[WIP] [Feat] Re-do mk admission after eviction in worker cluster #8477

mszadkow · 2026-01-08T19:22:15Z

What type of PR is this?

/kind feature

What this PR does / why we need it:

Which issue(s) this PR fixes:

Special notes for your reviewer:

Still working on the E2E as I want to achieve situation when worker that evicted the workload in the first place can not be admitted, but we don't know which one will it be.

Does this PR introduce a user-facing change?

k8s-ci-robot · 2026-01-08T19:22:18Z

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

k8s-ci-robot · 2026-01-08T19:22:18Z

Adding the "do-not-merge/release-note-label-needed" label because no release-note block was detected, please follow our release note process to remove it.

Details

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.

netlify · 2026-01-08T19:22:21Z

✅ Deploy Preview for kubernetes-sigs-kueue canceled.

Name	Link
🔨 Latest commit	`16cb9d9`
🔍 Latest deploy log	https://app.netlify.com/projects/kubernetes-sigs-kueue/deploys/6960b6f7af02f80008c826dc

k8s-ci-robot · 2026-01-08T19:22:23Z

[APPROVALNOTIFIER] This PR is NOT APPROVED

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

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

Details

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

mszadkow · 2026-01-08T19:22:35Z

/ok-to-test

mszadkow · 2026-01-09T08:10:27Z

/retest

k8s-ci-robot · 2026-01-09T16:57:36Z

@mszadkow: 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-e2e-main-1-35	`16cb9d9`	link	true	`/test pull-kueue-test-e2e-main-1-35`

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.

Details

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.

olekzabl

Looks good! Just a few minor comments.

olekzabl · 2026-01-09T16:12:51Z

pkg/controller/admissionchecks/multikueue/workload.go

+		}
+
+		// workload eviction on worker cluster
+		log.V(5).Info("Workload gets evicted in the remote cluster", "cluster", evictedRemote)


Nit: this present tense feels slightly confusing; IIUC the workload already got evicted.

olekzabl · 2026-01-09T16:49:24Z

pkg/controller/admissionchecks/multikueue/workload.go

+				acs.LastTransitionTime = metav1.NewTime(w.clock.Now())
+				workload.SetAdmissionCheckState(&wl.Status.AdmissionChecks, *acs, w.clock)
+				wl.Status.ClusterName = nil
+				wl.Status.NominatedClusterNames = nil


Out of curiosity - can this have an effect, given this rule?
(or is this intended to replace an empty list with nil in some edge cases?)

olekzabl · 2026-01-09T21:08:32Z

pkg/controller/admissionchecks/multikueue/workload.go

+		}
+
+		for cluster := range group.remotes {
+			if err := client.IgnoreNotFound(group.RemoveRemoteObjects(ctx, cluster)); err != nil {


In most cases, other calls to RemoveRemoteObjects in this file are followed by group.remotes[cluster] = nil.
Should it be done also here?

olekzabl · 2026-01-09T21:19:30Z

test/e2e/multikueue/e2e_test.go

+
+			createdAtWorker := ""
+
+			ginkgo.By("Checking that the workload is created in one of the workers", func() {


Below, you'll modify worker's CQ limits to manipulate what it can fit.
If so, how about using that trick once more, to get control over where the wl lands initially?

AFAICS the initial CPU quotas are: 2 at worker1 and 1 at worker2.
So if you set 1.5 for your workload, it'll certainly land on worker 1.
Then, you could swap the quotas (say, first set 2 at w2, then set 1 at w1) and verify if the workload moved to w2.

Compared to what you have now, +1 CQ update but -4 if blocks. (If I'm not mistaken).

olekzabl · 2026-01-09T21:22:44Z

test/e2e/multikueue/e2e_test.go

+			ginkgo.By("Checking that the workload is re-admitted in the other worker cluster", func() {
+				gomega.Eventually(func(g gomega.Gomega) {
+					g.Expect(k8sManagerClient.Get(ctx, wlKey, managerWl)).To(gomega.Succeed())
+					g.Expect(managerWl.Status.ClusterName).NotTo(gomega.HaveValue(gomega.Equal(createdAtWorker)))


You could also check that it's not empty.

olekzabl · 2026-01-09T21:23:18Z

test/integration/multikueue/jobs_test.go

 			}, gomega.Equal(completedJobCondition))))
 		})
 	})
+	ginkgo.It("Should redo the admission process once the workload looses Admission in the worker cluster", func() {


