kubernetes / k8s.io

Code and configuration to manage Kubernetes project infrastructure, including various *.k8s.io sites
https://git.k8s.io/community/sig-k8s-infra
Apache License 2.0
690 stars 782 forks source link

N2 Quota changes for Scale Projects #6271

Open upodroid opened 5 months ago

upodroid commented 5 months ago

Kubernetes project uses E2 instances on GCP by default unless we are testing something that requires specific instance types(GPU tests, scale perf testing, arm64)

k/k change: https://github.com/kubernetes/kubernetes/pull/118626

With E2, the VMs issued by Google can run on modern AMD Epyc or ancient Intel Skylake hosts. However, scale job control plane nodes need to run on high-performance instances consistently and will be using N2 machine types with Ice Lake CPUs.

However, N2 quotas are not set properly and this issue will track quota failures from k8s-infra-e2e-scale-project-XX and fix them as reported.

Quotas for E2 CPUs will be bumped to 1000 in us-east1. Please ensure that jobs are running in this location

/sig testing /sig scalability /priority critical-urgent

marseel commented 5 months ago

Related failure in PR: https://github.com/kubernetes/perf-tests/pull/2494 Example run: https://prow.k8s.io/view/gs/kubernetes-jenkins/pr-logs/pull/perf-tests/2494/pull-perf-tests-clusterloader2/1744323040390418432 Project: k8s-presubmit-scale-36

 - Quota 'N2_CPUS' exceeded.  Limit: 24.0 in region us-east1.
    metric name = compute.googleapis.com/n2_cpus
    limit name = N2-CPUS-per-project-region
    limit = 24.0
    dimensions = region: us-east1
upodroid commented 5 months ago

projects starting with k8s-* are part of the google.com org that we don't manage. Please migrate those projects to the community infrastructure.

k8s-triage-robot commented 2 months ago

The Kubernetes project currently lacks enough contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

You can:

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

k8s-triage-robot commented 1 month ago

The Kubernetes project currently lacks enough active contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

You can:

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle rotten