scheduler_perf: create pods and PodGroups from the group ops by yashaggarwal85d · Pull Request #143388 · kubernetes/kubernetes

yashaggarwal85d · 2026-10-25T06:17:15Z

What type of PR is this?

/kind cleanup
/sig scheduling
/wg workload-aware-scheduling

What this PR does / why we need it:

Follow-up from #141706. createPodGroups now has an optional pods (a createPods spec) and podsPerGroup/podsPerGroupParam, and createCompositePodGroups has an optional podGroups. The nested op gets its count and namespace from the parent, and podsPerGroup / parentLevel / childrenPodGroupsPerCPG are passed to its template. That lets us drop countMultiplierParam, the $podGroupsCount and $parentLevel params that createCompositePodGroups used to write into the workload params, and params.set. All configs that used the old pattern are moved over. Standalone createPodGroups and createPods still work as before.

Which issue(s) this PR is related to:

Fixes #143281

Special notes for your reviewer:

Ran go test ./test/integration/scheduler_perf/ and the TestSchedulerPerf integration tests (short and non-short) for cpg, podgroup/..., default_preemption and workload_preemption.

Does this PR introduce a user-facing change?

NONE

AI usage disclosure:

YES. This PR was written in part with the assistance of generative AI (drafting the change and migrating the configs). I reviewed and tested it, and I'm responsible for the changes.

kubernetes-prow · 2026-10-25T06:17:26Z

Hi @yashaggarwal85d. Thanks for your PR.

I'm waiting for a kubernetes member to verify that this patch is reasonable to test. If it is, they should reply with /ok-to-test on its own line. Until that is done, I will not automatically test new commits in this PR, but the usual testing commands by org members will still work.

Regular contributors should join the org to skip this step.

Once the patch is verified, the new status will be reflected by the ok-to-test label.

I understand the commands that are listed here.

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.

kubernetes-prow · 2026-10-25T06:17:27Z

This issue is currently awaiting triage.

If a SIG or subproject determines this is a relevant issue, they will accept it by applying the triage/accepted label and provide further guidance.

The triage/accepted label can be added by org members by writing /triage accepted in a comment.

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.

kubernetes-prow · 2026-10-25T06:17:57Z

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: yashaggarwal85d
Once this PR has been reviewed and has the lgtm label, please assign tosi3k 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:

test/integration/scheduler_perf/OWNERS

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

pacoxu · 2026-10-26T09:20:55Z

/ok-to-test

/test pull-kubernetes-scheduler-perf

/cc @iomarsayed

iomarsayed · 2026-10-27T16:35:59Z

/assign

iomarsayed

Thanks for taking this. Here are my first set of comments.

iomarsayed · 2026-10-27T16:41:55Z

 // It creates the configured number of PodGroups from a template and
 // then waits for them to be visible in the scheduler's informer cache.
-// This ensures that subsequent operations (like creating pods that reference these pod groups) won't fail due to cache lag.
+// This ensures that the pods referencing these pod groups won't fail due to cache lag.


I think the old comment still applies, let's just change the word subsequent because it is not the most accurate anymore:

Suggested change

// This ensures that the pods referencing these pod groups won't fail due to cache lag.

// This ensures that dependent operations (like creating pods that reference these pod groups) won't fail due to cache lag.

Made the suggested change

iomarsayed · 2026-10-27T16:46:44Z

+	PodsPerGroup int
+	// PodsPerGroupParam is the name of the parameter that determines PodsPerGroup.
+	PodsPerGroupParam string
+	// Pods describes the pods that get created for these PodGroups once they are cached.
+	// Count and Namespace must not be set, they are derived from the PodGroups.
+	// podsPerGroup is passed to the pod template.
+	// Optional.
+	Pods *createPodsOp


Please let's rename this variable into the operation name: createPodsOp.
It will have a better reflection in configs and readability.

Also this field cannot be optional, It is not sensible to have a CPG tree without PodGroups for our scheduler perf tests.
Let's also add tests to ensure this field is present.

Made the suggested change

iomarsayed · 2026-10-27T16:47:03Z

+	// Count and Namespace must not be set, they are derived from the hierarchy.
+	// parentLevel and childrenPodGroupsPerCPG are passed to the PodGroup template.
+	// Optional.
+	PodGroups *createPodGroups


Same comment, please let's rename this variable into the operation name: createPodGroupsOp.
It will have a better reflection in configs and readability.

Also this field cannot be optional, It is not sensible to have a CPG tree without PodGroups for our scheduler perf tests.
Let's also add tests to ensure this field is present.

Renamed!
But For createPodGroups I kept it optional for now because dra/workloadresourceclaims and starvation still use createPodGroups on its own

iomarsayed · 2026-10-27T17:03:29Z

+// This ensures that the PodGroups referencing these CompositePodGroups won't fail due to cache lag.
+// If op.PodGroups is set, it then creates the leaf PodGroups.


