WhatBrokeToday
GTA 6FAQ
Kubernetes

Kubernetes Topology Manager Leaks container_aligned_compute_resources_count on Retries

minor

Is it fixed? Not yet.

Ongoing since Nov 19, 00:03 UTC. First reported Nov 19, 00:03 UTC.

What happened

The Kubernetes topology manager incorrectly increments the container_aligned_compute_resources_count metric during deferred pod updates because it runs before the PredicateAdmitHandler. This causes the metric to grow continuously on every retry of a deferred pod update using the single-numa-node policy.

What caused it?

The topology manager admit handler executes too early during pod updates (before PredicateAdmitHandler), admitting the update and incrementing the container scope counter for every container on every retry.

Source: github.com

No verified workaround yet

None of the linked reports describe a mitigation. This page updates when one appears.

Official fix

Timeline

  1. Investigating(current)
  2. Workaround available(upcoming)
  3. Resolved(upcoming)
  1. First report on GitHub issue

    Topology manager: `container_aligned_compute_resources_count` grows on every retry of a deferred pod update

    Source

Evidence

1 report · 1 GitHubShow
  1. [1]
    GitHub First report
    Topology manager: `container_aligned_compute_resources_count` grows on every retry of a deferred pod update(opens GitHub issue in a new tab)

    What happened? Follow-up to #143400, reported in this review comment. The topology manager counts container_aligned_compute_resources_count{boundary="numa_node"} inside Admit. For a pod update (UpdateOperation, e.g. an in-place resize) this is too early: The topology manager admit handler runs before PredicateAdmitHandler. With the single-numa-node policy, the topology manager admits the update and increments the counter: container scope: once for every container of the pod…

    by lukaszwojciechowski 0upvotes 2comments

FAQ

What happened in the Kubernetes issue?

The Kubernetes topology manager incorrectly increments the container_aligned_compute_resources_count metric during deferred pod updates because it runs before the PredicateAdmitHandler. This causes the metric to grow continuously on every retry of a deferred pod update using the single-numa-node policy.

What is affected by the Kubernetes issue?

Kubernetes.

When did the Kubernetes issue start and end?

It started Nov 19, 00:03 UTC and is still ongoing (4 h 41 min so far). This page updates as it develops.

How do I get alerts for Kubernetes?

Subscribe to the Kubernetes RSS feed for every new Kubernetes incident, or connect an AI assistant to WhatBrokeToday over MCP and ask whether Kubernetes is down. Kubernetes feed

Where does this information come from?

WhatBrokeToday collects public status pages, developer community reports and security advisories, groups the ones about the same problem, and links every report under Evidence on this page. This one has 1 report from 1 independent source.