Ignore failure when running 'kubectl cp'

[luolanzone]

In some CI jobs, e.g. whole-conformance, a failure tar: antrea-agent.cov.out: file changed as we read it happened occasionally, the failure is caused by kubectl cp, this actually a warning from tar and can be ignored instead of failing the job since the test is actually passed.

[antoninbas]

Does this error mean that we could be uploading a corrupted coverage file to codecov, or is it truly harmless?

[luolanzone]

Does this error mean that we could be uploading a corrupted coverage file to codecov, or is it truly harmless?

I think it should be harmless, this issue doesn't happen frequently, I feel it's OK to not have the right coverage data for one or two PRs.

[antoninbas]

Does this error mean that we could be uploading a corrupted coverage file to codecov, or is it truly harmless?

I think it should be harmless, this issue doesn't happen frequently, I feel it's OK to not have the right coverage data for one or two PRs.

If kubectl cp fails with this error, do you know if any file is created on the host? If not, then I don't have a concern. We want to make sure that run_codecov doesn't fail because of an invalid file, otherwise isn't that change pointless?

BTW, thinking about it some more, do you think that we should also add a short sleep (maybe just 1s) before running the kubectl cp command? It may be enough to get rid of the issue. We could have both (the sleep and || true to ignore a possible error).

[luolanzone]

@antoninbas I submitted another PR to fix kubectl cp issue here #6387, maybe we can on-hold this PR to see if the same kubectl cp with tar failure happens in the future considering the coverage related codes changes a lot.