element-hq / element-hq/matrix-authentication-service
Re-schedule jobs from lost workers
Open
A-Jobs
O-Uncommon
S-Minor
T-Defect
- Dominant language
- Rust
- Stars
- 280
- Forks
- 111
- Avg merge
- 3d 1h
- Merged PRs (30d)
- 25
Description
Currently, when a worker is lost, if it had any running job, it will keep them as 'running', never rescheduling them to another worker.
This should be done as part of the leader duties, to look at running jobs from workers that are shutdown and reschedule them
https://github.com/element-hq/matrix-authentication-service/blob/2590b7dade1540cab04c16f117cffa6618f618ab/crates/tasks/src/new_queue.rs#L690
Some context in #2785
Contributor guide
Assessment
This issue has not been assessed yet.