apache / apache/datafusion-ballista
Ballista: Executor must return statistics in CompletedTask / CompletedJob
- Dominant language
- Rust
- Stars
- 2.1k
- Forks
- 320
- Avg merge
- 1d 22h
- Merged PRs (30d)
- 66
Description
**Is your feature request related to a problem or challenge? Please describe what you are trying to do.**
We cannot fix the shuffle mechanism until we have partition stats, or ShuffleReaderExec will attempt to read empty partitions, causing an error.
**Describe the solution you'd like**
Scheduler should receive partition stats and only try and read from non-empty shuffle partitions.
**Describe alternatives you've considered**
As a workaround we could write empty shuffle files for empty partitions.
**Additional context**
None
Contributor guide
Research direction
Trace the scheduler and executor paths named in the issue, including CompletedTask, CompletedJob, and ShuffleReaderExec. First determine where partition statistics are produced and passed between these components; done means the scheduler receives those statistics and avoids reading empty shuffle partitions.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- rust
- Domain
- distributed-systems
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100