimit: ""
successfulJobsHistoryLimit: ""
startingDeadlineSeconds: 200
|
In an HA StackLight setup, when highAvailabilityEnabled is set to true,
all StackLight Persistent Volumes (PVs) use the Local Volume Provisioner (LVP)
storage class not to rely on dynamic provisioners such as Ceph, which are not
available in every deployment. In a non-HA StackLight setup, when no storage
class is specified, PVs use the default storage class of a cluster.
Key |
Description |
Example values |
|---|---|---|
|
Defines the |
|
|
Defines (overrides the |
componentStorageClasses:
elasticsearch: ""
opensearch: ""
fluentd: ""
postgresql: ""
prometheusAlertManager: ""
prometheusServer: ""
|
Verify StackLight after configuration¶
This section describes how to verify StackLight after configuring its parameters as described in StackLight configuration procedure and StackLight configuration parameters. Perform the verification procedure described for a particular modified StackLight key.
Key |
Verification procedure |
|---|---|
|
|
|
|
|
|
|
|
|
In the OpenSearch Dashboards web UI, click Discover and
verify that the |
|
In the Grafana web UI, verify that the OpenStack dashboards are present and not empty. |
|
|
Key |
Verification procedure |
|---|---|
|
Verify that Alerta is present in the list of StackLight resources. An empty output indicates that Alerta is disabled. kubectl get all -n stacklight -l app=alerta
|
|
In the Alertmanager web UI, navigate to Status and verify
that the Config section contains the |
|
In the Alertmanager web UI, navigate to Status and verify that the Config section contains the intended receiver(s). |
|
In the Alertmanager web UI, navigate to Status and verify that the Config section contains the intended route(s). |
|
Run the following command. An empty output indicates either a failure or that the feature is disabled. kubectl get cm -n stacklight prometheus-alertmanager -o \
yaml | grep -A 6 inhibit_rules
|
|
|
|
|
|
|
|
Verify that kubectl get deployment sf-notifier \
-o=jsonpath='{.spec.template.spec.containers[0].env}' | jq .
|
|
In case of any failure:
|
|
In the Alertmanager web UI, navigate to Status and verify
that the Config section contains the |
|
|
|
Verify that the kubectl get deployment.apps/prometheus-blackbox-exporter -n stacklight \
-o=jsonpath='{.spec.template.spec.containers[?(@.name=="blackbox-exporter")].args}'
For example, for |
|
|
|
|
elasticsearch.logstashRetentionTimeRemoved in MCC 2.26.0 (17.1.0, 16.1.0)
|
Verify that the kubectl get cm elasticsearch-curator-config -n \
stacklight -o=jsonpath='{.data.action_file\.yml}'
|
|
Verify that the PVC(s) capacity is equal or higher (in case of statically provisioned volumes) than specified: kubectl get pvc -n stacklight -l "app=opensearch-master"
|
|
|
|
|
|
In the Grafana web UI, verify that the desired dashboard is set as a home dashboard. |
grafana.renderer.enabledRemoved in MCC 2.27.0 (17.2.0, 16.2.0)
|
Verify the Grafana Image Renderer. If set to kubectl logs -f -n stacklight -l app=grafana \
--container grafana-renderer
|
|
Verify the number of service replicas for the HA or non-HA StackLight mode. For details, see Deployment architecture. kubectl get sts -n stacklight
|
|
In the Grafana web UI, verify that the Ironic BM dashboard displays valuable data (no false-positive or empty panels). |
|
Verify that the customization has applied: kubectl -n stacklight get cm opensearch-dashboards -o=jsonpath='{.data}'
Example of system response: {"opensearch_dashboards.yml":"opensearch.hosts: http://opensearch-master:9200\
\nopensearch.requestTimeout: 60000\
\nopensearchDashboards.defaultAppId: dashboard/2d53aa40-ad1f-11e9-9839-052bda0fdf49\
\nserver:\
\n host: 0.0.0.0\
\n name: opensearch-dashboards\n"}
|
|
Verify that OpenSearch, Fluentd, and OpenSearch Dashboards are present in the list of StackLight resources. An empty output indicates that the StackLight logging stack is disabled. kubectl get all -n stacklight -l 'app in
(opensearch-master,opensearchDashboards,fluentd-logs)'
|
|
To troubleshoot issues with Splunk, refer to No logs are forwarded to Splunk. |
|
Verify that files were created for the specified path in the Fluentd container: kubectl get pods -n stacklight -o name | grep fluentd-logs | \
xargs -I{} kubectl exec -i {} -c fluentd-logs -n stacklight -- \
ls <logging.externalOutputSecretMounts.mountPath>
|
|
Verify that the customization has applied: kubectl -n stacklight get cm opensearch-master-config -o=jsonpath='{.data}'
Example of system response: {"opensearch.yml":"cluster.name: opensearch\
\nnetwork.host: 0.0.0.0\
\nplugins.security.disabled: true\
\nplugins.index_state_management.enabled: false\
\npath.data: /usr/share/opensearch/data\
\ncompatibility.override_main_response_version: true\
\ncluster.max_shards_per_node: 5000\n"}
|
logging.levelRemoved in MCC 2.26.0 (17.1.0, 16.1.0)
|
|
|
For details, see steps 4.2 and 4.3 in Create logs-based metrics. |
|
|
|
Verify that kubectl get cm -n stacklight fluentd-logs -o \
yaml | grep packet_size
|
|
|
|
In the Prometheus web UI, navigate to Alerts and verify that
the |
|
|
|
In the Prometheus web UI, run the following PromQL queries. The result should not be empty. node_scrape_collector_duration_seconds{collector="<COLLECTOR_NAME>"}
node_scrape_collector_success{collector="<COLLECTOR_NAME>"}
|
|
|
|
Verify that the appropriate components pods are located on the intended nodes: kubectl get pod -o=custom-columns=NAME:.metadata.name,\
STATUS:.status.phase,NODE:.spec.nodeName -n stacklight
|
|
|
|
|
|
In the Prometheus web UI, navigate to Alerts and verify that the list of alerts has changed according to your customization. |
|
|
|
It may take up to 10 minutes for the change to apply. |
|
Verify that the PVC(s) capacity equals or is higher (in case of statically provisioned volumes) than specified: kubectl get pvc -n stacklight -l "app=prometheus,component=server"
|
|
|
|
|
|
Verify that files were created for the specified path in the Prometheus container: kubectl exec -it prometheus-server-0 -c prometheus-server -n \
stacklight -- ls <remoteWriteSecretMounts.mountPath>
|
|
In the Prometheus web UI, navigate to Alerts and verify that
the list of alerts contains the |
|
|
|
|
|
Verify that the appropriate components PVCs have been created according
to the configured kubectl get pvc -n stacklight
|
Tune OpenSearch performance¶
The following hardware recommendations and software settings apply for better OpenSearch performance in a MOSK cluster.
