You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
(1) data fetch was initially given 3.8GB of RAM per job. More 95% of these jobs will not run out of memory with 3.8G. There is a small subset of jobs that have run out even with 8G, so this has been raised to 10G. It would be good to find the job types that consistently use >4GB (saved in job metric data) so that these jobs can be given mem: 4 and only the others have 10G - it will make it easier to schedule
(2) Give data fetch jobs a higher slurm priority than other jobs.
The text was updated successfully, but these errors were encountered:
(1) data fetch was initially given 3.8GB of RAM per job. More 95% of these jobs will not run out of memory with 3.8G. There is a small subset of jobs that have run out even with 8G, so this has been raised to 10G. It would be good to find the job types that consistently use >4GB (saved in job metric data) so that these jobs can be given
mem: 4
and only the others have 10G - it will make it easier to schedule(2) Give data fetch jobs a higher slurm priority than other jobs.
The text was updated successfully, but these errors were encountered: