-
Notifications
You must be signed in to change notification settings - Fork 0
Compute Canada queue priority
Gavin Douglas edited this page Oct 25, 2022
·
2 revisions
The sshare -l -A command is important for understanding our lab (and each individual person's) priority in the SLURM queue.
This is an example usage for assessing the priority of the def-shapiro account:
[gdouglas@narval2 projects]$ sshare -l -A def-shapiro_cpu
Account User RawShares NormShares RawUsage NormUsage EffectvUsage FairShare LevelFS GrpTRESMins TRESRunMins
-------------------- ---------- ---------- ----------- ----------- ----------- ------------- ---------- ---------- ------------------------------ ------------------------------
def-shapiro_cpu 1 0.000000 555742879 0.008608 0.008608 0.000001 cpu=815652,mem=20695904766,en+
def-shapiro_cpu gdouglas 1 0.041667 1274823 0.000020 0.002294 0.008224 18.164050 cpu=0,mem=0,energy=0,node=0,b+
The key column is LevelFS, which is added only when the -l option is used. It represents the relative usage of this account vs all other accounts (0-1 when overused, which means that priority will be lower, and > 1 when priority is preferred).
So in the above example the overall def-shapiro priority is low, but the user gdouglas would otherwise have high priority (I swear I didn't choose this example to be passive-aggressive ;), I just noticed now that this is what it shows).
This is the description of LevelFS from the sshare manual:
This is the association's fairshare value compared to its siblings, calculated as Norm Shares / Effectv Usage.
If an association is over-served, the value is between 0 and 1. If an association is under-served,
the value is greater than 1. Associations with no usage receive the highest possible value, infinity.