...
| Code Block |
|---|
$ sinfo -p thor PARTITION AVAIL TIMELIMIT NODES STATE NODELIST thor up infinite 16 4 alloc thor[005011-012,015-016,022-023],thorbf0thorbf2a[15-16,22-23011-012] thor up infinite 4876 idle thor[001-004010,013-014,017-021,024032],thorbf2a[001-010,013-032],thorbf0thorbf3a[01-14,17-21,24-32] 001-016] $squeue -p thor JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON) 494553 139854 thor thor interact ... R 1:05:49 8 thor[005-012] 139853 thor interact ... R 1:40:5007 4 thor[015001-016002],thorbf0thorbf3a[15001-16002] 494552 139849 thor thor interact ... R R 325:15:3851 4 thor[022011-023012],thorbf0thorbf2a[22011-23012] |
To allocate nodes interactively
...
| Code Block |
|---|
# CPU nodes only $ salloc -N 2 -p thor --time=1:00:00 -w thor001,thor002dthor002 # CPU and BlueField nodes $ salloc -N 24 -p thor --time=1:00:00 -w thor001,thorbf001thor00[1-2],thorbf3a00[1-2] |
To submit a batch job
| Code Block |
|---|
# CPU nodes only $ sbatch -N 4 -p thor --time=1:00:00 -w thor00[1-4] <slurm script> # CPU and BlueField nodes $ sbatch -N 4 -p thor --time=1:00:00 -w thor00[1-2],thorbf00thorbf2a00[1-2] <slurm script> |
Note: Thor cluster has NVIDIA BlueField-2 & BlueField-3 cards with ARM processors on it. Those adapters can also be seen in slurm as “nodes” marked with thorbf0thorbf2a0[01-32] & thorbf3a0[01-16], while the hosts are named thor[001-032].
...