Release notes in Managed Kubernetes

2026

July

added information about kernel modules used in pre-installed drivers for GPU node groups. Learn more in the Drivers for GPU node groups in a Managed Kubernetes cluster instruction.

June

added support for user data in clusters on dedicated servers. Learn more about user data in the User data in Managed Kubernetes node groups instruction;
updated the list of applications available for installation in a Managed Kubernetes cluster through the control panel — added the ability to install the Envoy Gateway application and removed the NGINX Ingress Controller installation.

May

added a new cluster type: a fault-tolerant multi-zone cluster. Available only in the multi-zone pool ru-6. Read more about cluster types in the Cluster types subsection of the About Managed Kubernetes instruction;
added mks service events related to the Envoy Gateway application to audit logs.

April

updated the list of supported Kubernetes versions;
updated the containerd configuration storage format on worker nodes — the configuration is now stored in the /etc/containerd/certs.d/docker.io/ directory in the hosts.toml file. This allows adding directories for custom registries and configuring them separately, for example, specifying a CA certificate for self-signed TLS or disabling certificate verification. Learn more about configuring containerd registries via the hosts.toml file in the containerd documentation on GitHub.

For Managed Kubernetes clusters version 1.31 and higher, the new containerd configuration storage format is applied by default upon cluster creation. In existing clusters, the changes will be applied automatically upon worker node reinstallation or when creating a new node group.

March

added mks service events related to the integration with Container Registry to audit logs;
updated the Karpenter provider — Karpenter now correctly calculates the amount of memory available after operating system boot and selects flavors with sufficient resources to run pods;
added the cni_type parameter to the selectel_mks_cluster_v1 resource — the CNI plugin type (CNI) can now be selected when creating a cluster via Terraform.

February

added mks service events to audit logs;
lifted the network card limitation for dedicated server clusters — you can now use network cards with connection speeds of 10 Gbps and 25 Gbps;
added support for ephemeralStorage when autoscaling with Cluster Autoscaler and Karpenter;
added the ability to view user data in the control panel that you specified when creating a cloud server cluster or adding a new node group;
added the pigz utility to the worker node image, which speeds up container image unpacking through multi-threaded layer processing. The containerd runtime automatically uses the pigz utility. No additional setup or configuration is required.

January

updated the autoscaling process with Karpenter:
- accelerated the scaling process by creating multiple NodeClaims in parallel;
- added support for UserData in NodeClass;
- added PriorityClassName to the Karpenter Helm chart.

December

updated the list of Kubernetes versions: added version 1.34 and removed Kubernetes version 1.31.

The list of available versions can be viewed in the Versions subsection of the About Managed Kubernetes instruction;
added the ability to create Managed Kubernetes clusters on dedicated servers with GPUs;
added the ability to specify a CIDR for the node group subnet in a Managed Kubernetes dedicated server cluster when creating a cluster and adding a node group;
added information about the CNI plugin (CNI) you selected during cluster creation to the control panel on the cluster page.

November

preemptible node groups are now available in the St. Petersburg, Moscow, and Novosibirsk regions;
added a field to the cluster deletion form to specify the reason for deletion.