-
Notifications
You must be signed in to change notification settings - Fork 40k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
A pod never terminated if a container image registry was unavailable #23746
A pod never terminated if a container image registry was unavailable #23746
Conversation
Labelling this PR as size/XS |
LGTM |
A good summary here: |
We should also fix the default container state to eliminate the need to enumerate all the special errors. This will help prevent breakage in the future. #23742 (comment) |
Test case? |
@Random-Liu's pending PR (#21224) will add more unit tests for this part of the code. I am okay with merging this PR as it is to patch 1.2 |
GCE e2e build/test passed for commit eeeccd0. |
LGTM |
Ideally, we should add a test in |
@k8s-bot test this [submit-queue is verifying that this PR is safe to merge] |
Will look at an e2e node test next week. |
GCE e2e build/test passed for commit eeeccd0. |
@k8s-bot test this [submit-queue is verifying that this PR is safe to merge] |
I am ok to merge this as is, but could you add a test case in a separate pr? |
GCE e2e build/test passed for commit eeeccd0. |
Automatic merge from submit-queue |
Nice finding @derekwaynecarr ! |
Thanks, @derekwaynecarr! |
@dchen1107 @vishh - I was trying to think how to write an e2e for this, but to reproduce I need to simulate docker returning a 50x error on a particular image pull request which is proving non-trivial for me to figure out how to plumb through. any tips would be appreciated. From a unit test perspective, I agree that #21224 provides coverage and removes the open list around kubecontainer error states. |
I think unit tests are enough in this case. As @derekwaynecarr mentioned, we don't have a good way to inject specific errors in docker client at the node/cluster e2e level, and I'm also not sure how much it'd help for testing one specific error case. |
I didn't expect an e2e test for this. unitttests provided by #21224 sounds good to me. Thanks! |
I did not realize that the error condition was non-trivial. Sorry for the
confusion!
|
release-note-none is not allowed for cherrypicks |
That is, if it matters enough to warrant a cherrypick, it needs a release note |
@bgrant0607, will the PR description be included in the release note as well? |
@yujuhong Currently I think only the title is included in the release note, though I would like to be able to extract a section from the description. |
It is currently only the title. #23743 will address collecting more detail. |
…f-#23746-upstream-release-1.2 Automated cherry pick of #23746 upstream release 1.2
Commit found in the "release-1.2" branch appears to be this PR. Removing the "cherrypick-candidate" label. If this is an error find help to get your PR picked. |
…rry-pick-of-#23746-upstream-release-1.2 Automated cherry pick of kubernetes#23746 upstream release 1.2
…rry-pick-of-#23746-upstream-release-1.2 Automated cherry pick of kubernetes#23746 upstream release 1.2
Fixes #23742
Fixes #22045
Pods will now show proper status when this happens as well:
Where as previously they would never report a reason.
I also removed the "temporary" part of the message because we have no idea if its temporary or permanent.
/cc @kubernetes/sig-node @kubernetes/rh-cluster-infra