[AFS] Implement cache for AFS consumed resources

[IrvingMg]

What type of PR is this?

Implements cache-based storage for AFS consumed resources, fixing race condition where consumed resources aren't visible to scheduler when entry penalties are deleted.

/kind bug

What this PR does / why we need it:

Which issue(s) this PR fixes:

Fixes #6710
Fixes #7688
Fixes #7693

Special notes for your reviewer:

Does this PR introduce a user-facing change?

AdmissionFairSharing: Fix the bug that occasionally a workload may get admitted from a busy LocalQueue,
bypassing the entry penalties.

[k8s-ci-robot]

Skipping CI for Draft Pull Request.
If you want CI signal for your change, please convert it to an actual PR.
You can still manually trigger a test run with /test all

[netlify]

✅ Deploy Preview for kubernetes-sigs-kueue canceled.

Name	Link
🔨 Latest commit	19b4c91
🔍 Latest deploy log	https://app.netlify.com/projects/kubernetes-sigs-kueue/deploys/69270155998f2300086e9ebd

[IrvingMg]

I’ve added a new subpackage for Admission Fair Sharing under the queue package, with the intention of also moving the AfsEntryPenalties struct there.

I think this will help tidy up the code and avoid the circular dependency in CalcLocalQueueFSUsage between the queue and workload packages when using AFS structs. This can be a follow-up

[IrvingMg]

I’ve opened draft PR for early review.

@PBundyra PTAL.

[IrvingMg]

/test pull-kueue-test-e2e-multikueue-main
/test pull-kueue-test-e2e-main-1-33

Look like network issue:

shasum: standard input: no properly formatted SHA checksum lines found
Failed to install helm-unittest
For support, go to https://github.com/kubernetes/helm
Error: plugin install hook for "unittest" exited with error
make: *** [Makefile-deps.mk:110: helm] Error 1

See: https://prow.k8s.io/view/gs/kubernetes-ci-logs/pr-logs/pull/kubernetes-sigs_kueue/7780/pull-kueue-test-e2e-multikueue-main/1992955152935424000

[k8s-ci-robot]

LGTM label has been added.

Git tree hash: 2e5e2b2d75000123f886a0eb1f630e054f142054

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: IrvingMg, mimowo

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [mimowo]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[IrvingMg]

1. manual test on a live cluster for Kueue restart

Here are the logs with the steps I followed for the manual restart test on a Kind cluster:
afs-restart-results.log

[mimowo]

Thank you, the raport lgtm

[PBundyra]

Just some nits, other than that lgtm! Thanks @IrvingMg great job!

[PBundyra]

Leaving unholding to @mimowo

[PBundyra]

I think we should also delete from afsConsumedResources here

[mimowo]

I can see it done here: https://github.com/kubernetes-sigs/kueue/compare/1c312e28b8dae582bf2617a0b4f9afffb475b648..19b4c9198a57c66e7dc79ff40f05b169f7710440#diff-9177ca87118e92b78007b8e08ffc92f1a18afa6fd6b4168b464e497e09ec53ecR225-R227

[mimowo]

Leaving unholding to @mimowo

sg, I think the remaining aspects of the PR:

1. ensure the Delete is called when the LQ is deleted to make sure no memory leak.
2. potentially simplify handling of the corner case discussed in [AFS] Implement cache for AFS consumed resources #7780 (comment) (to be checked)

[mimowo]

Thank you 👍
/unhold
/lgtm

[k8s-ci-robot]

LGTM label has been added.

Git tree hash: 0672030c67b03082b7cb1c179a7bf82ce3674ceb

[mimowo]

@IrvingMg please check if we can cherrypick. Since AFS is alpha on 0.13 and 0.14 this is not "must", but if we can, then it would be great. It would be great to try to backport to 0.14 at least.