KarmaGYZ commented on PR #17873:
URL: https://github.com/apache/flink/pull/17873#issuecomment-1132533635

   > > @xinbinhuang Thanks for the update. Could you explain why you need to 
derive the total task slots required from the JSON graph?
   > 
   > Our team wants to derive the total slots required for the job before 
redeploying. This help avoid the situation where the old job is stopped and the 
new job doesn't have enough resources to start. With this change we can make 
sure the cluster has enough resources for the new job first before stopping the 
old one. In general, I think slots sharing groups is an useful information to 
better understand the job for debugging and performance tuning. WDYT?
   
   I'm afraid that even with this information, you cannot calculate the exact 
number of slots your job requires because the slot sharing group will not 
restrict the scheduling of operators[1]. It is just a hint for the scheduling 
and might not be obeyed in the future. However, I agree that this is indeed 
helpful information. @godfreyhe WDYT about the compatibility issue introduced 
by this PR?
   
   [1] 
https://nightlies.apache.org/flink/flink-docs-master/docs/deployment/finegrained_resource/#notice


-- 
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]

Reply via email to