[SLURM] Checking consumed resources on SLURM
Phil Kauffman via Slurm
slurm at mailman.cs.uchicago.edu
Tue May 8 11:42:01 CDT 2018
Take a look at 'man squeue' and look at the -O option. It has a ton of useful things you can check.
Here are a couple of examples that show you how to check the consumed generic resources (GRES) on any machine in the cluster:
Here we can see that 'pramodkm' is currently consuming 3x GPUs and 'ady' is consuming only 1x.
kauffman at hester2:~$ squeue -O username,nodelist,gres
USER NODELIST GRES
ady gpu3 gpu:1
pramodkm gpu3 gpu:3
alexliu0809 slurm4 (null)
alexliu0809 slurm4 (null)
alexliu0809 slurm3 (null)
alexliu0809 slurm3 (null)
alexliu0809 slurm3 (null)
kauffman at hester2:~$ squeue -O jobid,partition,name,username,nodelist,gres,state
JOBID PARTITION NAME USER NODELIST GRES STATE
78237 titan bash ady gpu3 gpu:1 RUNNING
78174 titan bash pramodkm gpu3 gpu:3 RUNNING
78322 general xato-m-10 alexliu0809 slurm4 (null) RUNNING
78321 general xato-m-9 alexliu0809 slurm4 (null) RUNNING
78320 general xato-m-8 alexliu0809 slurm3 (null) RUNNING
78319 general xato-m-7 alexliu0809 slurm3 (null) RUNNING
78318 general xato-m-6 alexliu0809 slurm3 (null) RUNNING
--
Phil Kauffman
Systems Administrator
Dept. of Computer Science
University of Chicago
kauffman at cs.uchicago.edu
773-702-3913
More information about the Slurm
mailing list