RoanHeNaN commented on issue #70206: URL: https://github.com/apache/doris/issues/70206#issuecomment-6629365308
### Update: reproduced and fix verified on a local build The root-cause analysis above is no longer just code-derived — it is now confirmed on a local single-node cluster built with the proposed change. A regression test reproduces the resurrection with the legacy behavior and verifies suppression with the grace watermark: ``` grace=0 -> k=1 visible (resurrected -> reproduces the bug) grace=3600 -> k=1 deleted (suppressed -> the fix) grace=-1 -> k=1 deleted (never prune -> fallback) Test 1 suites, failed 0 suites ``` Scenario (MOW table with a sequence column): 1. delete `k=1` with a high sequence (`seq=1000`) via delete sign 2. full compaction prunes the delete-sign tombstone 3. late write for `k=1` with a lower sequence (`seq=900`) With `grace=0` (current behavior) `k=1` is resurrected; with the grace window the tombstone is retained and the late lower-seq write is correctly suppressed. This also confirms the key step of the analysis — while the tombstone is alive, `lookup_row_key` + the sequence comparison already suppress the late write. Draft PR with the fix + regression test: apache/doris#70217 Still open for maintainer input on the direction (grace watermark vs. persisting a per-key sequence watermark) and on whether this should be a table property rather than a BE config. -- 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]
