fix: skip ephemeral volume topology requirements for bound pods during consolidation

[moko-poi]

Summary

Fix for #2803: Ephemeral volumes with zone TopologySpreadConstraints block consolidation with Unconsolidatable events.

Skip ephemeral volume topology requirements when evaluating bound pods for consolidation, since ephemeral PVCs are deleted with the pod and a new PVC will be created on rescheduling. This primarily targets WaitForFirstConsumer StorageClasses, which is the common case for ephemeral volumes. See inline comment for discussion on
Immediate binding mode.

Test plan

• Unit tests: bound pod + ephemeral volume returns no topology requirements
• Unit tests: unbound pod + ephemeral volume returns topology requirements (unchanged behavior)
• Unit tests: bound pod + regular PVC returns topology requirements (unchanged behavior)
• Integration test: consolidation replaces node with ephemeral volumes and zone TSC
• Full scheduling test suite passes (339 tests)
• Full disruption test suite passes (233 tests)

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is NOT APPROVED

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

[k8s-ci-robot]

Hi @moko-poi. Thanks for your PR.

I'm waiting for a kubernetes-sigs 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.

Tip

We noticed you've done this a few times! Consider joining the org to skip this step and gain /lgtm and other bot rights. We recommend asking approvers on your previous PRs to sponsor you.

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.

[moko-poi]

@jmdeal Thanks for the analysis on this issue! I wanted to get your thoughts on one thing.

In your comment on #2803, you noted

One nuance that I'll note is that if the VolumeBindingMode is Immediate, we would still want to honor the ephemeral volume constraints.

The current implementation skips topology requirements for all ephemeral volumes on bound pods, regardless of VolumeBindingMode:

if volume.Ephemeral != nil && pod.Spec.NodeName != "" {
    return nil, nil
}

For WaitForFirstConsumer, this is correct — the new PVC will bind to whatever zone the pod is scheduled to, so the old PVC's zone constraint is irrelevant.

For Immediate, the situation is more nuanced. If we skip the constraint, Karpenter may create a replacement node in a zone that differs from where the provisioner binds the new PVC, which could leave the pod temporarily unschedulable. However, if we don't skip, the old PVC's zone constraint will continue to cause incorrect TSC
counts, blocking consolidation entirely (which is the current bug). Since the old PVC is deleted with the pod and the new PVC's zone is determined independently by the provisioner, honoring the old constraint doesn't guarantee correctness either.

Given that Immediate + ephemeral volumes is an uncommon combination and skipping is still an improvement over the current behavior (permanently blocked consolidation), I went with the simpler approach for now. That said, if you think Immediate mode warrants additional handling, I'd like to discuss what the right approach would
be. Let me know your thoughts.