rmetzger opened a new pull request #13424:
URL: https://github.com/apache/flink/pull/13424
## What is the purpose of the change
If a job gets stuck in the INITIALIZING state, we can not properly cancel it
in this stage.
What effectively happens is that we are waiting for the initialization to be
over before cancelling the job.
This change runs the initialization in a separate thread, that we can
interrupt in case the user issues a cancel call.
This approach does not guarantee that the initialization is cooperatively
aborting the current operation. If it is not cooperating, we will wait and
follow the old approach of cancelling.
## Brief change log
- Introduce `InitializingJobManagerRunner` class that hides the thread
handling
- adjusted `Dispatcher` and `DispatcherJob` to cooperate with the new class
## Verifying this change
I adjusted some existing tests.
This is a draft PR, once it seems to be ok, I'll add a test for
`InitializingJobManagerRunner`.
## Does this pull request potentially affect one of the following parts:
- Dependencies (does it add or upgrade a dependency): (yes / *no*)
- The public API, i.e., is any changed class annotated with
`@Public(Evolving)`: (yes / *no*)
- The serializers: (yes / *no* / don't know)
- The runtime per-record code paths (performance sensitive): (yes / *no* /
don't know)
- Anything that affects deployment or recovery: JobManager (and its
components), Checkpointing, Kubernetes/Yarn/Mesos, ZooKeeper: (*no* / no /
don't know)
- The S3 file system connector: (yes / *no* / don't know)
## Documentation
- Does this pull request introduce a new feature? (yes / *no*)
- If yes, how is the feature documented? (not applicable / docs /
*JavaDocs* / not documented)
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
