Details
-
Improvement
-
Status: Closed
-
Minor
-
Resolution: Fixed
-
1.12.0
-
None
Description
Follow the discussion in this PR[1].
Currently, ResourceManager#internalDeregisterApplication is a blocking call, so all the specific deployment implementation will do the Flink cluster deregister synchronously.
After FLINK-17024, we could get a safe thread pool to execute the I/O operations, including the "deregister". This ticket is to make the interface asynchronous. Benefit from this, the shutdown process will be a little faster.
[1]. https://github.com/apache/flink/pull/11427#issuecomment-613416114
Attachments
Issue Links
- is a child of
-
FLINK-15788 Various Kubernetes integration improvements
- Closed
- is superceded by
-
FLINK-23118 Drop Mesos support
- Closed