Same suggestion as here: https://github.com/kubernetes/kubernetes/pull/143388/changes#r4391479010

iomarsayed · 2026-10-28T15:41:23Z

+	if cpg.TemplatePath == "" {
+		return fmt.Errorf("templatePath must be set")
+	}
+	if cpg.Pods == nil {


Once the operation field is not optional, this validation should be adjusted to return an error immediately.

iomarsayed · 2026-10-28T15:42:31Z

+	if err := cpg.isValid(false); err != nil {
+		return nil, err
+	}
+	if cpg.Pods != nil {


Once the operation field is not optional, we don't need this check anymore, where isValid() will check for it and fail immedaitely.

iomarsayed · 2026-10-28T15:43:25Z

 	if !allowParameterization && ccpg.ChildrenPodGroupsPerCPG < 1 {
 		return fmt.Errorf("childrenPodGroupsPerCPG must be greater than 0, got %d", ccpg.ChildrenPodGroupsPerCPG)
 	}
+	if ccpg.PodGroups != nil {


Same here: Once the operation field is not optional, this validation should be adjusted to return an error immediately.

Done, it returns an error now if createPodGroupOp isn't set

yashaggarwal85d · 2026-10-29T12:03:19Z

Hey @iomarsayed, thanks for the review, I was making these changes and forgot to reply. Thanks for waiting. I have made some fixes as suggested. lemme reply to the comments as well

yashaggarwal85d · 2026-10-29T12:24:23Z

In DRA, measurePods is a total that the pod template spreads round-robin with Mod.Index.PodGroupCount, so per-group creation would change which pods land in which group. Starvation runs a churn op between creating the PodGroup and the pods. Do you want me to move those over in this PR too, or keep it optional for those cases?

yashaggarwal85d · 2026-11-02T04:18:58Z

/test pull-kubernetes-unit

github-project-automation Bot added this to SIG Scheduling and WG Workload-aware Scheduling Oct 25, 2026

github-project-automation Bot moved this to Needs Triage in SIG Scheduling Oct 25, 2026

kubernetes-prow Bot added the cncf-cla: yes Indicates the PR's author has signed the CNCF CLA. label Oct 25, 2026

github-project-automation Bot moved this to Backlog in WG Workload-aware Scheduling Oct 25, 2026

kubernetes-prow Bot added needs-triage Indicates an issue or PR lacks a `triage/foo` label and requires one. needs-ok-to-test Indicates a PR that requires an org member to verify it is safe to test. labels Oct 25, 2026

kubernetes-prow Bot added needs-priority Indicates a PR lacks a `priority/foo` label and requires one. area/test sig/testing Categorizes an issue or PR as relevant to SIG Testing. labels Oct 25, 2026

kubernetes-prow Bot requested review from damemi and denkensk October 25, 2026 06:18

kubernetes-prow Bot requested a review from iomarsayed October 26, 2026 09:20

yashaggarwal85d force-pushed the scheduler-perf-inline-podgroup-ops branch from e885c5e to 699e57f Compare October 27, 2026 05:15

kubernetes-prow Bot removed the needs-rebase Indicates a PR cannot be merged because it has merge conflicts with HEAD. label Oct 27, 2026

kubernetes-prow Bot assigned iomarsayed Oct 27, 2026

iomarsayed reviewed Oct 28, 2026

View reviewed changes

yashaggarwal85d force-pushed the scheduler-perf-inline-podgroup-ops branch from e697a88 to bab6463 Compare October 28, 2026 17:03

yashaggarwal85d requested a review from iomarsayed October 29, 2026 12:26

kubernetes-prow Bot added the needs-rebase Indicates a PR cannot be merged because it has merge conflicts with HEAD. label Oct 30, 2026

yashaggarwal85d added 2 commits October 31, 2026 06:38

scheduler_perf: create pods and PodGroups from the group ops

acb9f37

scheduler_perf: rename nested ops and require createPodGroupsOp

0340d75

yashaggarwal85d force-pushed the scheduler-perf-inline-podgroup-ops branch from bab6463 to 0340d75 Compare October 31, 2026 06:55

kubernetes-prow Bot removed the needs-rebase Indicates a PR cannot be merged because it has merge conflicts with HEAD. label Oct 31, 2026

	// This ensures that the pods referencing these pod groups won't fail due to cache lag.
	// This ensures that dependent operations (like creating pods that reference these pod groups) won't fail due to cache lag.

		// This ensures that the PodGroups referencing these CompositePodGroups won't fail due to cache lag.
		// If op.PodGroups is set, it then creates the leaf PodGroups.

Conversation

yashaggarwal85d commented Oct 25, 2026

What type of PR is this?

What this PR does / why we need it:

Which issue(s) this PR is related to:

Special notes for your reviewer:

Does this PR introduce a user-facing change?

AI usage disclosure:

Uh oh!

kubernetes-prow Bot commented Oct 25, 2026

Uh oh!

kubernetes-prow Bot commented Oct 25, 2026

Uh oh!

kubernetes-prow Bot commented Oct 25, 2026

Uh oh!

pacoxu commented Oct 26, 2026

Uh oh!

iomarsayed commented Oct 27, 2026

Uh oh!

iomarsayed 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!

yashaggarwal85d Oct 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

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!

yashaggarwal85d commented Oct 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

yashaggarwal85d commented Oct 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

yashaggarwal85d commented Nov 2, 2026

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

yashaggarwal85d Oct 29, 2026 •

edited

Loading

yashaggarwal85d commented Oct 29, 2026 •

edited

Loading

yashaggarwal85d commented Oct 29, 2026 •

edited

Loading