gcbmgr release frequently fails at "push git objects"

[tpepper]

Lately on patch releases we've had as high maybe 5-30% failure rate in "gcbmgr release" at the "push git objects" step. Eg:

Today's run of:

gcbmgr release --nomock --official release-1.14 --buildversion=v1.14.9-beta.0.44+500f5aba80d712

Failed with:

Step #1: ================================================================================
Step #1: PUSH GIT OBJECTS  (11/14)
Step #1: ================================================================================
Step #1:
Step #1: Checkout master branch to push objects: OK
Step #1: Pushing tags
Step #1: * v1.14.9: OK
Step #1: * v1.14.10-beta.0: OK
Step #1: Pushing release-1.14 branch: OK
Step #1: Checkout master branch to push objects: OK
Step #1: Rebase master branch: OK
Step #1: Pushing master branch: FAILED
Step #1: [2019-Nov-13 17:50:09 UTC] push_git_objects in 26s
Step #1: FAILED in push_git_objects.
Step #1:
Step #1: RELEASE INCOMPLETE! Exiting...

Frequently this does not reproduce. A second run of the exact same 'gcbmgr release' command usually will succeed, and this was the case for this 1.14 failure above.

Today we also had the 1.16 'gcbmgr release' fail and refuse to continue, noting the v1.16.3 and v1.16.4-beta.0 tags already existed. This contradicts the above run for 1.14 which failed initially, had pushed tags, yet succeed on re-run despite the existing tags.

Running:

git push --delete upstream v1.16.3
git push --delete upstream v1.16.4-beta.0

and then re-running the same 'gcbmgr release' command for the 1.16 release then succeeded.

This feels like a race or incomplete error handling, perhaps in GCB, gcbmgr, or maybe even in GitHub API accesses?

[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

[fejta-bot]

Rotten issues close after 30d of inactivity.
Reopen the issue with /reopen.
Mark the issue as fresh with /remove-lifecycle rotten.

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

[k8s-ci-robot]

@fejta-bot: Closing this issue.

In response to this:

Rotten issues close after 30d of inactivity.
Reopen the issue with /reopen.
Mark the issue as fresh with /remove-lifecycle rotten.

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

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.

[justaugustus]

Removing the rotten label so this doesn't pop up again in sweeps.
/remove-lifecycle rotten

I think I'm fine w/ this staying closed.
I believe intermittent failures in accessing GitHub can cause this. We saw something similar in #1302 / https://kubernetes.slack.com/archives/CJH2GBF7Y/p1589889924258900

[tpepper]

I wonder if it's the same root cause as whatever causes a notable amount of cherry picks to not have labels correctly applied? Eg: I infer some throttling or limit in API usage.