You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by GitBox <gi...@apache.org> on 2020/12/05 13:04:26 UTC

[GitHub] [flink] becketqin commented on pull request #14314: [FLINK-20492][runtime] Add implementations of cancelTask() and finish…

becketqin commented on pull request #14314:
URL: https://github.com/apache/flink/pull/14314#issuecomment-739248485


   @StephanEwen As far as I understand, we rely on the `afterInvoke()` to close all the operators after the mailbox loop exits. However, if the task was canceled, `StreamTask.invoke()` would just throw a `CancelTaskException` and bypass the `afterInvoke()` call. In this case, `cleanUpInvoke()` will run but it does not close the operators. It does confuses me a bit and might be another bug. This bug fix did not touch that part but just follow the way `SourceStreamTask` does.
   
   BTW, I think this problem only has practical impact on standalone clusters. Otherwise, the TM will be teared down anyways after the job cancellation, so there won't be actual resource leak. But from what I understand, the standalone cluster are still used by some users. So it seems still something necessary to fix.


----------------------------------------------------------------
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:
users@infra.apache.org