Izeren opened a new pull request, #26663:
URL: https://github.com/apache/flink/pull/26663
## What is the purpose of the change
Address local recovery issues when Adaptive scheduler is enabled.
1. Pass latest completed checkpoint in addition to execution graph to
`StateSizeEstimates` (that is needed because execution graph goes through
cancelling/cancelled state and checkpoint coordinator is `nulled` by the time
we run calculations).
2. Assign positive priority score to allocations that have overlapping key
groups even when state size is zero (currently we would only give priority
score if `managedKeyedState` is present, but local recovery semantics doesn't
`require` state presence).
**Context**: When job can be recovered locally, we should keep slot
allocation after restart to maintain
## Verifying this change
`LocalRecoveryITCase#testRecoverLocallyFromProcessCrashWithWorkingDirectory`
now passes when `AdaptiveScheduler` is enabled.
## 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: issues-unsubscr...@flink.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
