Vamsi-klu commented on PR #19083: URL: https://github.com/apache/pinot/pull/19083#issuecomment-5642791805
Thanks @noob-se7en. I agree, and I am closing this. There is no post-#17062 production incident to justify an auto force-commit path. #17062 already covers the transient Kafka recreate case that used to mark a replica OFFLINE. Partial OFFLINE that still happens is either a sick server (force-commit would punish the healthy replicas) or the short Helix-thread init retry, which is a server-side problem. The safer controller repair, OFFLINE to CONSUMING for mixed IN_PROGRESS replicas, already shipped in #17754 as controller.realtime.segment.partialOfflineReplicaRepairEnabled (default off). A second default-off flag is more config than it is worth. I am not taking the N-run gate or rewriting this PR as a reset. The reset exists. If we still want to harden first consumer create, that belongs on the server (#19370), not as RSVM force-commit. Happy to reopen if someone has a concrete post-#17062 incident that neither #17062 nor #17754 covers. cc @Jackie-Jiang @sajjad-moradi @swaminathanmanish -- 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]
