Addressed issues

The following issues have been addressed in the MOSK 26.2.1 release:

  • [OpenStack] [67664] Resolved the issue that caused the Horizon database purge job to fail to start due to the missing Horizon image on the MOSK management content delivery network.

  • [OpenStack] [Gazpacho] [65683] Resolved the issue that caused the neutron-periodic-workers pod to crash-loop after migration from Neutron OVS to OVN due to stale logical routers left in the OVN Northbound database.

  • [OpenSDN] [65652] Resolved the issue that caused the OpenSDN DNS service to remove the DNS records of an instance after live or cold migration of the instance to another compute node.

  • [OpenSDN] [61618] Resolved the issue that caused ZooKeeper to produce an excessive amount of logs. ZooKeeper now logs at the INFO level by default instead of DEBUG, and you can change the level using the new zookeeper:logLevel parameter. For details, see Enable debug logs for the Tungsten Fabric services.

  • [LCM] [65611] Resolved the issue that caused the underlayKubernetesLogs collector of supportdump-controller to fail to collect per-node MKE logs on clusters running Mirantis Container Runtime 25.0.17 or later.

  • [Management cluster] [65700] Resolved the issue that caused a management cluster with custom k0s certificates to stop reconciling due to the outdated management cluster kubeconfig secret after replacing the CA certificate or changing the certificate host name.

  • [Management cluster] [65108] Resolved the issue that prevented kube-apiserver of the management cluster from applying a new CA certificate after rotation of a custom Keycloak certificate.

  • [Ceph] [66416] Resolved the issue that caused Ceph to fail to collect SMART health metrics for all OSD devices because sudo failed in the Ceph OSD containers. The issue was caused by the upstream Ceph issues 76415 and 75274.

  • [Ceph] [65246] Improved the performance of node label and annotation verification in Pelagia. On clusters with more than 400 nodes, the verification now takes up to 2 minutes instead of more than 5 minutes.

  • [Ceph] [64772] Resolved the issue with lifecycle management of Ceph OSDs whose BlueStore DB logical volumes consist of multiple LVM segments on the same device. The issue caused the CephDeploymentHealth object to report incorrect device verification for such OSDs.

  • [Air-gap] [68364] Resolved the issue that caused an offline copy containing only the k0s- and mosk- Cluster releases to lack the MKE images required by MOSK clusters. The air-gap metadata of MOSK management 2.32.1 now includes the MKE images in the image collections of the mosk- Cluster releases.

  • [Air-gap] [65885] Resolved the issue that caused the airgapped.sh sync command to fail with the 404 Client Error: Not Found error for the files of the k0s- Cluster release because airgapped-tools did not URL-encode the + sign in file paths.