[VL] Hash shuffle may leave local spill payloads unmerged when partition-buffer eviction writes pids out of order
- Dominant language
- Scala
- Stars
- 1.6k
- Forks
- 657
- Avg merge
- 2d 21h
- Merged PRs (30d)
- 85
Description
### Backend
VL (Velox)
### Bug description
Hash shuffle can fail during ShuffleWriter.stop() with:
```
Native shuffle write: ShuffleWriter stop failed - Invalid:
Merging from spill N is not exhausted. pid: X
```
### Gluten version
_No response_
### Spark version
None
### Spark configurations
_No response_
### System information
_No response_
### Relevant logs
```bash
```
Contributor guide
Research direction
Start at ShuffleWriter.stop() and trace the hash-shuffle path involving partition-buffer eviction and local spill merging. Use the reported “Merging from spill N is not exhausted” failure as the symptom to reproduce or follow; done means stop succeeds without leaving spill payloads unmerged when pids are written out of order.
Written by the indexing model from the issue text.
Assessment
- Domain
- data-engineering, distributed-systems
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 38/100