Enhancements#
Detail on the enhancements introduced in MKE 4.2.0 includes:
Windows worker nodes#
You can now run Windows worker nodes in an MKE 4 cluster.
- Fresh installs can join Windows workers, including in air-gapped environments using mirrored Windows images and the offline image bundle.
- gMSA (group Managed Service Accounts) is supported for Windows workloads on both fresh installs as well as MKE 3 to MKE 4 upgrades.
- Existing MKE 3 clusters with Windows workers can be upgraded to MKE 4.
- Windows ingress is supported.
Optional FIPS 140-3 mode#
MKE 4.2.0 can be installed in a FIPS 140-3 configuration.
- A new FIPS mode field in the
mke4.yamlconfiguration file, with validation, controls FIPS installs. - FIPS builds are provided for the MKE operator and the etcd maintenance service, and the FIPS k0s binary is synchronized to hosts during install, including for air-gapped clusters.
- The Dex REST API server runs in a FIPS configuration when FIPS mode is enabled.
Run without an external load balancer#
Clusters that do not have an external load balancer in front of the control plane can now use the built-in Control Plane Load Balancer (CPLB). In providing CPLB, MKE 4 makes it so that a single virtual IP can front the manager nodes.
VMware vSphere child clusters#
MKE 4 can provision and manage child clusters on VMware vSphere, including in air-gapped environments. A vSphere cluster template is included for child-cluster deployments.
High Availability (HA) for system components#
The built-in MKE system components now default to an HA configuration so that the loss of a single node does not take down the management plane.
- Critical single-replica components now run with multiple replicas by default.
PodDisruptionBudgets,PriorityClasses, node selectors, and topology spread constraints are applied to the system components so they are scheduled across manager nodes and protected during voluntary disruptions.
Cluster node lifecycle commands#
- The
mkectl node addandmkectl node removecommands add and drain/remove worker nodes, respectively, using lease-based cluster state, without the need to re-run a fullmkectl applycommand. - The
mkectl statuscommand reports the current state of the cluster and its addons from the command line. - The
mkectl applycommand now waits for all addons to reach a ready state before returning.
etcd maintenance service#
The etcd maintenance service:
- Now runs on child clusters.
- Automatically raises the etcd storage quota when etcd enters a NOSPACE alarm, instead of requiring manual intervention.
- Records an audit trail of maintenance actions.
- Rolls back if an MKE 4 to MKE 4+ upgrade fails, to prevent etcd from being left in a half-changed state.
CSI Secrets Store driver#
The CSI Secrets Store driver is integrated into MKE 4.x, so secrets from an external store can be mounted into Pods through a CSI volume.
Configurable resource limits and Prometheus retention#
- Resource requests and limits for system components are configurable through
the
mke4.yamlconfiguration file. - Prometheus data retention is now configurable.
Addon status in the MKE API#
- k0rdent service statuses for MKE addons are surfaced in the MKE configuration status API.
- The
mkectl applycommand now waits on the reported MKE addon status instead of a fixed Helm wait, so the command reflects the real readiness of each addon.
MKE Dashboard improvements#
- Self-service password change for local users.
- You can exec into a Pod from the MKE Dashboard.
- New MKE Dashboard pages:
- Envoy Gateway controller
- Internal gateways configuration
- Gateway settings section
- Updated Ingress page behavior.
- The MKE Dashboard honors the
localUserssetting and hides or blocks direct user management when local users are disabled. - Visual alignment with the k0rdent product family, a refactored login screen, a "Cancel" button on the YAML edit view, and display of the k0rdent version.
Update of kube-prometheus-stack to 86.2.3#
MKE 4.2.0 upgrades the kube-prometheus-stack from chart 79.5.0 to 86.2.3.
CRD upgrade
Mirantis recommends that you upgrade the CRDs for
kube-prometheus-stack v0.86.2 before you upgrade your cluster to MKE
4.2.0. For detailed instruction, refer to Upgrade
Considerations.