ShivsundarR commented on PR #19789: URL: https://github.com/apache/kafka/pull/19789#issuecomment-2912377957
I noticed @adixitconfluent 's PR [here](https://github.com/apache/kafka/pull/19824) which adds a `future.get()` to `produceAbortedTransaction()`. This could explain the flakiness in `testAlterReadUncommittedToReadCommittedIsolationLevel` as well. If the abort transaction does not complete in time, then we could possibly see these records in a different `poll()`. So that PR should fix the problem and we would get both the records on a single `poll()`. It is now merged. This also explains why the other tests which use explicit mode and expect multiple records are not flaky. Initially I thought this could be a race condition in the client, but it seems unlikely now. We can observe the builds on AK to see if the flakiness persists and then we can close this PR if it is resolved. We can keep the PR open till then. The refactors in `waitedPoll()` should also not be required as we would ideally get all the produced records in a single `poll()`. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: jira-unsubscr...@kafka.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
