guoweiM commented on pull request #13898:
URL: https://github.com/apache/flink/pull/13898#issuecomment-721442906


   > From the solution, it seems that you think that the source of the 
instability is that the notification for the completion of the checkpoint does 
not come before the test exits, right?
   
   Yes. In the streaming execution mode the committer/globalCommitter depends 
on the checkpoint complete so I let the source waits until the 
Committer/GlobalCommitter receive all the datas. 
   (Of course if we finish the 《FLIP-147: Support Checkpoints After Tasks 
Finished》 there will be no this problem.)
   
   
   Another thing that makes the test unstable is that the order in which the 
`Committer` and `GlobalCommitter` receives the checkpoint complete 
notification. It leads that we could not know the output of `GlobalCommitter` 
when a checkpoint is complete. So This pr also change how to verify the output 
of `GlobalCommitter`.
   This is an example that `GlobalCommitter` receive the checkpoint complete 
before the no.3 `Committer`.
   
   
   
   
![image](https://user-images.githubusercontent.com/26054913/98054342-a436e500-1e75-11eb-9f72-a829dbe4f260.png)
   
   
   
   


----------------------------------------------------------------
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]


Reply via email to