[WIP] Set ownership of spec.replicas to scale subresource

[mariantalla]

(with @hoegaarden )

What type of PR is this?
/kind bug

What this PR does / why we need it:
This PR implements field ownership on replicas.scale for the scale subresource, so that changes to that field would behave the same way (including conflicts where applicable) as a server-side apply of the main resource.

Which issue(s) this PR fixes:

xref: #82046

Special notes for your reviewer:

• The path that we're exploring in this PR is adding a managedField entry for spec.replicas, on behalf of the scale subresource (i.e. the scale subresource appears as the manager of the field).
  • To begin with we've implemented it for replica sets; if we see that the approach works we'd have to extend that to other resources too... I'm cautious of duplication there.

Does this PR introduce a user-facing change?:

NONE

Additional documentation e.g., KEPs (Kubernetes Enhancement Proposals), usage docs, etc.:

/cc @apelisse @jennybuckley @julianvmodesto @hoegaarden

[k8s-ci-robot]

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

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

In response to this:

(with @hoegaarden )

What type of PR is this?
/kind bug

What this PR does / why we need it:
This PR implements field ownership on replicas.scale for the scale subresource, so that changes to that field would behave the same way (including conflicts where applicable) as a server-side apply of the main resource.

Which issue(s) this PR fixes:

xref: #82046

Special notes for your reviewer:

• The path that we're exploring in this PR is adding a managedField entry for spec.replicas, on behalf of the scale subresource (i.e. the scale subresource appears as the manager of the field).
• To begin with we've implemented it for replica sets; if we see that the approach works we'd have to extend that to other resources too... I'm cautious of duplication there.

Does this PR introduce a user-facing change?:

NONE

Additional documentation e.g., KEPs (Kubernetes Enhancement Proposals), usage docs, etc.:

/cc @apelisse @jennybuckley @julianvmodesto @hoegaarden

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/test-infra repository.

[apelisse]

@julianvmodesto @jennybuckley

[julianvmodesto]

Not sure if you've seen @jennybuckley's idea yet here:
#83294 (comment)

Seems nice since we could keep the logic to update fieldmanager in the internal package and we could re-use it for ReplicaSets, Deployments etc. as well as CRDs.

[apelisse]

Adding the entry in the ManagedFieldsEntry is not good enough to grab the ownership from other users. You also have to remove it from other users, which is definitely not trivial. The best way to do that is to call FieldManager.Update() with the old and new object (I think both are available in the ScaleREST). Jenny and Julian have discussed the wiring of the FieldManager in ScaleREST.

[k8s-ci-robot]

@mariantalla: PR needs rebase.

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/test-infra repository.

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: mariantalla
To complete the pull request process, please assign deads2k
You can assign the PR to them by writing /assign @deads2k in a comment when ready.

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

Needs approval from an approver in each of these files:

• pkg/registry/OWNERS
• staging/src/k8s.io/apiserver/OWNERS
• test/OWNERS

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

[mariantalla]

Alright, the commit state is a bit (or, a lot!) messy at the moment, so here's where work for this PR is at for now:

• In pkg/registry/apps/replicaset/storage/storage.go there's logic that ensures that there only ever is one field manager for spec.replicas. It does this by looping through all managed fields, removing any that mention spec.replicas and adding an entry that specifies that the current manager is the field manager.

  • That's not a very effective implementation, for a number of reasons, so I don't see us sticking with it eventually. For one, it would need to be duplicated for other main resources too. On top of that, it opens space for a number of bugs (for example the case where there's an entry for a manager of spec.replicas AND spec.somethingelse).

• We've started working through @jennybuckley 's idea (from [WIP] Server-side Apply: Track ownership for scale subresource for Deployments #83294 (comment)). There's a bare-minimum implementation of a scaleFieldManager in staging/src/k8s.io/apiserver/pkg/endpoints/handlers/fieldmanager/fieldmanager.go, and then logic (in staging/src/k8s.io/apiserver/pkg/endpoints/installer.go) to install scaleFieldManager as THE FieldManager in the case of scale subresources.

Next steps:

• Ensure that the end-to-end flow is as expected, i.e. that scaleFieldManager does get called when requests to /scale happen.
• Implement the Apply() and Update() logic for scaleFieldManager.

Yesterday at the working group meeting we said we'd speak with the rest of sig-api-machinery on whether a more reasonable approach would be to redirect subresource Apply/Updates to Apply/Updates of the main resource. Related context: #84530 .

[fejta-bot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle stale

[apelisse]

/remove-lifecycle stale

[fejta-bot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle stale

[apelisse]

/remove-lifecycle stale

[k8s-ci-robot]

@mariantalla: The following tests failed, say /retest to rerun all failed tests:

Test name	Commit	Details	Rerun command
pull-kubernetes-node-e2e-containerd	e65ab39	link	/test pull-kubernetes-node-e2e-containerd
pull-kubernetes-conformance-kind-ipv6	e65ab39	link	/test pull-kubernetes-conformance-kind-ipv6
pull-kubernetes-e2e-gce-device-plugin-gpu	e65ab39	link	/test pull-kubernetes-e2e-gce-device-plugin-gpu
pull-kubernetes-kubemark-e2e-gce-big	e65ab39	link	/test pull-kubernetes-kubemark-e2e-gce-big
pull-kubernetes-files-remake	e65ab39	link	/test pull-kubernetes-files-remake
pull-kubernetes-e2e-gce-ubuntu-containerd	e65ab39	link	/test pull-kubernetes-e2e-gce-ubuntu-containerd
pull-kubernetes-e2e-gce	e65ab39	link	/test pull-kubernetes-e2e-gce
pull-kubernetes-e2e-kind	e65ab39	link	/test pull-kubernetes-e2e-kind
pull-kubernetes-node-e2e	e65ab39	link	/test pull-kubernetes-node-e2e
pull-kubernetes-bazel-build	e65ab39	link	/test pull-kubernetes-bazel-build
pull-kubernetes-dependencies	e65ab39	link	/test pull-kubernetes-dependencies
pull-kubernetes-conformance-kind-ipv6-parallel	e65ab39	link	/test pull-kubernetes-conformance-kind-ipv6-parallel
pull-kubernetes-e2e-gce-100-performance	e65ab39	link	/test pull-kubernetes-e2e-gce-100-performance
pull-kubernetes-bazel-test	e65ab39	link	/test pull-kubernetes-bazel-test
pull-kubernetes-verify	e65ab39	link	/test pull-kubernetes-verify
pull-kubernetes-typecheck	e65ab39	link	/test pull-kubernetes-typecheck

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/test-infra repository. I understand the commands that are listed here.

[fejta-bot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle stale

[fejta-bot]

Stale issues rot after 30d of inactivity.
Mark the issue as fresh with /remove-lifecycle rotten.
Rotten issues close after an additional 30d of inactivity.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle rotten

[apelisse]

/close

@nodo is working on an alternative implementation of this.

[k8s-ci-robot]

@apelisse: Closed this PR.

In response to this:

/close

@nodo is working on an alternative implementation of this.

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/test-infra repository.