Migrate merge-blocking jobs to dedicated cluster: pull-kubernetes-node-e2e

[spiffxp]

What should be cleaned up or changed:

This is part of #18550

To properly monitor the outcome of this, you should be a member of [email protected]. PR yourself into https://github.com/kubernetes/k8s.io/blob/master/groups/groups.yaml#L603-L628 if you're not a member.

Migrate pull-kubernetes-node-e2e to k8s-infra-prow-build by adding a cluster: k8s-infra-prow-build field to the job:

• Make sure any release-branch variants are also updated
• Reference this issue, but do not "fix" it; instead leave this open to monitor the behavior of the job before/after
• Here's an example PR that did this for pull-kubernetes-integration: Migrate pull-kubernetes-integration to k8s-infra-prow-build #18814

NOTE: migrating this job is not as straightforward as some of the other #18550 issues, because we also need to:

• Switch away from a fixed GCP project to boskos-managed projects
  • Replace --gcp-project=k8s-jkns-pr-node-e2e with --gcp-project-type=gce-project
• If this turns out to break things, revert and ask for help

Once the PR has merged, note the date/time it merged. This will allow you to compare before/after behavior.

Things to watch for the job

• https://prow.k8s.io/?job=pull-kubernetes-node-e2e
  • does the job start failing more often?
  • does the job start going into error state?
• https://testgrid.k8s.io/presubmits-kubernetes-blocking#pull-kubernetes-node-e2e&graph-metrics=test-duration-minutes
  • does the job duration look worse than before? spikier than before?
• https://storage.googleapis.com/k8s-gubernator/triage/index.html?pr=1&job=pull-kubernetes-node-e2e
  • do more failures show up than before?
• https://prow.k8s.io/job-history/gs/kubernetes-jenkins/pr-logs/directory/pull-kubernetes-node-e2e
  • (can be used to answer some of the same questions as above)
• metrics explorer: CPU limit utilization for pull-kubernetes-node-e2e for 6h
  • is the job wildly underutilizing its CPU limit? if so, perhaps tune down (if uncertain, post evidence in this issue and ask)
  • (it will probably be helpful to look at different time resolutions like 1h, 6h, 1d, 1w)
• metrics explorer: Memory limit utilization for pull-kubernetes-node-e2e for 6h
  • is the job wildly underutilizing its memory limit? if so, perhaps tune down (if uncertain, post evidence in this issue and ask)
  • (it will probably be helpful to look at different time resolutions like 1h, 6h, 1d, 1w)

Things to watch for the build cluster

• prow-build dashboard 1w
  • is the build cluster scaling as needed? (e.g. maybe it can't scale because we've hit some kind of quota)
  • (it will probably be helpful to look at different time resolutions like 1h, 6h, 1d, 1w)
• prowjobs-experiment 1w
  • (shows resource consumption of all job runs, pretty noisy but putting this here for completeness)
• https://monitoring.prow.k8s.io/d/wSrfvNxWz/boskos-resource-usage?orgId=1
  • are there still plenty of free projects in the "GCE project (k8s-infra) usage status" graph?

Keep this open for at least 24h of weekday PR traffic. If everything continues to look good, then this can be closed.

/wg k8s-infra
/sig testing
/area jobs
/help

[spiffxp]

/remove-help
/assign

[spiffxp]

Opened #18915

[RobertKielty]

@spiffxp I marked this as In Progress based on #18915 having merged.

When can we call this complete?

[spiffxp]

PR merged 2020-08-19, which is too far ago to be able to cleanly show before/after data using testgrid or prow.k8s.io

From a local grafana instance I have that runs queries against k8s-gubernator:build, it looks like the job runs more reliably and with a comparable failure rate under load.

https://storage.googleapis.com/k8s-gubernator/triage/index.html?date=2020-08-30&pr=1&job=pull-kubernetes-node-e2e

A screenshot of triage from 2020-08-30 is early enough to pick up the before/after performance, and things look no worse that I can see. I'm guessing the spike of failures immediately after is unrelated, or has been corrected since then

CPU limit usage

CPU limit looks reasonable. As with other jobs, we need most of the CPU up front for building; in the case all the testing cpu usage happens on nodes spun up elsewhere. If we had a shared build we could take the CPU requirements way down.

Memory limit usage

Same story with memory limit usage

[spiffxp]

/close
I think this is good enough

Apologies for falling behind on this one, it should have been in Monitoring, and I just didn't have time to sit still and check in on it until now.

[k8s-ci-robot]

@spiffxp: Closing this issue.

In response to this:

/close
I think this is good enough

Apologies for falling behind on this one, it should have been in Monitoring, and I just didn't have time to sit still and check in on it until now.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.