leixm commented on PR #2725: URL: https://github.com/apache/celeborn/pull/2725#issuecomment-2333063844
> After discussing with @SteNicholas , we thought of two reasons why removeBatch is called after cleanup: > > 1. stage retry without [[CELEBORN-1496] Differentiate map results with only different stageAttemptId #2609](https://github.com/apache/celeborn/pull/2609) , the retried task has the same (shuffleId, mapId, attemptId), which has already mapper-ended in last retry > 2. attempt 0 succeeds in `write` and celeborn client successfully called mapperEnd, but Spark thinks it's failed and rerun other attempts, then other attempts cleans up pushState because of mapper ended. Agree, scene 2 often happens in our production environment. -- 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]
