deepakpanda93 commented on issue #8835: URL: https://github.com/apache/hudi/issues/8835#issuecomment-4845031463
This is fixed in Hudi 0.14.0+ (and all 1.x). Root cause: a known global-index bug — with GLOBAL_BLOOM and hoodie.bloom.index.update.partition.path=true on a MOR table, when a record's partition changes the delete from the old partition wasn't generated correctly, leaving the record duplicated in the new partition. Fix: HUDI-5968 / [#8490](https://github.com/apache/hudi/pull/8490) ("Fix global index duplicate and handle custom payload when update partition"), first released in 0.14.0. As @ad1happy2go noted in this thread, the patch was not in 0.13.1 (it targeted 0.14.0), which is why upgrading to 0.13.1 didn't resolve it. Two notes: - Please test on a recent 1.x release (which includes this fix). - The fix prevents new duplicates on update-partition-path; any records already duplicated by the older version won't disappear on upgrade — you'll need a one-time dedup/rewrite of the affected keys. Closing as fixed in 0.14.0+. If you still observe duplicates with GLOBAL_BLOOM + update-partition-path on 1.x (with data written after the upgrade), please reopen with a minimal repro. Thanks! -- 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]
