WIP: Add integration test for DRA config change

[nojnhuh]

What type of PR is this?

/kind cleanup

What this PR does / why we need it:

This PR addresses the first "Nice to have" item listed in #5873:

• Add integration for config change Add support for DRA devices in kueue workloads #5873 (comment)

This PR extends the existing DRA integration test that ensures a workload with an unmapped DeviceClass is not admitted. After the existing steps are done, this PR adds steps to restart the controllers with new configuration that maps the DeviceClass and ensures the workload is admitted.

Which issue(s) this PR fixes:

Fixes #

Special notes for your reviewer:

Does this PR introduce a user-facing change?

NONE

[netlify]

✅ Deploy Preview for kubernetes-sigs-kueue canceled.

Name	Link
🔨 Latest commit	98914fc
🔍 Latest deploy log	https://app.netlify.com/projects/kubernetes-sigs-kueue/deploys/68cc7c7fce3095000834a3de

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: nojnhuh
Once this PR has been reviewed and has the lgtm label, please assign mimowo for approval. For more information see the Code Review Process.

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

Needs approval from an approver in each of these files:

• OWNERS

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

[nojnhuh]

/cc @alaypatel07 @helayoty

This is WIP because the test was failing for me locally but I think it's pretty close. It fails because after restarting the controllers, the workload is admitted as expected, but doesn't seem to have reserved any DRA resources. A plain restart of the controllers without changing the config at all also admits the workload, which I wouldn't expect, but makes sense given that first observation.

If anything jumps out as the cause for the failure I'm all ears!

[k8s-ci-robot]

@nojnhuh: GitHub didn't allow me to request PR reviews from the following users: helayoty.

Note that only kubernetes-sigs members and repo collaborators can review this PR, and authors cannot review their own PRs.

In response to this:

/cc @alaypatel07 @helayoty

This is WIP because the test was failing for me locally but I think it's pretty close. It fails because after restarting the controllers, the workload is admitted as expected, but doesn't seem to have reserved any DRA resources. A plain restart of the controllers without changing the config at all also admits the workload, which I wouldn't expect, but makes sense given that first observation.

If anything jumps out as the cause for the failure I'm all ears!

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-sigs/prow repository.

[nojnhuh]

Changes here are almost entirely from indentation, so the "hide whitepsace" checkbox should show a more meaningful diff.

[k8s-ci-robot]

@nojnhuh: The following test failed, say /retest to rerun all failed tests or /retest-required to rerun all mandatory failed tests:

Test name	Commit	Details	Required	Rerun command
pull-kueue-test-integration-baseline-main	98914fc	link	true	/test pull-kueue-test-integration-baseline-main

Full PR test history. Your PR dashboard. Please help us cut down on flakes by linking to an open issue when you hit one in your PR.

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-sigs/prow repository. I understand the commands that are listed here.

[alaypatel07]

Thanks @nojnhuh for raising this PR.

the workload is admitted as expected, but doesn't seem to have reserved any DRA resources. A plain restart of the controllers without changing the config at all also admits the workload, which I wouldn't expect, but makes sense given that first observation.

There is a difference between how DRA worklaods are processed and how other workloads are processed in Kueue. DRA workloads need some extra get calls, so they are skipped in event handlers and are instead processed in reconcile loop. Considering this backgroud, the scenario you described smells like some event handler is adding DRA workload to scheduler queue before reconcile function gets a chance to pre-process the data.

@nojnhuh do we have traces of logs where workload is successfully admitted and where dra-preprocess is complete? the timestamps on those entries will verify the theory

also cc @mimowo

[alaypatel07]

@nojnhuh found the culprits that are added workload to scheduler queue: #6949. Once these changes are merged, the test case here should pass.

[nojnhuh]

@nojnhuh found the culprits that are added workload to scheduler queue: #6949. Once these changes are merged, the test case here should pass.

Thanks! I'll take a look there.

One thing I forgot to mention is that when I test this scenario by hand on a real cluster, I don't see the same issue. To me, it looks like I'm not restarting things correctly in the test in case that helps inform what the root cause is.

[nojnhuh]

@nojnhuh do we have traces of logs where workload is successfully admitted and where dra-preprocess is complete? the timestamps on those entries will verify the theory

@alaypatel07 Does the "DRA Integration when DRA is configured via ConfigMap Should admit workload with DRA resource claim templates" test case match those conditions? If not I would need some help tweaking the tests to show what you're looking for.

[alaypatel07]

@nojnhuh I think the test you have implemented looks like the right place. It found some bugs in the implementation that is fixed by #6949, with the latest fixes I am able to make the test go green, PTAL there

[alaypatel07]

/close

changes here were picked and merged in #6949, thanks @nojnhuh!

[k8s-ci-robot]

@alaypatel07: Closed this PR.

In response to this:

/close

changes here were picked and merged in #6949, thanks @nojnhuh!

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-sigs/prow repository.