Zakelly commented on PR #29445: URL: https://github.com/apache/flink/pull/29445#issuecomment-6125916727
> @Zakelly I don't think this covers what is left after [FLINK-40921](https://issues.apache.org/jira/browse/FLINK-40921): this branch still crashes in the same materialization path when the fork stays alive 254.6 s after `ChangelogStateBackendMigrationTest` starts. Can we close those backends instead, as [FLINK-40921](https://issues.apache.org/jira/browse/FLINK-40921) did for `StateBackendTestBase`? Agreed, we should close the remaining backends in the tests. I'll create a separate PR for that. In these tests, `SyncMailboxExecutor` runs `initMaterialization()` on the periodicExecutor thread, so it can overlap with close(). In production, materialization initialization and normal cleanup run serially on the task thread. This wait therefore only helps with concurrent race between `close()` and `initMaterialization()`. I'd suggest we keep it to reduce the possibility. -- 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]
