Kubernetes

Kubernetes

  • [27634] Fixed the issue with active VMs being killed by a container during a Virtlet pod restart.

  • [28450] Fixed the issue with runc using more memory during a container startup. For details, see the corresponding GitHub issue.

  • [28979] Added separate stages for runConformance in the Deploy - update Kubernetes cluster pipeline job to prevent the runConformance failures during the pipeline job execution.

  • [28976] Fixed the issue with a Kubernetes Node being uncordoned during a reboot.

  • [26881] Renamed the targetHosts parameter for etcd to targetHostsEtcd to prevent the Deploy - update Kubernetes cluster pipeline job from failing with the The current scope already contains a variable of the name targetHosts error.

  • [27425] Changed the CRI Proxy logging storage from /tmp to journald to prevent CRI and kubelet failures once /tmp is out of space.

  • [27534] Adjusted the logic for the Kubernetes namespaces to fix the namespaces creation and removal errors.

  • [27977] Implemented the installKubernetesClient function that is designed to install the Kubernetes resources to fix the issue with an MCP Kubernetes cluster deployment failure due to labels assigned to compute nodes.

  • [28730] Fixed the issue with containerd installation failure caused by the absent configuration directory for the containerd config.toml.

  • [28767] Fixed the issue with the etcd upgrade failure due to the tmp etcd directory not being cleared after the etcd installation. Added the default overwrite: true parameter to the extract_etcd section of etcd/server/service.sls in the etcd Salt formula.

  • [28771] Added the missing etcd v3 configenv to /var/lib/etcd/. Previously, configenv could be used only for v2 and variables for v3 must have been created manually.

  • [28953] Fixed the issue with the CNI plugins not being upgraded during the execution of the Deploy - update Kubernetes cluster pipeline job. The issue occurred because Salt formula verified only the presence of the CNI plugins installation files and not their versions. The fix adds overwrite: true to the /opt/cni/bin:archive.extracted: section of kubernetes/_common.sls in the Kubernetes Salt formula.

  • [28975] Fixed the incorrect logic of the Deploy - update Kubernetes cluster pipeline job that used grep with nodeShortName for verifying the nodes statuses. This logic caused an incorrect of the nodes for clusters containing more than 10 compute nodes. For example, not only cmp1 was selected but any other cmp1X nodes such as cmp10, cmp11 are selected as well.