eubnara commented on PR #7893: URL: https://github.com/apache/hadoop/pull/7893#issuecomment-5335724636
We hit this issue in production on a ~600 NodeManager cluster (DefaultContainerExecutor): kill-during-localization left resources stuck in DOWNLOADING and leaked the localizer's DFS sockets inside the NM JVM, piling up FIN_WAIT1 connections on DataNodes (HDFS side tracked in HDFS-17965). Since this PR was stale-closed, I've revived it in #8691 with the fix unchanged and a unit test added, which was the remaining blocker. Credit to @zheng-weihao is kept in the commit message. Please let me know if you'd rather reopen this one instead — happy to close mine in that case. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