To tune OpenSearch performance:
Depending on your cluster size, set the required disk and CPU size along with memory limit and heap size.
Heap size is calculated in StackLight as ⅘ of the specified memory limit. If the calculated heap size exceeds 32 GB, slightly crossing this threshold causes significant waste of memory due to loss of Ordinary Object Pointers (OOPS) compression, which allows storing 64-bit pointers in 32-bits.
Since Container Cloud 2.25.0 (Cluster releases 17.0.0 and 16.0.0), to prevent this behavior, for the memory limit in the 31-50 GB range, the heap size is set to fixed 31 GB using the
enforceOopsCompressionparameter, which is enabled by default. For details, see Logging: Enforce OOPS compression. Exceeding the range causes loss of benefit of OOPS compression, so the ⅘ formula applies again.OpenSearch is write-heavy, so SSD is preferable as a disk type.
Hardware recommendations for OpenSearch¶ Cluster size
Memory limit (GB)
Heap size (GB)
CPU (# of cores)
Small
16
12.8
2
Medium
32
25.6
4
Large
64
51.2
8
To configure hardware settings for OpenSearch, refer to Resource limits in the StackLight configuration procedure section.
Configure the maximum count of
mmapfiles. OpenSearch uses mmapfs to map shards stored on disk, which is set to65530by default.To verify
max_map_count:sysctl -n vm.max_map_count
To increase
max_map_count, follow the Create MOSK host profiles procedure.Example configuration:
kernelParameters: sysctl: vm.max_map_count: "<value>"
Extended retention periods, which depend on open shards, require increasing this value significantly. For example, to
262144.Configure swap as it significantly degrades performance. Lower
swappinessto1or0(to disable swap). For details, use the Create MOSK host profiles procedure.Example configuration:
kernelParameters: sysctl: vm.swappiness: "<value>"
Configure the kernel I/O scheduler to improve timing of disk writing operations. Change it to one of the following options:
none- applies the FIFO queue.mq-deadline- applies three queues: FIFO read, FIFO write, and sorted.
Changing I/O scheduling is also possible through
BareMetalHostProfile. However, the specific implementation highly depends on the disk type used:cat /sys/block/sda/queue/scheduler mq-deadline kyber bfq [none]
Export logs from OpenSearch Dashboards to CSV¶
Available since MCC 2.23.0 (12.7.0 and 11.7.0)
This section describes how to export logs from the OpenSearch Dashboards navigation panel to the CSV format.
Caution
The log limit is set 10 000 rows, and it does not take into account the resulted file size.
Note
The following instruction describes how to export all logs from the
opensearch-master-0 node of an OpenSearch cluster.
To export logs from the OpenSearch Dashboards navigation panel to CSV:
Log in to the OpenSearch Dashboards web UI as described in Getting access.
Navigate to the Discover page.
In the left navigation panel, select the required log index pattern from the top drop-down menu. For example, system* for system logs and audit* for audit logs.
In the middle top menu, click Add filter and add the required filters. For example:
event.provider matches the
opensearch-masterloggerorchestrator.pod matches the
opensearch-master-0node name
In Search field names, search for required fields to be present in the resulting CSV file. For example:
orchestrator.pod for
opensearch-master-0message for the log message
In the right top menu:
Click Save to save the filter after naming it.
Click Reporting > Generate CSV.
When the report generation completes, download the file depending on your browser settings.
OpenSearch Dashboards¶
This section describes OpenSearch Dashboards that enable you to observe visual representation of logs and Kubernetes events of your cluster.
View OpenSearch Dashboards¶
OpenSearch Dashboards is part of the StackLight logging stack. Using the OpenSearch Dashboards web UI, you can view the visual representation of your OpenStack deployment notifications, logs, Kubernetes events, and other cluster notifications related to your deployment.
Note
By default, StackLight logging stack, including OpenSearch Dashboards, is disabled. For details, see Mirantis Container Cloud Reference Architecture: Deployment architecture.
To view the OpenSearch Dashboards:
Log in to the OpenSearch Dashboards web UI as described in Getting access.
Click the required dashboard to inspect the visualizations or perform a search:
Dashboard
Description
Notifications
Provides visualizations on the number of notifications over time per source and severity, host, and breakdowns. The dashboard includes search.
K8s events
Provides visualizations on the number of Kubernetes events per type, and top event-producing resources and namespaces by reason and event type. Includes search.
System Logs
Available for clusters created since Container Cloud 2.26.0 (Cluster releases 17.1.x, 16.1.x, or later).
Provides visualizations on the number of log messages per severity, source, and top log-producing host, namespaces, containers, and applications. Includes search.
