SEZ9 commented on PR #12134: URL: https://github.com/apache/seatunnel/pull/12134#issuecomment-6568028686
Quick status update on head `c6fd7fc0d575c9c316fb41beb94eb1fd6dcaefa0`. **CI:** the fork run [41162516391](https://github.com/DanielLeens/seatunnel/actions/runs/41162516391) finished as `completed` / `cancelled` (success=31, cancelled=52, skipped=13). The cancelled jobs include [Run / engine-v2-it (11, ubuntu-latest)](https://github.com/DanielLeens/seatunnel/actions/runs/41162516391/job/124442885699), [Run / transform-v2-it-part-2 (17, ubuntu-latest)](https://github.com/DanielLeens/seatunnel/actions/runs/41162516391/job/124442885546) and [Run / jdbc-connectors-it-part-1 (17, ubuntu-latest)](https://github.com/DanielLeens/seatunnel/actions/runs/41162516391/job/124442885758). Since they were cancelled without producing results, there is no green signal for the new test yet. Could you re-run the cancelled jobs (or point me to a newer attempt)? No unrelated source/dependency edits should be needed for that. **Open points from the previous review (unchanged):** - **F1 (HIGH):** The `CoordinatorService` / `JobMaster` changes aren't visible in the diff body, and the test's terminal `SAVEPOINT_DONE`-after-failover assertion depends on them. Could you summarize in the PR description what the new master does with a pending savepoint it inherits (redrive vs. abort-and-restart), so the test can be read against the intended behavior? - **F2 / F5 (MEDIUM):** The window between observing the in-flight savepoint via `pendingCheckpoints` and `masterNode1.shutdown()` means the savepoint can complete on the old master, and the test would still pass without exercising the failover path. Please add an assertion that proves failover actually happened before `SAVEPOINT_DONE`, or otherwise make the test fail loudly if it degrades into the no-failover scenario. - **F3 / F6 (MEDIUM):** The reflection into `CheckpointCoordinator`'s private `pendingCheckpoints` can throw mid-transition and abort the Awaitility wait; at minimum add `ignoreExceptions()` / null-guards. If there is an existing or cheap test-visible accessor for pending checkpoints, I'd prefer that over reaching into internals this PR also modifies. - **F4 / F7 / F8 (MEDIUM/LOW):** `savepointCaller` should be shut down during test teardown, and the outcome of the `savepointCall` future should be checked (with a timeout) after the final assertion so a pre-failover `savePointJob` failure surfaces as itself rather than as an unrelated 30s Awaitility timeout. Once the cancelled jobs are green and the test hardening above is in, I'm happy to take another look. <!-- streview-comment:2213 --> -- 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]
