SEZ9 commented on PR #12710: URL: https://github.com/apache/seatunnel/pull/12710#issuecomment-6159926500
Thanks for keeping this rebased on `dev`. davidzollo has approved the current head (`0852d6ae`), and the files this PR owns are unchanged since my last pass on `c3a2538f`, so my earlier three points are still the things I'd like to close out before merging: 1. **Master chart – shutdown policy (PR12710-F1).** The `exec java` change means SIGTERM now reaches the JVM, which is the right root-cause fix. However, with Hazelcast's default shutdown-hook policy the hook performs `TERMINATE`, and `SeaTunnelServer.shutdown` doesn't act on the terminate flag, so the member leaves abruptly rather than performing the clean stop the PR description promises. Could you either configure `hazelcast.shutdownhook.policy` for a graceful leave in the chart, or adjust the description to state exactly what happens today? 2. **Worker chart – `terminationGracePeriodSeconds` (PR12710-F2).** Could you point me to where `terminationGracePeriodSeconds` is wired for the worker (and master) deployment, and confirm the default is large enough for a graceful Hazelcast leave? Otherwise Kubernetes will still SIGKILL mid-shutdown. 3. **Docs (PR12710-F3).** The Kubernetes graceful-shutdown docs still present the `preStop` + `stop-seatunnel-cluster.sh` workaround as the way to get a clean stop. If the en/zh docs were updated, please link the sections so I can verify they describe the new SIGTERM → shutdown-hook behaviour and clarify whether the `preStop` hook is still needed. Once those are confirmed (or the remaining gaps are addressed) I'm happy to go ahead with the merge. <!-- streview-comment:1741 --> -- 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]
