June 2026

Managed Kubernetes clusters now follow a support lifecycle for Kubernetes versions, keeping clusters on supported and secure versions.

each Kubernetes version now has three lifecycle dates: maintenance start, end-of-life (EOL), and automatic upgrade
notifications appear in the portal and by email as a cluster approaches each date
when a cluster reaches its automatic upgrade date and a safe upgrade path exists, the cluster is upgraded to a supported version, using the configured maintenance window where possible

Current version status and upgrade dates are shown in the Managed Kubernetes section of the portal.

Note: a cluster that fails auto-upgrade three times is flagged for unmanaged conversion, which starts the offboarding notice sequence.

Cluster API access can now stay reachable through DNS endpoints instead of a single IP, reducing disruption during failover.

Portal, and kubeconfig generation support public and private DNS endpoints in the form public.<zone>:6443 / private.<zone>:6443
enabled by default for new clusters
existing IP-based kubeconfigs remain fully compatible

NewManaged Kubernetes

L2 Segments can now be organized and access-controlled through IAM the same way other resources are.

a resource group type — Aggregative or Resource-based — must be chosen on creation
a default aggregative resource group is created automatically and existing L2 Segments assigned to it
L2 Segment pages now respect IAM permissions for Hosts, Kubernetes nodes, and Direct Connect

NewL2 SegmentsPlatform

New load balancers can no longer be created with a protocol combination known to cause issues.

creating or editing an L7 load balancer is now blocked when GRPC/GRPCs protocols are combined with HTTP/2 enabled

Note: existing load balancers already using this combination are unaffected and continue to work.

ImprovedLoad Balancer

Customers interested in AI Compute can now request access directly from the Customer Portal instead of contacting sales separately.

NewPlatform

Dedicated server reinstalls can now be scripted as part of infrastructure-as-code workflows.

Reinstalling the operating system on dedicated servers is now supported through the Terraform provider.

NewPlatform