Informer example requires fix: infomer restarted after onerror event doesn't dispatch other events

[faxioman]

Describe the bug The informer example shows a way to deal with errors:

informer.on('error', (err: k8s.V1Pod) => {
  console.error(err);
  // Restart informer after 5sec
  setTimeout(() => {
    informer.start();
  }, 5000);
});

But when an error occurs (eg. the api server pod is restarted) if you restart the informer with informer.start(), all the events (add, delete, update) are never fired again. I don't know if this is a desired behaviour and the example is wrong or the example is fine and the informer has a bug.

** Client Version ** 0.16.3

** Server Version ** Kubernetes v1.23.3

To Reproduce Execute the informer, restart the api server, observe that informer events are never fired

Expected behavior After an error event and an informer restart, all add/delete/update event for the informer should be fired again

Environment (please complete the following information):

• OS: MacOs (minikube v1.25.2)
• NodeJS Version v16.14.0
• Cloud runtime minikube v1.25.2

[brendandburns]

That is expected. Events should only be fired for newly added things (since presumably you already received events for the earlier things).

This should be consistent with the way informers work in other languages also.

[faxioman]

Yes, this behaviour is clear… but after informer.start inside an onerror handler, if I remove a k8s resource and recreate that resource the events delete/add are never fired.

[brendandburns]

@faxioman that's true, in that case you should receive an update event rather than a delete/add event. If you don't receive an update in that situation then there is a bug.

Given the way that the Kubernetes API works, if the watch drops for some reason (forcing a re-list to occur) then I believe it isn't possible to tell the difference between whether an object was deleted and re-added, or that it was simply updated.

Does this behavior cause problems for you? I wouldn't expect that it would cause problems for your operator as it should always be looking at the 'current' state.

As mentioned above, you should get an 'update' event, if you don't get that, then we should figure out why not.

[faxioman]

@brendandburns ok, I'll check if an update event is raised. Anyway, I see that if I simply remove a crd (after an informer restart) the event delete is never fired.

[brendandburns]

Hrm, that seems like a bug. Once the informer is restarted, if you delete something, you should see the delete event.

Can you provide more details on the repro (or sample code)

thanks

[brendandburns]

Friendly ping for more details if you have them.

Thanks!

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues and PRs.

This bot triages issues and PRs according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue or PR as fresh with /remove-lifecycle stale
• Mark this issue or PR as rotten with /lifecycle rotten
• Close this issue or PR with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

[dominykas]

I have a suspicion that some recent changes affected this to the opposite - I'm now seeing unhandled ECONNRESET in 0.17.0, whereas 0.15.x seems to reconnect OK, while 0.16.x seems to just stay silent.

Will be digging deeper...

[k8s-triage-robot]

The Kubernetes project currently lacks enough active contributors to adequately respond to all issues and PRs.

This bot triages issues and PRs according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue or PR as fresh with /remove-lifecycle rotten
• Close this issue or PR with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle rotten

[k8s-triage-robot]

The Kubernetes project currently lacks enough active contributors to adequately respond to all issues and PRs.

This bot triages issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Reopen this issue with /reopen
• Mark this issue as fresh with /remove-lifecycle rotten
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/close not-planned

[k8s-ci-robot]

@k8s-triage-robot: Closing this issue, marking it as "Not Planned".

In response to this:

The Kubernetes project currently lacks enough active contributors to adequately respond to all issues and PRs.

This bot triages issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Reopen this issue with /reopen
• Mark this issue as fresh with /remove-lifecycle rotten
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/close not-planned

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.