Suggested change

ginkgo.It("Should redo the admission process once the workload looses Admission in the worker cluster", func() {

ginkgo.It("Should redo the admission process once the workload loses Admission in the worker cluster", func() {

olekzabl · 2026-01-09T21:28:01Z

test/integration/multikueue/jobs_test.go

+		ginkgo.By("check manager's workload ClusterName reset", func() {
+			gomega.Eventually(func(g gomega.Gomega) {
+				managerWl := &kueue.Workload{}
+				g.Expect(managerTestCluster.client.Get(worker1TestCluster.ctx, wlLookupKey, managerWl)).To(gomega.Succeed())


Suggested change

g.Expect(managerTestCluster.client.Get(worker1TestCluster.ctx, wlLookupKey, managerWl)).To(gomega.Succeed())

g.Expect(managerTestCluster.client.Get(managerTestCluster.ctx, wlLookupKey, managerWl)).To(gomega.Succeed())

(or are there reasons to use the other context here?)

olekzabl · 2026-01-09T21:32:34Z

pkg/controller/admissionchecks/multikueue/workload.go

+		// workload eviction on worker cluster
+		log.V(5).Info("Workload gets evicted in the remote cluster", "cluster", evictedRemote)
+		needsACUpdate := acs.State == kueue.CheckStateReady
+		if err := workload.PatchAdmissionStatus(ctx, w.client, group.local, w.clock, func(wl *kueue.Workload) (bool, error) {


Why would we do it when needsACUpdate is false?
(Currently, the update func does nothing in this case, but it still returns true - looks like we'd send an empty patch request to the apiserver?)

k8s-ci-robot added the cncf-cla: yes Indicates the PR's author has signed the CNCF CLA. label Jan 8, 2026

k8s-ci-robot requested review from gabesaba and mimowo January 8, 2026 19:22

k8s-ci-robot added the size/XXL Denotes a PR that changes 1000+ lines, ignoring generated files. label Jan 8, 2026

k8s-ci-robot added the ok-to-test Indicates a non-member PR verified by an org member that is safe to test. label Jan 8, 2026

mszadkow changed the title ~~Fix/8302 reset mk admission cluster~~ WIP Fix/8302 reset mk admission cluster Jan 8, 2026

Restart MK workload admission after remote workload eviction

16cb9d9

mszadkow force-pushed the fix/8302-reset-mk-admission-cluster branch from 7d879a3 to 16cb9d9 Compare January 9, 2026 08:06

k8s-ci-robot added size/L Denotes a PR that changes 100-499 lines, ignoring generated files. and removed size/XXL Denotes a PR that changes 1000+ lines, ignoring generated files. labels Jan 9, 2026

mszadkow changed the title ~~WIP Fix/8302 reset mk admission cluster~~ [WIP] [Feat] Re-do mk admission after eviction in worker cluster Jan 9, 2026

mszadkow marked this pull request as ready for review January 9, 2026 14:25

k8s-ci-robot requested a review from kannon92 January 9, 2026 14:25

olekzabl reviewed Jan 9, 2026

View reviewed changes


		createdAtWorker := ""

		ginkgo.By("Checking that the workload is created in one of the workers", func() {

	ginkgo.It("Should redo the admission process once the workload looses Admission in the worker cluster", func() {
	ginkgo.It("Should redo the admission process once the workload loses Admission in the worker cluster", func() {

	g.Expect(managerTestCluster.client.Get(worker1TestCluster.ctx, wlLookupKey, managerWl)).To(gomega.Succeed())
	g.Expect(managerTestCluster.client.Get(managerTestCluster.ctx, wlLookupKey, managerWl)).To(gomega.Succeed())

[WIP] [Feat] Re-do mk admission after eviction in worker cluster #8477

Are you sure you want to change the base?

[WIP] [Feat] Re-do mk admission after eviction in worker cluster #8477

Conversation

mszadkow commented Jan 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What type of PR is this?

What this PR does / why we need it:

Which issue(s) this PR fixes:

Special notes for your reviewer:

Does this PR introduce a user-facing change?

Uh oh!

k8s-ci-robot commented Jan 8, 2026

Uh oh!

k8s-ci-robot commented Jan 8, 2026

Uh oh!

netlify bot commented Jan 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✅ Deploy Preview for kubernetes-sigs-kueue canceled.

Uh oh!

k8s-ci-robot commented Jan 8, 2026

Uh oh!

mszadkow commented Jan 8, 2026

Uh oh!

mszadkow commented Jan 9, 2026

Uh oh!

k8s-ci-robot commented Jan 9, 2026

Uh oh!

olekzabl left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

mszadkow commented Jan 8, 2026 •

edited

Loading

netlify bot commented Jan 8, 2026 •

edited

Loading