davidzollo commented on PR #10808: URL: https://github.com/apache/seatunnel/pull/10808#issuecomment-5549314368
Pushed a fix for the Issue 1 blocker from the 2026-08-22/09-01 reviews (worker-first container startup never actually forming a cluster). **Root cause confirmed**: `LiteNodeDropOutTcpIpJoiner` blocks a lite (worker-only) node from self-promoting to Hazelcast master at two separate gates (the all-blacklisted self-promotion branch, and `isThisNodeMasterCandidate()`), so the worker's join retry loop only succeeds once the master becomes reachable. The test started the worker container and waited for its *entire* sequential setup (image, jar copy, in-container exec) to finish before even starting the master container — pushing the master's availability out far enough to exhaust the worker's join-retry window, reproducing identically in 3/3 prior CI runs. The sibling `JobLogUrlPortIT` starts master-first (which always works, since a non-lite node can self-promote alone); this test deliberately reverses that order to exercise the worker-sees-no-master-yet scenario, which needs both containers' setup to run concurrently, not sequentially. **Fix**: start both `createServer()` calls on separate threads and join both, so the master's setup overlaps with the worker's retry window instead of only beginning after it finishes. **Also fixed**: `JobStateEventTest` had a real compile break (`FAILED_JOB_EVENT_TIMEOUT_SECONDS` undefined) introduced by the dev-sync merge commit — the file already imports and correctly uses the replacement `RESTORE_TO_FAILED_TIMEOUT_SECONDS` at line 94, only line 165 was left on the old name. Fixed to match. Neither fix touches `src/main/**`. Both changes are verified to compile cleanly locally; the container-startup fix needs the Docker-backed E2E to actually run to prove out, which I'm leaving to CI. Issue 2 (the `seatunnel-hazelcast-shade` module removal in the dev-sync) is still open for a confirmation that it was intentional — I don't have the context to answer that myself. -- 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]
