[GitHub] [spark] Ngone51 commented on pull request #30650: [SPARK-24818][CORE] Support delay scheduling for barrier execution

[GitBox <gi...@apache.org>]

Ngone51 commented on pull request #30650:
URL: https://github.com/apache/spark/pull/30650#issuecomment-771326460


   > If there is insufficient resources to run the entire taskset at (some) highest locality level, and there is atleast one task that can be launched at that level - we will end up always failing the barrier taksset (unless delay is set to 0) ?
   
   Yes
   
   > If yes, as a follow up, do we want to make delay as 0 for barrier tasksets ?
   
   Before this PR, we always recommend users disable delay scheduling by setting delay to 0 as a workaround in the error message. And after this PR, we recommend users use the non-legacy delay scheduling. Unless there's a strong reason that users must use legacy delay scheduling, I personally don't have a strong feeling to have the followup since the non-legacy delay scheduling is enabled by default too. But yes, set delay to 0 would let legacy way work anyway. Although I think this PR can't benefit it in that case since we'll never have a chance to revert. 


----------------------------------------------------------------
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



---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org