KKcorps commented on PR #19998: URL: https://github.com/apache/pinot/pull/19998#issuecomment-6464298712
**This closes the last order the model found, where a segment destroyed before the watermark cap read its max made the round persist no watermark. Nothing new turned up.** I replayed the model's 52 known bad orders and none changed. An hour of search over 8 table shapes found nothing that breaks only with this change, and neither did a second search started next to a marked segment that gets destroyed. Each segment's comparison max is now cached before any thread can mark the segment, and `markSegmentUpdated` publishes it with the marker. So outside tests `getWatermarkToPersist` never meets an unset marker, and a destroyed segment caps with the same max today's code reads from its metadata. That also closes a Helix thread marking a segment already removed out of TTL, which the model cannot build. The model covers one replica and one partition, and not the mapped metadata sidecar itself. <!-- upsert-model-check head=578a04af15a82a22ee3875ca8c6450931936a573 --> -- 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]
