KKcorps commented on PR #19817: URL: https://github.com/apache/pinot/pull/19817#issuecomment-6165471939
**This closes the race it targets, where the next consumer reads rows while a revert is still rolling their key back and loses partial-upsert updates, plus two known pauseless bad orders. Nothing new turned up, and the pauseless reset from my last comment is gone.** The pauseless orders close because a pauseless partial table in PROTECTED mode now waits for the build before the next segment consumes, so that consumer can no longer merge against a half-added segment or take keys off the building one. I replayed the model's known bad orders with and without the change, and none of the other 45 change. A 61-minute run of 1,296 searches over 24 table shapes found nothing new. Two 20-minute searches that start right before a reverting download or a pauseless build found 32 histories that break today and pass with the change. The model has no dedup tables, does not measure the new wait, and covers one replica's handoff of one partition. <!-- theseus-pr-check head=ba1dd6367c198c9a765e5002c4a5c60f02f095f7 --> -- 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]
