-
Notifications
You must be signed in to change notification settings - Fork 207
Closed
Labels
bugSomething isn't workingSomething isn't working
Description
Steps to reproduce
- Create a run with many job submissions (e.g. a retry run that fails to provision due to no offers, may require adjusting
_PENDING_RETRY_DELAYSto generate job submissions quickly). - It's suspected this can lead to high memory usage but I'm yet to reproduce it locally.
Actual behaviour
These SELECTs may fetch too many rows:
.options(joinedload(RunModel.jobs)) dstack/src/dstack/_internal/server/background/tasks/process_runs.py
Lines 156 to 157 in fae73ce
selectinload(RunModel.jobs) .joinedload(JobModel.instance)
Expected behaviour
Select only the necessary (latest) submissions for every job.
dstack version
master
Server logs
Additional information
No response
Metadata
Metadata
Assignees
Labels
bugSomething isn't workingSomething isn't working