Closed chapman39 closed 3 months ago
We can change the deadline time to whatever makes sense. We could also use the immediate
option instead.
Sort of a funny way to test this argument is by allocating 60 minutes and setting the deadline to an hour from now - requiring you immediately get an allocation.
[meemee@ruby965:repo]$ s1 -t60 --deadline=now+1hour echo "hello"
srun -N1 -Aeng -t60 --deadline=now+1hour echo hello
srun: job 2630692 queued and waiting for resources
srun: Job has been cancelled
srun: error: Unable to allocate resources: Job/step already completing or completed
All modified and coverable lines are covered by tests :white_check_mark:
Project coverage is 90.63%. Comparing base (
dad99cd
) to head (cf8ab9d
).
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
With this PR, CI jobs now have a deadline on Quartz. This way, if jobs get stuck on the queue after gitlab fails, it won't clog up the system. I choose an hour because that's the same deadline as Gitlab.
If
--deadline=now+1hour
and-t25
, the job will dequeue after 35 minutes (60-25).I couldn't find a similar argument on
lalloc
/bsub
commands (Lassen).Also, I snuck in some more copyright updates.