Caution
Due to a known issue, this dashboard does not exist in Container Cloud 2.26.0 (Cluster releases 17.1.0 and 16.1.0). The issue is addressed in Container Cloud 2.26.1 (Cluster releases 17.1.1 and 16.1.1). To work around the issue in 2.26.0, you can map the fields of the
logstashindex to thesystemone and view logs in the deprecated Logs dashboard. For mapping details, see System index fields mapped to Logstash index fields.Logs Deprecated in 2.26.0 (17.1.0 and 16.1.0)
Available only for clusters created before Container Cloud 2.26.0 (Cluster releases 17.0.x, 16.0.x, or earlier).
Analogous to System Logs but contains logs generated only for the mentioned Cluster releases.
Search in OpenSearch Dashboards¶
OpenSearch Dashboards provide the following search tools:
Filters
Queries
Full-text search
Filters enable you to organize the output information using the interface tools. You can search for information by a set of indexed fields using a variety of logical operators.
Queries enable you to construct search commands using OpenSearch query domain-specific language (DSL) expressions. These expressions allow you to search by the fields not included in the index.
In addition to filters and queries, you can use the Search input field for full-text search.
From the dashboard view, click Add filter.
In the dialog that opens, select the field of search in the Field drop-down menu.
Select the logical operator in the Operator drop-down menu.
Type or select the filter value from the Value drop-down menu.
Available since MCC 2.23.0 (12.7.0 and 11.7.0)
For the orchestrator.labels field of the system and audit log indices,
you can use the flat_object field type to apply the filtering using
value or valueAndPath. For example:
Using
value: to obtain all logs produced byiam-proxy, add the following filters:orchestrator.typethat matcheskubernetesorchestrator.labels._valuethat matchesiam-proxy
Using
valueAndPath: to obtain all logs produced by the OpenSearch cluster, add the following filters:orchestrator.typethat matcheskubernetesorchestrator.labels._valueAndPaththat matchesorchestrator.labels.app=opensearch-master
From the dashboard view, click Add filter.
In the dialog that opens, click Edit as Query DSL and type in the search request.
View Grafana dashboards¶
Using the Grafana web UI, you can view the visual representation of the metric graphs based on the time series databases.
Most Grafana dashboards include a View logs in OpenSearch Dashboards link to immediately view relevant logs in the OpenSearch Dashboards web UI. The OpenSearch Dashboards web UI displays logs filtered using the Grafana dashboard variables, such as the drop-downs. Once you amend the variables, wait for Grafana to generate a new URL.
Note
Due to the known issue, the View logs in OpenSearch Dashboards link does not work in Container Cloud 2.26.0 (Cluster releases 17.1.0 and 16.1.0). The issue is addressed in Container Cloud 2.26.1 (Cluster releases 17.1.1 and 16.1.1).
Caution
The Grafana dashboards that contain drop-down lists are limited to 1000 lines. Therefore, if you require data on a specific item, use the filter by name instead.
Note
Grafana dashboards that present node data have an additional Node identifier drop-down menu. By default, it is set to machine to display short names for Kubernetes nodes. To display Kubernetes node name labels, change this option to node.
To view the Grafana dashboards:
Log in to the Grafana web UI as described Getting access.
From the drop-down list, select the required dashboard to inspect the status and statistics of the corresponding service in your management or MOSK cluster:
Component
Dashboard
Description
Ceph cluster
Ceph Cluster
Provides the overall health status of the Ceph cluster, capacity, latency, and recovery metrics.
Ceph Nodes
Provides an overview of the host-related metrics, such as the number of Ceph Monitors, Ceph OSD hosts, average usage of resources across the cluster, network and hosts load.
This dashboard is deprecated since Container Cloud 2.25.0 (Cluster releases 17.0.0 and 16.0.0) and is removed in Container Cloud 2.26.0 (Cluster releases 17.1.0 and 16.1.0).
Therefore, Mirantis recommends switching to the following dashboards in the current release:
For Ceph stats, use the Ceph Cluster dashboard.
For resource utilization, use the System dashboard, which includes filtering by Ceph node labels, such as
ceph_role_osd,ceph_role_mon, andceph_role_mgr.
Ceph OSDs
Provides metrics for Ceph OSDs, including the Ceph OSD read and write latencies, distribution of PGs per Ceph OSD, Ceph OSDs and physical device performance.
Ceph Pools
Provides metrics for Ceph pools, including the client IOPS and throughput by pool and pools capacity usage.
Ironic
Ironic BM
Provides graphs on Ironic health, HTTP API availability, provisioned nodes by state and installed
ironic-conductorbackend drivers.Container Cloud
Clusters Overview
Represents the main cluster capacity statistics for all clusters of a Container Cloud deployment where StackLight is installed.
Note
Due to the known issue, the Prometheus Targets Unavailable panel of the Clusters Overview dashboard does not display data for managed clusters of the 11.7.0, 11.7.4, 12.5.0, and 12.7.x series Cluster releases after update to Container Cloud 2.24.0.
Etcd
Available since Container Cloud 2.21.0 (Cluster release 11.5.0). Provides graphs on database size, leader elections, requests duration, incoming and outgoing traffic.
MCC Applications Performance
Available since Container Cloud 2.23.0 (Cluster release 11.7.0). Provides information on the Container Cloud internals work based on Golang, controller runtime, and custom metrics. You can use it to verify performance of applications and for troubleshooting purposes.
Kubernetes resources
Kubernetes Calico
Provides metrics of the entire Calico cluster usage, including the cluster status, host status, and Felix resources.
Kubernetes Cluster
Provides metrics for the entire Kubernetes cluster, including the cluster status, host status, and resources consumption.
Kubernetes Containers
Provides charts showing resource consumption per deployed Pod containers running on Kubernetes nodes.
