LuciferYang opened a new issue, #13547: URL: https://github.com/apache/gluten/issues/13547
On Spark 4.1 and later, with `spark.sql.shuffle.orderIndependentChecksum.enabled` or `spark.sql.shuffle.orderIndependentChecksum.enableFullRetryOnMismatch`, `VeloxHashShuffleWriter::computeRowBasedChecksums` hashes each row to build the map output checksum. For hash and range partitioning it drops the first column, assuming it is the partition id, but every caller of `doSplit` has already removed the partition id column. So the first data column never contributes to the checksum, and a retried map task whose output differs only in that column is not detected. This issue was written with the assistance of AI (Claude Opus). -- 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]
