adiranjan25 commented on issue #18597:
URL: https://github.com/apache/iceberg/issues/18597#issuecomment-6830670733

   I've been investigating the subset-key `MERGE INTO` partition-grouping 
behavior described in this issue and have developed a reproducible regression 
test using synthetic Iceberg tables partitioned by `days(event_date), bucket(2, 
id)`.
   
   In my Spark 4.2 test environment, I observed:
   
   - With partially clustered distribution disabled, 10 input scan partitions 
were grouped into 2 execution groups.
   - With partially clustered distribution enabled, the same workload produced 
10 execution groups.
   - Both configurations completed the MERGE correctly.
   
   The test exercises the physical plan and validates the resulting table 
contents. This demonstrates the partition-concentration behavior, although it 
does not establish an OOM or a production performance improvement.
   
   I'm preparing the regression coverage for upstream review.
   
   I'd appreciate maintainer guidance on two questions:
   
   1. Does this test capture the relevant failure mode sufficiently to be 
useful as permanent regression coverage?
   2. Would an additional Iceberg-side planning change or node-local shuffle 
implementation be needed to address the original issue beyond the existing 
Spark configuration?
   
   I'm happy to extend the contribution based on the preferred approach.


-- 
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