wankunde commented on PR #45661:
URL: https://github.com/apache/spark/pull/45661#issuecomment-2017108745

   > I'm a bit worried about this change, as it changes the assumption of 
always having two phases during shuffle: first phase only write to temp files, 
the second phase "commit" it to the final destination in a short time. The 
shuffle and task scheduling process is quite convoluted in Spark and I can't be 
100% sure that this is a safe change.
   
   Hi, @cloud-fan , the shuffle in this PR still need two phases during shuffle:
   * Stage 1 will write to a TempShuffleBlockId file in the final shuffle data 
file directory, and will make sure no conflict.
   * Stage 2 is "commit phase", and will just rename the TempShuffleBlockId 
file to the final shuffle data file if only one shuffle spill file.


-- 
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]

Reply via email to