leaves12138 opened a new pull request, #10780:
URL: https://github.com/apache/paimon/pull/10780

   ### Purpose
   
   Row ID reassignment currently aborts as soon as a concurrent COMPACT 
snapshot is committed. Allow it to advance through compaction by rebinding the 
original mappings to the current manifests when logical row ranges are 
preserved. If compaction changes those ranges, including deletion-vector 
materialization or a file that only partially overlaps the plan, replan the 
originally selected partitions against the latest snapshot. Skip the operation 
when no selected partition still requires reassignment.
   
   Recover from expired planning manifest lists and input metadata removed 
during a rewrite when a newer snapshot is available. Retain the existing retry 
budget and rejection of concurrent OVERWRITE snapshots or missing required 
history. Keep historical ADD/DELETE pairs and current data, BLOB, 
deletion-vector and global-index metadata consistent.
   
   ### Tests
   
   - Passed 150 related tests on JDK 8 with normal Maven checks, and on JDK 11 
in an isolated Linux checkout:
   
     ```bash
     mvn -pl paimon-core -DwildcardSuites=none \
       
-Dtest=DataEvolutionRowIdReassignerTest,DataEvolutionRowIdReassignerCompactTest,CompactReassignReuseTest,DataEvolutionDeletionVectorTest
 test
     ```
   
   - Added coverage for manifest/data/BLOB/index compaction, partial overlap 
and repeated replanning, original partition scope, null partition filters, DV 
materialization with BLOB and partial-column updates, expired metadata during 
serial/parallel rewriting, and the rejection/retry-limit paths.
   - Additional isolated metadata-only replay at manifest scan parallelism 1 
and 6 passed entry-by-entry metadata comparison and an independent 
interval-union audit, including compaction that removes the original rewrite 
inputs.
   


-- 
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