Promoting Windows Unit Test CI job to release-informing

[Mark Rossetti]

Hi Release team,

I'd like to promote the CI job that runs unit tests from the k/k repo on Windows nodes to release informing.

Recently a lot of work has been done by SIG-Windows to address existing unit test failures and also bring the job duration down significantly. Promoting this to release informing should help the over all stability of Windows support on Kubernetes by identifying issues that may not show up in existing e2e test coverage and also keep future unit test failures on Windows from slowly creeping in over time.

Links:

• Process: https://github.com/kubernetes/sig-release/blob/master/release-blocking-jobs.md#promoting-or-demoting-jobs

• Prow Status: https://prow.k8s.io/?job=ci-kubernetes-unit-windows-master
• Dashboard: https://testgrid.k8s.io/sig-windows-signal#windows-unit-master 
• Triage: https://storage.googleapis.com/k8s-triage/index.html?job=ci-kubernetes-unit-windows-master
• PROW job config: https://github.com/kubernetes/test-infra/blob/9002febd24f80b2bb08501f208e1ee37163665ba/config/jobs/kubernetes-sigs/sig-windows/windows-unit.yaml#L2-L36

Thanks!
--Mark Rossetti

[Benjamin Elder]

Can we fix prow (individual run results) and testgrid pages (your link above) both failing to load first?

I'm guessing we're using unpruned raw go unit test results instead of pruned / per-package which will really stress testgrid/prow/triage, kubernetes/kubernetes has an enormous number of test cases.

It's difficult to use this signal as long as the results don't load.

https://github.com/kubernetes/kubernetes/tree/master/cmd/prune-junit-xml, we can follow up in slack to figure out how to set this up for the windows job. 

--
You received this message because you are subscribed to the Google Groups "kubernetes-sig-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-re...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/kubernetes-sig-release/SA1PR21MB4327568EDA400B7BA091CE8AC1D02%40SA1PR21MB4327.namprd21.prod.outlook.com.

[Benjamin Elder]

[re-adding d...@kubernetes.io]

Thanks for fixing this! 
Generally +1

I think Patrick Ohly had some additional concerns in slack about self-skipping tests?

On Thu, Mar 13, 2025 at 10:34 AM 'Mark Rossetti' via kubernetes-sig-release <kubernetes-...@googlegroups.com> wrote:

I submitted a fix for this yesterday.
Unfortunately, it looks like regression in some kubelet unit tests slipped in yesterday. 

I'll work on fixing this up too.

Thanks

--Mark

To view this discussion visit https://groups.google.com/d/msgid/kubernetes-sig-release/c4f25c9e-4b56-40ee-b2b1-7cc98bd1cb6en%40googlegroups.com.

[Patrick Ohly]

"'Benjamin Elder' via kubernetes-sig-release"

<kubernetes-...@googlegroups.com> writes:

> [re-adding d...@kubernetes.io]
>
> Thanks for fixing this!
> Generally +1
>
> I think Patrick Ohly had some additional concerns in slack about
> self-skipping tests?

I was just surprised about one of those (generic code in
pkg/controller/tainteviction failing on Windows?!). But that doesn't
block promoting the job because the argument that it needs to be running
and be watched to avoid regressions is sound.

--
Best Regards

Patrick Ohly
Cloud Software Architect