--- For OpenFoam and Code_Aster ---
This manage system for Gorina cluster uses Slurm, all jobs must be startet with Slurm !
Slurm head node .....: gorina11
Slurm working nodes .: gorina11-38
- Slurm commands
List nodes and status: sinfo
List jobs ...........: scontrol show jobs
List queue ..........: squeue
Simple test .........: srun -p cpu20 -N3 /bin/hostname
Cancel job ..........: scancel job
Command shell .......: srun -p cpu20 --pty $SHELL
Requesting two GPU ..: srun -p gpuA100 --gres=gpu:2
- Slurm partitions (All nodes have 256 Gbyte of memory.):
cpu36 = 8 nodes with 36 cpu cores @ 2.1GHz, job max runtime is one week
cpu28 = 2 nodes with 28 cpu cores @ 1.7GHz, job max runtime is one day
cpu20 = 18 nodes with 20 cpu cores @ 2.4GHz, job max runtime is two weeks
!!! IMPORTANT !!!
All jobs must be started from your /bhome/.. (or your /local/home).
This is very important if you run OpenFOAM or if your jobs do a lot of
file read or write, or creates big files. If unsure about how much IO
your job will do, then use /bhome.
Do NOT just run from your HOME, your jobs won't run as fast as they could, and they will bog down the /home file server for all of us...
Jobs causing heavy load on the /home file server will be terminated
without warning...
Note Contact [Theodor] theo@ux.uis.no if you need direct access to a node for some (debugging?) reason.