Closed nateharms closed 4 years ago
Merging #57 into master will increase coverage by
0.18%
. The diff coverage is68.13%
.
@@ Coverage Diff @@
## master #57 +/- ##
==========================================
+ Coverage 62.73% 62.91% +0.18%
==========================================
Files 27 27
Lines 4564 4635 +71
==========================================
+ Hits 2863 2916 +53
- Misses 1701 1719 +18
Impacted Files | Coverage Δ | |
---|---|---|
autotst/job/job_test.py | 98.63% <100%> (+0.16%) |
:arrow_up: |
autotst/job/job.py | 41.99% <62.82%> (+3.09%) |
:arrow_up: |
Continue to review full report at Codecov.
Legend - Click here to learn more
Δ = absolute <relative> (impact)
,ø = not affected
,? = missing data
Powered by Codecov. Last update dcebf50...792f17d. Read the comment docs.
Okay, I'm thinking this is okay to merge this in even though the patch wasn't successful. We only changed 15 seconds to 90 seconds so nothing should have changed. Any complaints?
Okay, I'm gonna do a rebase to fix the history and then add some tests to job.py
to account for the decrease in coverage, but then this should be good to go.
Okay, this is looking like all the checks pass and the coverage is good. I’m going to do a rebase to fix up the git history. But it should be ready for review soon.
Hi there @davidfarinajr and @rwest, can one of y'all give this a review when you get a chance? Thanks!
It looks like more changed than is described in the pull request description. Problem with rebasing?
@rwest sorry about that. This PR was originally just about a sleep issue but we broadened the scope of the PR after speaking with @davidfarinajr in person. The PR description now reflects these changes.
There was an issue where our rate of checking if a job was complete was too frequent and SLURM would frequently return errors looking like the following:
or
After speaking with some people at Research Computing at Northeastern, we were told that our code was running
squeue
andsbatch
too frequently and that we can check every minute (at minimum). This sets thetime.sleep()
to 90s so we avoid this issue. We also added changes to make sure that jobs are submitting by waiting for the queue to get smaller so our jobs don't overwhelm it.