Kubernetes: The API Server Learns to Stop Choking Slow Watchers

A five-PR series from contributor nami-ant reworks how the API server handles slow watch clients, moving from forced disconnects toward a stall-and-resume model, while two memory-pinning fixes and a webhook cache security patch round out a day focused on control plane resilience.

Duration: PT2M35S

Episode overview

This episode is a short developer briefing from Kubernetes.

It explains recent repository work in plain language.

  • Show: Kubernetes
  • Published: 2026-08-19T13:03:34Z
  • Audio duration: PT2M35S

Transcript excerpt

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

You're listening to Kubernetes, your developer briefing for August 19th, 2026.

Today's biggest story is a coordinated effort to fix how the API server treats watch clients that fall behind. Right now, a slow reader can block the shared dispatch path for other watchers, and if it doesn't catch up, the server force-closes it. PRs 141448, 141449, 141450, 141451, and 141452 build this out step by…

Second theme: memory discipline in the API server. Two separate PRs, 141454 and 141456, both cap the buffer pool used for encoding large protobuf responses — one at 256 kilobytes, one at 128 — to stop bursts of big responses from permanently pinning tens of megabytes of heap. Having two competing cap values open at…

Third, a security fix: PR 141458 closes a cache-poisoning risk in the webhook authorizer, where the caching gate wasn't accounting for label and field selectors when deciding whether a response was safe to cache.

Elsewhere, smaller but meaningful bug fixes: 141443 stops kubelet from aliasing volume data across pods during resize actions, 141459 fixes stale trust bundle cache entries keyed by name, and 141438 removes duplicate service account storage construction in the API server.

Wha…

Nearby episodes from Kubernetes

  1. Closing the Gaps Between List and Watch
  2. Test Coverage and Interface Cleanup Day
  3. Cleaning Up the Foundations
  4. Kubelet Hardening and the Slow March Off k8s.io/kubernetes
  5. Correctness Fixes Across the Stack
  6. Scheduler Cleanup and List Consistency Fixes
  7. Weekly Recap - Overflow Fixes and Admission Policy Hardening
  8. Correctness Fixes Across the Stack