rangareddy commented on issue #7829: URL: https://github.com/apache/hudi/issues/7829#issuecomment-4833180383
This is Spark non-determinism, not a Hudi bug — monotonically_increasing_id() (and uuid(), per SPARK-23599) regenerate different values on stage/task re-attempts, and since Hudi's upsert hash-routes on the record key, the recomputed rows land in different file groups, causing duplicates/overwrites. Fix: upgrade to Hudi 0.14.0+ and let Hudi auto-generate the record key (omit recordkey.field) so it's materialized and stable across retries, or use a deterministic content-derived key like sha2(concat_ws(...)) instead of a positional one. -- 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]