Kubernetes Deployments
Provides information on the desired and current state of all service replicas deployed on a Container Cloud cluster.
Kubernetes Namespaces
Provides the Pods state summary and the CPU, MEM, network, and IOPS resources consumption per name space.
Kubernetes Nodes
Provides charts showing resources consumption per Container Cloud cluster node.
Kubernetes Pods
Provides charts showing resources consumption per deployed Pod.
NGINX
NGINX
Provides the overall status of the NGINX cluster and information about NGINX requests and connections.
OpenStack
OpenStack - Overview
Provides general information on OpenStack services resources consumption, API errors, deployed OpenStack compute nodes and block storage usage.
OpenStack Ingress controller
Available since MOSK 23.3. Monitors the number of requests, response times and statuses, as well as the number of Ingress SSL certificates including expiration time and resources usage.
OpenStack Instances Availability
Available since MOSK 23.2. Provides information about the availability of instance floating IPs per OpenStack compute node and project. Also, enables monitoring of probe statistics for individual instance floating IPs.
OpenStack Network IP Capacity
Available since MOSK 25.1. Provides information about the statistics of IP address allocation for external networks and subnets on non-Tungsten Fabric based MOSK clusters. For configuration details, see Start monitoring IP address capacity.
OpenStack PortProber
Available since MOSK 24.2. Provides information about the availability of Neutron ports per OpenStack compute node, project, and port owner.
OpenStack PortProber [Deprecated]
Available since MOSK 25.1. Provides information about the availability of Neutron ports per OpenStack compute node, project, and port owner. Deprecated in favor of the OpenStack PortProber dashboard.
Use this deprecated dashboard only to access old data collected before MOSK 25.1.
OpenStack PowerDNS
Available since MOSK 24.3. Provides different stats about OpenStack PowerDNS servers such as connections, resources, queries, rings, errors, and other.
OpenStack Usage Efficiency
Available since MOSK 23.3. Provides information about requested (allocated) CPU and memory usage efficiency on a per-project and per-flavor basis. Aims to identify flavors that specific projects are not effectively using, with allocations significantly exceeding actual usage. Also, evaluates per-instance underuse for specific projects.
KPI - Provisioning
Provides provisioning statistics for OpenStack compute instances, including graphs on VM creation results by day.
Cinder
Provides graphs on the OpenStack Block Storage service health, HTTP API availability, pool capacity and utilization, number of created volumes and snapshots.
Glance
Provides graphs on the OpenStack Image service health, HTTP API availability, number of created images and snapshots.
Gnocchi
Provides panels and graphs on the Gnocchi health and HTTP API availability.
Heat
Provides graphs on the OpenStack Orchestration service health, HTTP API availability and usage.
Ironic OpenStack
Provides graphs on the OpenStack Bare Metal Provisioning service health, HTTP API availability, provisioned nodes by state and installed ironic-conductor backend drivers.
Keystone
Provides graphs on the OpenStack Identity service health, HTTP API availability, number of tenants and users by state.
Neutron
Provides graphs on the OpenStack networking service health, HTTP API availability, agents status and usage of Neutron L2 and L3 resources.
NGINX Ingress controller
Not recommended. Deprecated since MOSK 23.3 and is removed in MOSK 24.1. Use OpenStack Ingress controller instead.
Monitors the number of requests, response times and statuses, as well as the number of Ingress SSL certificates including expiration time and resources usage.
Nova - Availability Zones
Provides detailed graphs on the OpenStack availability zones and hypervisor usage.
Nova - Hypervisor Overview
Provides a set of single-stat panels presenting resources usage by host.
Nova - Instances
Provides graphs on libvirt Prometheus exporter health and resources usage. Monitors the number of running instances and tasks and allows sorting the metrics by top instances.
Nova - Overview
Provides graphs on the OpenStack compute services (
nova-scheduler,nova-conductor, andnova-compute) health, as well as HTTP API availability.Nova - Tenants
Provides graphs on CPU, RAM, disk throughput, IOPS, and space usage and allocation and allows sorting the metrics by top tenants.
Nova - Users
Provides graphs on CPU, RAM, disk throughput, IOPS, and space usage and allocation and allows sorting the metrics by top users.
Nova - Utilization
Provides detailed graphs on Nova hypervisor resources capacity and consumption.
Memcached
Memcached Prometheus exporter dashboard. Monitors Kubernetes Memcached pods and displays memory usage, hit rate, evicts and reclaims rate, items in cache, network statistics, and commands rate.
MySQL
MySQL Prometheus exporter dashboard. Monitors Kubernetes MySQL pods, resources usage and provides details on current connections and database performance.
RabbitMQ [Deprecated]
Not recommended. Deprecated since MOSK 25.1. RabbitMQ Prometheus exporter dashboard. Monitors Kubernetes RabbitMQ pods, resources usage and provides details on cluster utilization and performance.
Caution
This dashboard is renamed from RabbitMQ to RabbitMQ [Deprecated] in MOSK 25.1 and will be removed in one of the following releases for the sake of the RabbitMQ Overview and RabbitMQ Erlang dashboards.
For deprecation details, see Deprecation notes: RabbitMQ Prometheus Exporter.
RabbitMQ Erlang
Available since MOSK 25.1. Monitors RabbitMQ BEAM performance, memory details, load and distribution metrics using native Prometheus plugin metrics.
RabbitMQ Overview
Available since MOSK 25.1. Monitors RabbitMQ node performance, resource usage, message queue, channel, and connection statistics using native Prometheus plugin metrics.
Cassandra
Provides graphs on Cassandra clusters’ health, ongoing operations, and resource consumption.
Kafka
Provides graphs on Kafka clusters’ and broker health, as well as broker and topic usage.
