1996fanrui opened a new pull request, #24012:
URL: https://github.com/apache/flink/pull/24012
## What is the purpose of the change
[FLINK-33960][Scheduler] Fix the bug that Adaptive Scheduler doesn't respect
the lowerBound when one flink job has more than 1 tasks
When we using the adaptive scheduler and the rescale api, users will set the
lowerBound and upperBound for each job vertices. And users expect the
parallelism of all vertices between lowerBound and upperBound.
But when one flink job has more than 1 vertex, and resource isn't enough.
Some of lowerBound won't be respect.
How to reproduce this bug:
One job has 2 job vertices, we set the resource requirements are:
Vertex1: lowerBound=2, upperBound=2
Vertex2: lowerBound=8, upperBound=8
They are same slotSharingGroup, and we only 5 available slots. This job
shouldn't run due to the slots cannot meets the resource requiremnt for vertex2.
But the job can runs, and the parallelism of vertex2 is 5.
### Why does this bug happen?
Flink calculates the minimumRequiredSlots for each slot sharing group, it
should be the max lowerBound for all vertices of current slot sharing group.
But it's using the on the minimum lowerBound.
## Brief change log
- [FLINK-33752][JUnit5 migration] Migrate SlotSharingSlotAllocatorTest to
Junit5 and Assertj
- [FLINK-33960][Scheduler] Fix the bug that Adaptive Scheduler doesn't
respect the lowerBound when one flink job has more than 1 tasks
- [FLINK-33960][Scheduler][refactor] Simplify the implementation of
SlotSharingGroupMetaInfo
## Verifying this change
- Added
`SlotSharingSlotAllocatorTest#testDetermineParallelismWithLowerBoundsInsufficientSlotsForPartialVertices`
## Does this pull request potentially affect one of the following parts:
- Dependencies (does it add or upgrade a dependency): no
- The public API, i.e., is any changed class annotated with
`@Public(Evolving)`: no
- The serializers: no
- The runtime per-record code paths (performance sensitive): no
- Anything that affects deployment or recovery: JobManager (and its
components), Checkpointing, Kubernetes/Yarn, ZooKeeper: no
- The S3 file system connector: no
## Documentation
- Does this pull request introduce a new feature? 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]
