Addressed issues

The following issues have been addressed in the MOSK 26.2 release:

Bare metal

[63019] Resolved the issue with the AnsibleExtra configuration that prevented management cluster upgrade with one of the machines remaining in the prepare phase.
[63231] Resolved CVE-2026-46242 in the Linux kernel.

[56677] Resolved the issue that prevented management node replacement when automated node cleaning failed.
[58602] Resolved the issue that caused Kubernetes API server profiling to remain enabled in the MKE configuration after being disabled in the Cluster object.
[60956] Added /var/log/warn into the logrotate configuration to prevent disk space issues.

[62314] Resolved the issue that caused failure to add a new BGP router using the OpenSDN web UI.
[62399] Resolved the issue that caused the DNS backend to split during recreation of a virtual machine with the same DNS name.
[63740] Resolved the issue that caused regression to appear in the heat-engine pods causing issues with logical router updates due to the missing ref_uuid_list handling.

[Epoxy] [56111] Resolved the issue that caused server rebuild to fail with VirtualInterfaceCreateException (fault code 500) due to a race condition triggered by Nova creating Neutron ports without device_owner. Without it, Neutron skipped DHCP provisioning and later changed the port status to ACTIVE during instance teardown, causing both initial builds and rebuilds to enter the ERROR state.
[Epoxy] [OVS] [58690] Resolved the issue that caused floating IP deletion to fail due to NotFound errors from Designate when deleting PTR records or zones.
[Epoxy] [59184] Resolved the issue that caused Horizon to incorrectly display DHCP as enabled for a subnet in the Subnet Details tab of the Edit Subnet dialog when DHCP was actually disabled.
[Epoxy] [59296] Resolved the issue that caused creation of a server with security groups attached to fail when a project contained multiple security groups with the same name.
[Epoxy] [60381, 61640] Improved Pure Storage support in the Block Storage service (OpenStack Cinder).

[48581] Resolved the issue that caused the OpenSearchClusterStatusCritical alert to fire during the cluster update with StackLight in HA mode when the next OpenSearch node restarted before shards from the previous node finished assigning.
[64398] Resolved the issue that caused the MdSyncStuck alert to fire in bulk for RAID arrays that were healthy but queued behind an active monthly mdadm consistency check, rather than genuinely stuck.

[Ceph] [60923] Resolved the issue that caused Ceph OSD to ignore multipath devices with metadata devices and leave them unconfigured regardless of being discovered.
[Cluster update] [59171] Resolved the issue that caused the sync operation to fail with the Size mismatch and Digest mismatch errors during the preparation of an offline copy for an air-gapped cluster update.
[IAM] [62285] Resolved the issue that caused the iam-keycloak Pods to hang in the Init state while waiting for the iam-cluster-wait job to complete, even though the job had already completed successfully.
[Management console] [58691] Resolved the issue where updating tlsConfig using the MOSK management console incorrectly set dedicatedControlPlane: true.