Redis
Provides graphs on Redis clusters’ and pods’ health, connections, command calls, and resource consumption.
Tungsten Fabric
Tungsten Fabric Controller
Provides graphs on the overall Tungsten Fabric Controller cluster processes and usage.
Tungsten Fabric vRouter
Provides graphs on the overall Tungsten Fabric vRouter cluster processes and usage.
ZooKeeper
Provides graphs on ZooKeeper clusters’ quorum health and resource consumption.
StackLight
Alertmanager
Provides performance metrics on the overall health status of the Prometheus Alertmanager service, the number of firing and resolved alerts received for various periods, the rate of successful and failed notifications, and the resources consumption.
OpenSearch
Provides information about the overall health status of the OpenSearch cluster, including the resources consumption, number of operations and their performance.
OpenSearch Indices
Provides detailed information about the state of indices, including their size, the number and the size of segments.
Grafana
Provides performance metrics for the Grafana service, including the total number of Grafana entities, CPU and memory consumption.
PostgreSQL
Provides PostgreSQL statistics, including read (DQL) and write (DML) row operations, transaction and lock, replication lag and conflict, and checkpoint statistics, as well as PostgreSQL performance metrics.
Prometheus
Provides the availability and performance behavior of the Prometheus servers, the sample ingestion rate, and system usage statistics per server. Also, provides statistics about the overall status and uptime of the Prometheus service, the chunks number of the local storage memory, target scrapes, and queries duration.
Prometheus Relay
Provides service status and resources consumption metrics.
Telemeter Server
Provides statistics and the overall health status of the Telemeter service.
Note
Due to the known issue, the Telemeter Client Status panel of the Telemeter Server dashboard does not display data for managed clusters of the 11.7.0, 11.7.4, 12.5.0, and 12.7.x series Cluster releases after update to Container Cloud 2.24.0.
System
System
Provides a detailed resource consumption and operating system information per Container Cloud cluster node.
Mirantis Kubernetes Engine (MKE)
MKE Cluster
Provides a global overview of an MKE cluster: statistics about the number of the worker and manager nodes, containers, images, Swarm services.
MKE Containers
Provides per container resources consumption metrics for the MKE containers such as CPU, RAM, network.
Export data from Table panels of Grafana dashboards to CSV¶
This section describes how to export data from Table panels of
Grafana dashboards to .csv files.
Note
Grafana performs data exports for individual panels on a dashboard, not the entire dashboard.
To export data from Table panels of Grafana dashboards to CSV:
Log in to the Grafana web UI as described in Getting access.
In the right top corner of the required Table panel, click the kebab menu icon and select Inspect > Data.
In Data options of the Data tab, configure export options:
Enable Apply panel transformation
Leave Formatted data enabled
Enable Download for Excel, if required
Click Download CSV.
StackLight alerts¶
This section provides an overview of the available predefined StackLight alerts, including OpenStack, Tungsten Fabric, Container Cloud, Ceph, StackLight, MKE, and other alerts that can contain information about both OpenStack and MOSK clusters.
To view the alerts, use the Prometheus web UI. To view the firing alerts, use Alertmanager or Alerta web UI.
For alert troubleshooting guidelines, see Troubleshoot alerts.
OpenStack¶
This section describes the alerts available for OpenStack services.
This section describes the alerts available for the OpenStack system services.
This section lists the alerts for the libvirt service.
Severity |
Critical |
|---|---|
Summary |
Failure to gather libvirt metrics. |
Description |
Libvirt Exporter fails to gather metrics on the
|
Severity |
Major |
|---|---|
Summary |
Libvirt Exporter Prometheus target is down. |
Description |
Prometheus fails to scrape metrics from the Libvirt Exporter endpoint
on the |
Severity |
Critical |
|---|---|
Summary |
Libvirt Exporter Prometheus targets outage. |
Description |
Prometheus fails to scrape metrics from all Libvirt Exporter endpoints. |
This section lists the alerts for the MariaDB service.
Severity |
Critical |
|---|---|
Summary |
MariaDB cluster is down. |
Description |
The MariaDB |
Replaced with MariadbExporterTargetDown in 23.3
Severity |
Critical |
|---|---|
Summary |
MariaDB Exporter cluster Prometheus targets outage. |
Description |
Prometheus fails to scrape metrics from 2/3 of the
|
Since 23.3 to replace MariadbExporterClusterTargetsOutage
Severity |
Critical |
|---|---|
Summary |
MariaDB Exporter cluster Prometheus target down. |
Description |
Prometheus fails to scrape metrics from the
|
Severity |
Warning |
|---|---|
Summary |
MariaDB node is falling behind. |
Description |
The |
Severity |
Major |
|---|---|
Summary |
MariaDB cluster node is not ready. |
Description |
The |
Severity |
Warning |
|---|---|
Summary |
MariaDB cluster node is out of sync. |
Description |
The |
Severity |
Warning |
|---|---|
Summary |
MariaDB InnoDB log writes are stalling. |
Description |
The |
Severity |
Warning |
|---|---|
Summary |
MariaDB table lock waits are high. |
Description |
The |
This section lists the alerts for the Memcached service.
Severity |
Critical |
|---|---|
Summary |
Memcached cluster is down. |
Description |
The Memcached |
Severity |
Warning |
|---|---|
Summary |
Memcached has no open connections. |
Description |
The Memcached database cluster |
Severity |
Warning |
|---|---|
Summary |
Memcached has no open connections on all nodes. |
Description |
The Memcached database cluster |
Severity |
Warning |
|---|---|
Summary |
10 Memcached evictions. |
Description |
An average of |
Since 23.3 to replace MemcachedExporterClusterTargetsOutage
Severity |
Critical |
|---|---|
Summary |
Memcached Exporter cluster Prometheus target down. |
Description |
Prometheus fails to scrape metrics from the
|
Replaced with MemcachedExporterTargetDown in 23.3
Severity |
Critical |
|---|---|
Summary |
Memcached Exporter cluster Prometheus targets outage. |
Description |
Prometheus fails to scrape metrics from 2/3 of the
|
This section describes the alerts for the OpenStack SSL certificates.
