Kubernetes Ships May Patch Wave Across 1.33–1.36, Fixes IPv6 ServiceCIDR Allocation Bug

Kubernetes shipped its scheduled May patch wave: v1.36.1, v1.35.5, v1.34.8, and v1.33.12, all built off the May 8 cherry-pick deadline with target dates of May 12–13. No CVEs were rolled in this cycle, but the v1.34.8 changelog includes a handful of fixes that production operators should read before skipping the upgrade.

Notable fixes

  • IPv6 ServiceCIDR allocation correctness (#138385). A bug allowed 64-bit IPv6 ServiceCIDRs to allocate addresses outside the configured subnet range. If you run dual-stack or IPv6-only services, validate your allocations after upgrading.
  • Scheduler in-flight queue cleanup (#138435). Stale state could persist after failed scheduling attempts, with cascading effects on subsequent decisions.
  • Kube-proxy large-cluster sync (#138637). Full-sync was being skipped under “large cluster mode” in cases where it shouldn’t have been.
  • Windows HNS endpoint cleanup (#138601). Addresses DNS timeouts on L2Bridge networks — a long-standing pain point for Windows nodes.
  • Kubeadm and kubelet fixes (#138685 and others). Better handling of delayed load balancers during init, etcd cluster status checks, LocalAPIEndpoint defaulting, and pod-startup SLI metric accuracy.

Support context

v1.34 enters maintenance mode on August 27, 2026, with end-of-life on October 27. Operators still on v1.33 should be planning their move; the 1.33.12 patch is a continuity release rather than a destination.

Sources: Kubernetes patch release schedule, CHANGELOG-1.34 — May 12–13, 2026.

Cloud Native news weekly

Stay on top of the cloud-native release wire

Kubernetes, AI infra, and CNCF moves — delivered when they matter.