j1wonpark commented on PR #4197: URL: https://github.com/apache/amoro/pull/4197#issuecomment-4881929621
Thanks for the review @czy006! Both points are addressed: **1.** Fixed in 99367db3 — the optimizer group property `shutdown-timeout-ms` is now appended as `-st` in the startup args (same pattern as `-hb`), and the K8s `terminationGracePeriodSeconds` is derived from the same property instead of re-parsing the CLI string, so the two can never diverge. Documented in `managing-optimizers.md`. **2.** Added in 21d75def — `FlinkExecutor.close()` now drains the in-flight task up to the shutdown timeout (capped below `task.cancellation.timeout` to avoid failing the TaskManager) before force-interrupting. Since Flink cancels the toucher source before the operators, the drain also keeps the registration alive by touching AMS with the existing token at the heartbeat interval (e34d2bc2) — otherwise AMS would expire the optimizer after `optimizer.heart-beat-timeout` and reset the in-flight task mid-drain. It only touches, never re-registers, so once `close()` returns AMS cleans up through the normal expiration path. I also fixed several races I found in the new shutdown paths (b4e5ae37). -- 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]