By default, these alerts are disabled. To enable them, set
openstack.externalFQDNs.enabled to true. For details, see
Configuration options for SSL certificates.
Severity |
Critical |
|---|---|
Summary |
SSL certificate for an OpenStack service expires on |
Description |
The SSL certificate for the OpenStack |
Severity |
Major |
|---|---|
Summary |
SSL certificate for an OpenStack service expires on |
Description |
The SSL certificate for the OpenStack |
Severity |
Critical |
|---|---|
Summary |
SSL certificate probes for an OpenStack service are failing. |
Description |
The SSL certificate probes for the OpenStack
|
Severity |
Critical |
|---|---|
Summary |
OpenStack |
Description |
Prometheus fails to probe the OpenStack
|
This section lists the alerts for the RabbitMQ service.
Note
Before Container Cloud 2.29.0 (Cluster releases 17.4.0 and 16.4.0),
this alert was named RabbitMQNetworkPartitionsDetected.
Severity |
Major |
|---|---|
Summary |
RabbitMQ unreachable peers detected. |
Description |
The |
Deprecated since MCC 2.29.0 (17.4.0 and 16.4.0)
Severity |
Critical |
|---|---|
Summary |
RabbitMQ is down. |
Description |
The |
Deprecated since MCC 2.29.0 (17.4.0 and 16.4.0)
Severity |
Critical |
|---|---|
Summary |
|
Description |
Prometheus fails to scrape metrics from the
|
Severity |
Major |
|---|---|
Summary |
RabbitMQ operator Prometheus target is down. |
Description |
Prometheus fails to scrape metrics from the
|
Severity |
Warning |
|---|---|
Summary |
RabbitMQ file descriptors usage is high for the last 10 minutes. |
Description |
The |
Severity |
Warning |
|---|---|
Summary |
RabbitMQ disk space usage is high. |
Description |
The |
Severity |
Major |
|---|---|
Summary |
RabbitMQ memory usage is high. |
Description |
The |
Available since MCC 2.29.0 (17.4.0 and 16.4.0)
Severity |
Major |
|---|---|
Summary |
|
Description |
Prometheus fails to scrape metrics from the |
This section describes the alerts available for the OpenStack core services.
This section describes the alerts for the OpenStack services API:
Severity |
Critical |
|---|---|
Summary |
OpenStack ingress controller Prometheus targets outage. |
Description |
Prometheus fails to scrape metrics from all OpenStack ingress controller endpoints. |
Severity |
Critical |
|---|---|
Summary |
OpenStack API responds with HTTP 401. |
Description |
The OpenStack API |
Severity |
Critical |
|---|---|
Summary |
OpenStack API responds with HTTP 5xx. |
Description |
The OpenStack API |
Available since MOSK 24.1
Severity |
Critical |
|---|---|
Summary |
OpenStack |
Description |
The OpenStack |
Severity |
Critical |
|---|---|
Summary |
OpenStack public API responds with HTTP 401. |
Description |
The OpenStack |
Severity |
Critical |
|---|---|
Summary |
OpenStack Public API responds with HTTP 5xx. |
Description |
The OpenStack |
Removed in MOSK 24.1
Severity |
Critical |
|---|---|
Summary |
OpenStack |
Description |
The OpenStack |
Removed in MOSK 24.1
Severity |
Critical |
|---|---|
Summary |
OpenStack |
Description |
The OpenStack |
This section lists the alerts for Cinder:
Severity |
Critical |
|---|---|
Summary |
|
Description |
The |
Severity |
Major |
|---|---|
Summary |
|
Description |
The |
Severity |
Critical |
|---|---|
Summary |
|
Description |
The |
This section lists the alerts for Ironic.
Severity |
Major |
|---|---|
Summary |
|
Description |
The |
This section lists the alerts for Neutron:
Severity |
Critical |
|---|---|
Summary |
|
Description |
The |
Severity |
Critical |
|---|---|
Summary |
|
Description |
The |
Severity |
Critical |
|---|---|
Summary |
|
Description |
The |
This section lists the alerts for Nova:
Available since MOSK 24.3
Severity |
Major |
|---|---|
Summary |
Openstack Nova orphaned allocations detected. |
Description |
Orphaned resource allocations are detected on compute nodes. |
Severity |
Critical |
|---|---|
Summary |
|
Description |
The |
Severity |
Critical |
|---|---|
Summary |
|
Description |
The |
Severity |
Critical |
|---|---|
Summary |
|
Description |
The |
This section lists the alerts for the Cloudprober service:
Since 23.3 to replace OpenstackCloudproberTargetsOutage TechPreview
Severity |
Major |
|---|---|
Summary |
Openstack Cloudprober Prometheus target down. |
Description |
Prometheus fails to scrape metrics from the
|
Replaced with OpenstackCloudproberTargetDown in 23.3 Available since 23.2 TechPreview
Severity |
Major |
|---|---|
Summary |
Openstack Cloudprober Prometheus targets outage. |
Description |
Prometheus fails to scrape metrics from all OpenStack Cloudprober endpoints (more than 1/10 failed scrapes). |
This section lists the alerts for the Portprober service:
Available since MOSK 24.2 TechPreview
Severity |
Major |
|---|---|
Summary |
OpenStack Portprober Prometheus targets outage. |
Description |
Prometheus fails to scrape metrics from more than one OpenStack Portprober endpoint. |
Available since MOSK 24.3
This section lists the alerts for the PowerDNS service:
Severity |
Critical |
|---|---|
Summary |
DNS probe failure for |
Description |
The DNS probe failed at least 3 times for the DNS |
Severity |
Critical |
|---|---|
Summary |
DNS probe target experienced outage for |
Description |
Prometheus failed to probe the DNS |
Severity |
Warning |
|---|---|
Summary |
High DNS query duration for |
Description |
The DNS query duration for the DNS |
Severity |
Critical |
|---|---|
Summary |
PowerDNS Prometheus target is down |
Description |
Prometheus fails to scrape metrics from the |
Severity |
Critical |
|---|---|
Summary |
Openstack PowerDNS UDP checksum errors detected |
Description |
The number of UDP checksum errors has increased by |
Severity |
Critical |
|---|---|
Summary |
Openstack PowerDNS UDP input errors detected |
Description |
The number of UDP input errors has increased by |
Severity |
Critical |
|---|---|
Summary |
Openstack PowerDNS UDP receive buffer errors detected |
Description |
The number of UDP receive buffer errors has increased by |
Severity |
Critical |
|---|---|
Summary |
Openstack PowerDNS UDP send buffer errors detected |
Description |
The number of UDP send buffer errors has increased by |
Available since MOSK 24.1
This section lists the alerts for OpenStack credential rotation:
Note
You can adjust thresholds for the alerts included in this section. If needed, refer to Alerts configuration.
