zhuzhurk opened a new pull request #16842:
URL: https://github.com/apache/flink/pull/16842
## What is the purpose of the change
This fix is to avoid StackOverflowException which will lead to JM crash.
When requested slots are not fulfilled in time, task failure will be
triggered and all related tasks will be canceled and restarted. However, in
this process, if a task is already assigned a slot, the slot will be returned
to the slot pool and it will be immediately used to fulfill pending slot
requests of the tasks which will soon be canceled. The execution version of
those tasks are already bumped in DefaultScheduler#restartTasksWithDelay(...)
so that the assignment will fail immediately and the slot will be returned to
the slot pool and again used to fulfill pending slot requests. StackOverflow
can happen in this way when there are many vertices, and fatal error can happen
and lead to JM crash.
To fix the problem, this PR will cancel the pending requests of all the
tasks which will be canceled soon(i.e. tasks with version bumped) before
canceling these tasks.
## Verifying this change
This change added tests and can be verified as follows:
- *Added unit test
DefaultSchedulerTest#pendingSlotRequestsOfVerticesToRestartWillNotBeFulfilledByReturnedSlots*
## 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, ZooKeeper: (**yes** / 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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
