Kubernetes: Kubelet Probe Regression Gets a Full Fleet Backport

A kubelet probe cancellation bug from the pod worker context change is being fixed and cherry-picked across three release branches, while scheduling picks up new gang-scheduling flexibility and several e2e reliability fixes land to cut down on flaky tests.

Duration: PT2M54S

Episode overview

This episode is a short developer briefing from Kubernetes.

It explains recent repository work in plain language.

  • Show: Kubernetes
  • Published: 2026-07-29T13:03:05Z
  • Audio duration: PT2M54S

Transcript excerpt

This excerpt keeps the crawler page concise. Listen to the episode or use the RSS feed for the full update.

Good morning, it's July 29th, and here's your Kubernetes briefing.

The top story today is a probe reliability regression that's touching nearly every active release branch. Since a change in version 1.35, probe workers inherited the pod worker's sync context, so the moment a pod started terminating, that context got cancelled — killing exec probes mid-flight and leaving the pod's…

Second theme: scheduling is getting more flexible and more deterministic. Jdzikowski's PR 141023 makes minimum group count mutable for composite pod groups, extending work already done for pod group mutability, and a companion PR, 141021, adds finalizer-based garbage collection protection so parent groups clean up…

Third theme: hardening the end-to-end test framework itself. Kairosci landed four related PRs — 140997, 140998, 140996, and 140995 — fixing silently swallowed errors, adding retry logic for transient API server failures, improving daemon readiness checks, and closing a PVC deletion race during stateful set cleanup.…

Smaller but worth noting: client-go's fake custom store had its cache store conformance restored in PR 140966, already cherry-picked to 1.36, and the event broadcaster…

Expect the…

Nearby episodes from Kubernetes

  1. The Great Test Cleanup, Plus Scheduler and Networking Fixes
  2. Correctness Fixes Ripple Across Kubelet and Networking
  3. Weekly Recap - Scheduling Gets a Major Rework
  4. Scheduler Speed and Cleanup Fixes
  5. Reliability Fixes and API Memory Cleanup
  6. DRA Hardening and the API Server's Fine Print
  7. Feature Graduation Day
  8. Kubelet Resource Accounting Gets a Reliability Pass