Severity |
Warning |
|---|---|
Summary |
OpenStack administrator credentials are overdue for rotation |
Description |
The OpenStack administrator credentials have not been rotated since
|
Severity |
Warning |
|---|---|
Summary |
OpenStack service user credentials are overdue for rotation |
Description |
The OpenStack service user credentials have not been rotated since
|
See also
This section describes the alerts available for the OpenStack Controller (Rockoon).
Available since MOSK 24.3
Severity |
Major |
|---|---|
Summary |
Collector failure in the OpenStackDeployment Exporter. |
Description |
The |
Severity |
Critical |
|---|---|
Summary |
OpenStackDeployment Exporter Prometheus target is down. |
Description |
Prometheus fails to scrape metrics from the OpenStackDeployment Exporter endpoint. |
Severity |
Warning |
|---|---|
Summary |
SSL certificate for an OpenStack service expires on |
Description |
The SSL certificate |
Severity |
Major |
|---|---|
Summary |
SSL certificate for an OpenStack service expires on |
Description |
The SSL certificate |
Tungsten Fabric¶
This section describes the alerts available for the Tungsten Fabric services.
This section lists the alerts for Cassandra.
Severity |
Warning |
|---|---|
Summary |
Cassandra authentication failures. |
Description |
The |
Severity |
Major |
|---|---|
Summary |
Cassandra cache hit rate is too low. |
Description |
The average hit rate for the |
Severity |
Major |
|---|---|
Summary |
Cassandra client |
Description |
The |
Available since 23.3 to replace CassandraClusterTargetsOutage
Severity |
Critical |
|---|---|
Summary |
Cassandra cluster target down. |
Description |
Prometheus fails to scrape metrics from the
|
Replaced by CassandraClusterTargetDown in 23.3
Severity |
Critical |
|---|---|
Summary |
Cassandra cluster Prometheus targets outage. |
Description |
Prometheus fails to scrape metrics from 2/3 of the
|
Severity |
Warning |
|---|---|
Summary |
Cassandra |
Description |
The |
Severity |
Warning |
|---|---|
Summary |
Cassandra compaction executor tasks are blocked. |
Description |
The |
Severity |
Warning |
|---|---|
Summary |
Cassandra has too many pending compactions. |
Description |
The pending compaction tasks in the
|
Severity |
Critical |
|---|---|
Summary |
Cassandra connection timeouts. |
Description |
The |
Severity |
Warning |
|---|---|
Summary |
Cassandra flush writer tasks are blocked. |
Description |
The |
Severity |
Major |
|---|---|
Summary |
Cassandra has too many hints. |
Description |
The |
Severity |
Warning |
|---|---|
Summary |
Cassandra repair tasks are blocked. |
Description |
The |
Severity |
Critical |
|---|---|
Summary |
Cassandra storage exceptions. |
Description |
The |
Severity |
Critical |
|---|---|
Summary |
Cassandra scanned 50000 tombstones. |
Description |
The |
Severity |
Major |
|---|---|
Summary |
Cassandra scanned 25000 tombstones. |
Description |
The |
Severity |
Warning |
|---|---|
Summary |
Cassandra scanned 10000 tombstones. |
Description |
The |
Severity |
Warning |
|---|---|
Summary |
Cassandra high view/write latency. |
Description |
The |
This section lists the alerts for Kafka.
Since 23.3 to replace KafkaClusterTargetsOutage
Severity |
Critical |
|---|---|
Summary |
Kafka cluster Prometheus target down. |
Description |
Prometheus fails to scrape metrics from the
|
Replaced with KafkaClusterTargetDown in 23.3
Severity |
Critical |
|---|---|
Summary |
Kafka cluster Prometheus targets outage. |
Description |
Prometheus fails to scrape metrics from 2/3 of the
|
Severity |
Critical |
|---|---|
Summary |
Kafka cluster has missing brokers. |
Description |
The |
Severity |
Critical |
|---|---|
Summary |
Kafka cluster controller is missing. |
Description |
The |
Severity |
Critical |
|---|---|
Summary |
Unavailable partitions in Kafka cluster. |
Description |
Partitions without a primary replica have been detected in the
|
Severity |
Critical |
|---|---|
Summary |
Kafka cluster has too many controllers. |
Description |
The |
Severity |
Major |
|---|---|
Summary |
Unclean Kafka broker was elected as cluster leader. |
Description |
A Kafka broker that has not finished the replication state has been
elected as leader in |
Severity |
Warning |
|---|---|
Summary |
Kafka cluster has underreplicated partitions. |
Description |
The topics in the |
This section lists the alerts for Redis.
