bgeng777 opened a new pull request, #29861: URL: https://github.com/apache/flink/pull/29861
## What is the purpose of the change Backport #29677 ([FLINK-41271](https://issues.apache.org/jira/browse/FLINK-41271)) to `release-1.20`. When the Beam data RPC terminates, a Python task can keep waiting for bundle output and an idle task can keep the failed worker alive. ## Brief change log Observe termination through the server factory provided by Beam 2.43.0. Notify runners sharing the factory, wake result waits and interrupt an in-progress bundle close. This branch tracks the closing thread directly and unregisters the listener before releasing the runner's factory or shared lease. ## Verifying this change Compiled the changed production and test classes against the release-line artifacts and ran both test classes with JUnit. All 17 tests passed, including shared-runner notification, interruption of bundle close, early cancellation and normal teardown. The flink-python Spotless and Checkstyle checks passed. ## Does this pull request potentially affect one of the following parts: - Dependencies, public API, serializers and S3 connector: No. - Runtime per-record path: Yes. Python runners check for a recorded data RPC failure. - Recovery: Yes. Tasks sharing a failed worker can fail and recover together. ## Documentation Bug fix. No user-facing configuration changes. - [X] Yes, generative AI tooling was used. Generated-by: Codex (GPT-6) -- 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]
