SEZ9 commented on issue #13319: URL: https://github.com/apache/seatunnel/issues/13319#issuecomment-7022925937
@SEPURI-SAI-KRISHNA thanks for the update, and sorry for the slow answer. What you describe covers the failure reported here: in thread-share mode an `init()` failure led to `threadShareTaskQueue.add(null)`, the NPE left the published `TaskGroupContext` stuck in `executionContexts`, and later redeploys were skipped so the job hung. Calling `taskDone` for the tasks that are not submitted and queueing only non-null trackers is the right place to fix it. The two new checks in `testThreadShareTaskInitFailureCompletesTheGroup` target the stuck-context and skipped-redeploy symptoms directly: the location leaves `executionContexts`, and a second `deployTask` succeeds with the new execution id. It also helps that the test times out without the `taskDone` calls, since that shows it really guards the fix. Leaving `deployLocalTask` unchanged and noting the boundary with #12218 in the PR description is fine with me. I have not gone through the diff in #13343 yet, so this is based on your description, not a review. I need nothing more from you on this issue, and the rest will happen on the PR. Two things to confirm there: 1. **Original assertions:** does the test still assert the items from your plan (no NPE, the second task is not initialized, and the group completes FAILED with the injected `init()` error) alongside the two new checks? 2. **Modes covered:** which thread-share mode does the test run in? The problem applies to both ALL and PART, so please say whether both are covered, or why one is enough. <!-- streview-comment:2455 --> -- 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]
