SEZ9 commented on PR #12454: URL: https://github.com/apache/seatunnel/pull/12454#issuecomment-5855932790
@DanielLeens The Java 8 failover investigation from my preceding comment is addressed in new head `64c0222b6095ac63733b33b723ef788f8f5e19be`. `PhysicalVertex.cancel()` now reprocesses a terminal vertex only when `!taskFuture.isDone()`. This preserves completion of a late result after recovery while retaining the original no-op behavior for an already-completed result. It is a one-line change, folded into the original recovery commit; the PR still has two commits. The guarded implementation passed the ordinary master-failover case twice on OpenJDK 8/two active processors (approximately 25 and 28 seconds, 1 test/0 failures/errors/skips each), following the failing/current and passing/baseline comparisons already reported. The final current-head tree then passed: - Spotless and the full-reactor verification on OpenJDK 11. - All 45 engine unit tests, including the six late-terminal-result cases, plus 32 PostgreSQL/PayPal unit tests, with zero failures/errors/skips. - The three existing Java 8 integration cases together: ordinary master failover, failover during the close handshake, and checkpoint-trigger dispatch failure, with zero failures/errors/skips. [New complete Build 36319880506](https://github.com/SEZ9/seatunnel/actions/runs/36319880506) is running. Its results are being tracked separately from the old head, and I am not calling remote CI green. The PR description now includes this correction and validation. No CI workflow, timeout, exclusion, or dependency patch was added. Please take another look at the updated recovery guard when available. -- 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]
