zhouyejoe commented on pull request #29392: URL: https://github.com/apache/spark/pull/29392#issuecomment-673037833
@yanxiaole From the stacktrace, there is indeed race condition between the replayTask and checkForLogs()/cleanLogs(), where the latter two has codes which transfers the whole listing.ldb to Scala List, but the replayTask may delete entries from listing.ldb. I do think we can refactor the PR https://github.com/apache/spark/pull/24966 which introduces the listing.delete in replay task. Hi, @dongjoon-hyun. Since you were the original author for this PR, do you think there will be race condition? If there is, how should we resolve it? In relayTask, it can put this deletion into a ConcurrentLinkedQueue, and let the cleanLogs() to delete those entries. ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