Severity |
Major |
|---|---|
Summary |
Redis cluster is flapping. |
Description |
Changes have been detected in the |
Since 23.3 to replace RedisClusterTargetsOutage
Severity |
Major |
|---|---|
Summary |
Redis cluster Prometheus targets outage. |
Description |
Prometheus fails to scrape metrics from the
|
Replaced with RedisClusterTargetDown in 23.3
Severity |
Major |
|---|---|
Summary |
Redis cluster Prometheus targets outage. |
Description |
Prometheus fails to scrape metrics from 2/3 of the
|
Severity |
Warning |
|---|---|
Summary |
Redis has disconnected replicas. |
Description |
The |
Severity |
Critical |
|---|---|
Summary |
Redis Pod is down. |
Description |
The |
Severity |
Critical |
|---|---|
Summary |
Redis cluster has no primary node. |
Description |
The |
Severity |
Major |
|---|---|
Summary |
Redis has multiple primaries. |
Description |
The |
Severity |
Major |
|---|---|
Summary |
Redis cluster has rejected connections. |
Description |
Some connections to the |
Severity |
Major |
|---|---|
Summary |
Redis replication is broken. |
Description |
The |
This section lists alerts for the Tungsten Fabric Operator.
Available since MOSK 23.3
Severity |
Critical |
|---|---|
Summary |
Tungsten Fabric Operator Prometheus target is down. |
Description |
Prometheus fails to scrape metrics from the Tungsten Fabric Operator metrics service. |
This section lists the alerts for Tungsten Fabric.
Severity |
Critical |
|---|---|
Summary |
Tungsten Fabric API responds with HTTP 401. |
Description |
The Tungsten Fabric API responds with HTTP 401 for more than 5% of requests for the last 10 minutes. |
Severity |
Critical |
|---|---|
Summary |
Tungsten Fabric API responds with HTTP 5xx. |
Description |
The Tungsten Fabric API responds with HTTP 5xx for more than 1% of requests for the last 10 minutes. |
Severity |
Warning |
|---|---|
Summary |
Tungsten Fabric BGP sessions are down. |
Description |
|
Severity |
Warning |
|---|---|
Summary |
No active Tungsten Fabric BGP sessions. |
Description |
There are no active Tungsten Fabric BGP sessions on the
|
Severity |
Warning |
|---|---|
Summary |
No established Tungsten Fabric BGP sessions. |
Description |
There are no established Tungsten Fabric BGP sessions on the
|
Severity |
Warning |
|---|---|
Summary |
Tungsten Fabric Controller is down. |
Description |
The Tungsten Fabric Controller on the |
Severity |
Critical |
|---|---|
Summary |
All Tungsten Fabric Controllers are down. |
Description |
All Tungsten Fabric Controllers are down for 2 minutes. |
Severity |
Critical |
|---|---|
Summary |
Tungsten Fabric Controller Prometheus targets outage. |
Description |
Prometheus fails to scrape metrics from 2/3 of the Tungsten Fabric Controller exporter endpoints. |
Severity |
Warning |
|---|---|
Summary |
Tungsten Fabric vRouter is down. |
Description |
The Tungsten Fabric vRouter on the |
Severity |
Warning |
|---|---|
Summary |
Tungsten Fabric vRouter LLS sessions changes reached the limit of 5. |
Description |
The Tungsten Fabric vRouter LLS sessions on the |
Severity |
Warning |
|---|---|
Summary |
Tungsten Fabric vRouter LLS sessions reached the limit of 10. |
Description |
|
Severity |
Critical |
|---|---|
Summary |
Tungsten Fabric metadata is unavailable. |
Description |
The Tungsten Fabric metadata on the |
Severity |
Critical |
|---|---|
Summary |
All Tungsten Fabric vRouters are down. |
Description |
All Tungsten Fabric vRouters are down for 2 minutes. |
Severity |
Major |
|---|---|
Summary |
Tungsten Fabric vRouter Prometheus target is down. |
Description |
Prometheus fails to scrape metrics from the Tungsten Fabric vRouter
exporter endpoint on the |
Severity |
Critical |
|---|---|
Summary |
Tungsten Fabric vRouter Prometheus targets outage. |
Description |
Prometheus fails to scrape metrics from all Tungsten Fabric vRouter exporter endpoints. |
Severity |
Warning |
|---|---|
Summary |
Tungsten Fabric vRouter XMPP sessions changes reached the limit of 5. |
Description |
The Tungsten Fabric vRouter XMPP sessions on the |
Severity |
Warning |
|---|---|
Summary |
Tungsten Fabric vRouter XMPP sessions reached the limit of 10. |
Description |
|
Severity |
Warning |
|---|---|
Summary |
No Tungsten Fabric vRouter XMPP sessions. |
Description |
There are no Tungsten Fabric vRouter XMPP sessions on the
|
Severity |
Warning |
|---|---|
Summary |
Tungsten Fabric XMPP sessions changes reached the limit of 100. |
Description |
The Tungsten Fabric XMPP sessions on the |
Severity |
Warning |
|---|---|
Summary |
Tungsten Fabric XMPP sessions are down. |
Description |
|
Severity |
Warning |
|---|---|
Summary |
Missing Tungsten Fabric XMPP sessions. |
Description |
|
Severity |
Warning |
|---|---|
Summary |
Missing established Tungsten Fabric XMPP sessions. |
Description |
|
Severity |
Warning |
|---|---|
Summary |
Tungsten Fabric XMPP sessions reached the limit of 500. |
Description |
|
This section lists the alerts for ZooKeeper.