SEZ9 commented on PR #12454: URL: https://github.com/apache/seatunnel/pull/12454#issuecomment-5855685830
@DanielLeens I found a reproducible Java 8 failover problem on the current head `4f86cb0ac63e` while investigating the engine job's stalled log. Please treat this head as not ready pending the fix. Using the existing `CheckpointCoordinatorFailoverIT#testBatchJobCompletesAfterMasterFailover`, OpenJDK 8 and two active processors, with no changes to the test or CI: - Current head: two local executions did not complete within the 600-second diagnostic limit. - Unmodified base `025d08f9bcb0`: two executions passed (1 test, 0 failures/errors/skips each), each in about 27 seconds. - Current head with only the `TaskExecutionService` change removed, in an isolated local worktree: still did not complete within 600 seconds. - Current head with only the `PhysicalVertex.cancel()` change removed, retaining the task-startup fix: passed in about 27 seconds (1/0/0/0). The thread dumps show tasks waiting in `READY_START` and the client waiting for job completion. This strongly implicates the cancellation/recovery change, although I am still tracing the exact interaction before proposing a correction. The remote Java 8 engine job is still running; its last log is from this same ordinary master-failover case. This is separate from the earlier close-handshake case. The successful PostgreSQL result and Java 11 engine result reported above remain valid. No workflow changes, skips, or unverified source patch have been pushed during this diagnosis. I am working on the correction and will report its local and remote validation. -- 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]
