Add GC workqueue Forget to stop the rate limiter

[astraw99]

What type of PR is this?

/kind bug

What this PR does / why we need it:

In controller sync func, after the workqueue item AddRateLimited, we need Forget to stop the rate limiter from tracking it if the sync finished without error.

Does this PR introduce a user-facing change?

NONE

[k8s-ci-robot]

Hi @astraw99. 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.

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.

[fedebongio]

/assign @deads2k @caesarxuchao
/triage accepted

[deads2k]

I cannot comment outside the diff, but in the cases where we are not adding back to the queue, why shouldn't we .Forget? A standard controller does the forget in a layer outside the worker itself (example: https://github.com/kubernetes/community/blob/master/contributors/devel/sig-api-machinery/controllers.md#rough-structure).

I think it may be cleaner to model it like this. @caesarxuchao what do you think?

[astraw99]

@deads2k @caesarxuchao PTAL thanks.

[deads2k]

This is a good catch. I'm wondering if it's caused because the structure doesn't follow https://github.com/kubernetes/community/blame/master/contributors/devel/sig-api-machinery/controllers.md#L146-L190 . If we followed that pattern of "return an error to requeue, return nil to be forgotten" we could side-step all the edge cases we're catching with one-off's here, right?

[caesarxuchao]

Good catch. There is no call to Forget() in the entire GC code base. Perhaps the GC code was written before the Forget() method was introduced?

If we followed that pattern of "return an error to requeue, return nil to be forgotten" we could side-step all the edge cases we're catching with one-off's here, right?

That would be ideal but I don't see a clear way to refactor all three non-conforming cases:

• no requeuing if err==enqueuedVirtualDeleteEventErr
• no requeuing if err==namespacedOwnerOfClusterScopedObjectErr
• requeuing if err==nil but !n.isObserved()

I feel it's not worth to refactor.

[deads2k]

I gave the refactor a try and it got sticky for cases where we want to requeue, but we do not want a message in the log. this happens on RESTMapper errors for instance.

Lacking a refactor, I think we need to go through every return and either add

1. AddRateLimited
2. Forget
3. Comment

It's painful, but I think it's the only way to help us avoid misses. For instance, in attemptToDeleteWorker, we don't currently forget in three of the early returns and in attemptToOrphanWorker we're missing one.

[astraw99]

@deads2k @caesarxuchao Tried to make a refactor, PTAL thanks.

[dims]

This is back on reviewers plate! author has already incorporated suggestions.

[deads2k]

This is back on reviewers plate! author has already incorporated suggestions.

Working through some factorization tweaks on slack with the author.

[deads2k]

Great find and good job on the fix.

/lgtm
/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: astraw99, deads2k

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

The pull request process is described here

Needs approval from an approver in each of these files:

• pkg/controller/garbagecollector/OWNERS [deads2k]

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