imback82 commented on pull request #29473:
URL: https://github.com/apache/spark/pull/29473#issuecomment-682212843


   > Just a question; we cannot set a simpler rule to select which strategy 
(reapportioning or coalescing) we use when reading buckets? I think it is a bit 
annoying to set true to `repartitionBucketsInJoin.enabled` or 
`coalesceBucketsInJoin.enabled` data-by-data.
   
   Good point. One use case for repartition over coalesce is when there are 
enough cores available in the cluster, not to reduce the parallelism by 
coalescing. @c21, did you observe any patterns or heuristics on your workloads 
where repartition is preferred?
   
   > For example, if one has too small buckets (this case is extreme though), 
repartitioning for higher cardinality might not be able to help for better 
performance.
   
   This is still guarded by 
`spark.sql.bucketing.coalesceOrRepartitionBucketsInJoin.maxBucketRatio`, so 
this scenario is a little less concerning?


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



---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to