bowenli86 commented on PR #28716: URL: https://github.com/apache/flink/pull/28716#issuecomment-5169590731
@suez1224 Resolved, thanks for catching this. Managed-memory shutdown now first finishes this runner's active `RemoteBundle`, which removes its state registration even if other leases keep the shared `JobBundleFactory` alive. It then releases `sharedResources`; on the final lease, that closes the shared factory while the handler is still open. Only after both paths are quiesced do we close/gate `stateRequestHandler`. The non-managed path keeps its factory-first shutdown behavior. I also synchronized `flush()` so an async flush and `close()` cannot double-close the same bundle. Regression tests now cover the owned-factory path, final and non-final managed-memory leases, and concurrent `flush()` / `close()`; the concurrency test fails when synchronization is removed. Current head `8bedabaa159` has Azure CI SUCCESS. -- 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]
