Enhancements¶
This section outlines new features implemented in the Cluster release 14.0.0 that is introduced in the Container Cloud release 2.24.0.
Dropping of redundant Ceph components from management and regional clusters
Major version update of OpenSearch and OpenSearch Dashboards
Support for MKE 3.6.5 and MCR 20.10.17¶
Introduced support for Mirantis Container Runtime (MCR) 20.10.17 and Mirantis Kubernetes Engine (MKE) 3.6.5 that supports Kubernetes 1.24 for the Container Cloud management, regional, and managed clusters. On existing clusters, MKE and MCR are updated to the latest supported version when you update your managed cluster to the Cluster release 14.0.0.
Caution
Support for MKE 3.5.x is dropped. Therefore, new deployments on MKE 3.5.x are not supported.
Note
For MOSK-based deployments, the feature support is available since MOSK 23.2.
Automatic upgrade of Ceph from Pacific to Quincy¶
Upgraded Ceph major version from Pacific 16.2.11 to Quincy 17.2.6 with an automatic upgrade of Ceph components on existing managed clusters during the Cluster version update.
Note
For MOSK-based deployments, the feature support is available since MOSK 23.2.
Dropping of redundant Ceph components from management and regional clusters¶
As the final part of Ceph removal from Container Cloud management clusters, which reduces resource consumption, removed the following Ceph components that were present on clusters for backward compatibility:
Helm chart of the Ceph Controller (
ceph-operator
)Ceph deployments
Ceph namespaces
ceph-lcm-mirantis
androok-ceph
Learn more
Monitoring of network connectivity between Ceph nodes¶
Introduced healthcheck metrics and the following Ceph alerts to monitor network connectivity between Ceph nodes:
CephDaemonSlowOps
CephMonClockSkew
CephOSDFlapping
CephOSDSlowClusterNetwork
CephOSDSlowPublicNetwork
Note
For MOSK-based deployments, the feature support is available since MOSK 23.2.
Learn more
Improvements to StackLight alerting¶
Implemented the following improvements to StackLight alerting:
Changed severity for multiple alerts to increase visibility of potentially workload-impacting alerts and decrease noise of non-workload-impacting alerts
Renamed
MCCLicenseExpirationCritical
toMCCLicenseExpirationHigh
,MCCLicenseExpirationMajor
toMCCLicenseExpirationMedium
For Ironic:
Removed
IronicBmMetricsMissing
in favor ofIronicBmApiOutage
Removed inhibition rules for
IronicBmTargetDown
andIronicBmApiOutage
Improved expression for
IronicBmApiOutage
For Kubernetes applications:
Reworked troubleshooting steps for
KubeStatefulSetUpdateNotRolledOut
,KubeDeploymentOutage
,KubeDeploymentReplicasMismatch
Updated descriptions for
KubeStatefulSetOutage
andKubeDeploymentOutage
Changed expressions for
KubeDeploymentOutage
,KubeDeploymentReplicasMismatch
,CephOSDDiskNotResponding
, andCephOSDDown
Learn more
Major version update of OpenSearch and OpenSearch Dashboards¶
Updated OpenSearch and OpenSearch Dashboards from major version 1.3.7 to 2.7.0. The latest version includes a number of enhancements along with bug and security fixes.
Note
For MOSK-based deployments, the feature support is available since MOSK 23.2.
Caution
The version update process can take up to 20 minutes, during
which both OpenSearch and OpenSearch Dashboards may become temporarily
unavailable. Additionally, the KubeStatefulsetUpdateNotRolledOut
alert
for the opensearch-master
StatefulSet may fire for a short period of time.
Note
The end-of-life support of the major version 1.x ends on December 31, 2023.
Performance tuning of Grafana dashboards¶
Tuned the performance of Grafana dashboards for faster loading and a better UX by refactoring and optimizing different Grafana dashboards.
This enhancement includes extraction of the OpenSearch Indices dashboard out of the OpenSearch dashboard to provide detailed information about the state of indices, including their size, the size of document values and segments.
Learn more
Dropped and white-listed metrics¶
To improve Prometheus performance and provide better resource utilization with faster query response, dropped metrics that are unused by StackLight. Also created the default white list of metrics that you can expand.
The feature is enabled by default using the
prometheusServer.metricsFiltering.enabled:true
parameter. Thus, if you
have created custom alerts, recording rules, dashboards, or if you were
actively using some metrics for different purposes, some of those metrics can
be dropped. Therefore, verify the white list of Prometheus scrape jobs to ensure that the required metrics are not dropped.
If a job name that relates to the required metric is not present in this list,
its target metrics are not dropped and are collected by Prometheus by default.
If the required metric is not present in this list, you can whitelist it
using the prometheusServer.metricsFiltering.extraMetricsInclude
parameter.