kafka: avoid unknown partition error during partition movement

[dotnwat]

If we trigger a partition movement and then immediately start issuing list_offset requests to the partition, librdkafka will produce the error % ERROR: offsets_for_times failed: Local: Unknown partition.

This was noticed in ducktape tests, and the retry mechanisms in place cause this to not be a test failure.

However, we should probably think about how to avoid this in practice by returning an error that will cause the client to backoff and refresh metadata? Ideally leadership transfer is transparent to the client with the exception of perhaps some temporary performance issues.

JIRA Link: CORE-778

[github-actions[bot]]

This issue hasn't seen activity in 3 months. If you want to keep it open, post a comment or remove the stale label – otherwise this will be closed in two weeks.

[github-actions[bot]]

This issue was closed due to lack of activity. Feel free to reopen if it's still relevant.