RoanHeNaN commented on PR #70217: URL: https://github.com/apache/doris/pull/70217#issuecomment-6763394865
> Thanks for working on this. I still see two issues with this approach: > > 1. In shared-nothing mode, the three replicas still compact independently. After the grace period, one replica may have removed the tombstone while another retains it. The same lower-sequence write can then produce different results across replicas. > 2. The grace check uses the newest `creation_time` among all input rowsets. Base compaction can consume rowsets just produced by cumulative compaction, whose `creation_time` reflects when they were built. If base compaction keeps including these recently created rowsets, tombstone cleanup may remain disabled indefinitely, even for very old `delete_sign = 1` rows. > > I have an alternative solution that fully addresses both replica consistency and tombstone cleanup. I’ll submit a fix soon. Thank you @bobhan1. Looking forward to the new solution. -- 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]
