SEZ9 commented on issue #12383: URL: https://github.com/apache/seatunnel/issues/12383#issuecomment-5724754506
Thanks for the detailed report. Could you attach the full `RocketMqIT` test log and the broker container log for one failing leg (ideally the fork CI run link)? Broker-side lines around the offset commit and any exception logged or swallowed inside `RocketMqAdminUtil.currentOffsets` would help tell a never-persisted commit apart from a failed or silently-empty admin query. On your suspect areas: the rebalance race looks plausible, since the test consumer is shut down right after `commitSync()`. Two things worth trying: (1) after the commit, poll `queryConsumerOffset` from the same consumer until it returns the offset before shutting the consumer down, and (2) make sure the commit actually reaches the broker synchronously rather than only being enqueued client-side. Since the window was already widened from 30s to 60s without effect, a longer timeout is not the fix. Please also check whether `RocketMqAdminUtil.currentOffsets` returns an empty map or omits queues on any exception (e.g. the consumer group not yet registered on `broker-a`, or an admin query to the wrong broker); if so, logging or rethrowing that error would make the IT fail with the real cause instead of a generic `ConditionTimeoutException`. <!-- streview-comment:1146 --> -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
