...
PC2 cluster system use SLURM as a scheduler/workload manager.
sharing of compute nodes among jobs is enabled
cpu-cores, memory and GPUs are allocated exclusively to jobs
SMT (aka. hyperthreading) is disabled on the compute nodes
start your MPI jobs with
srun
and notmpirun
ormpiexec
You only see your own compute jobs. If you are the project administrator of a compute-time project you see all jobs of the project.
extra tools:
scluster
,spredict
,squeue_pretty
...
Line | Mandatory | Meaning | ||
---|---|---|---|---|
| YES | specify the time limit of your job. Acceptable time formats for | ||
| no, default is 1 | use at least the number of | ||
| no, default is 1 | run | ||
| no, default is memory-per-node/number-of-cores | memory per allocated cpu core, e.g. 1000M or 2G for 1000 MB or 2 GB respectively. | ||
| no, default is the file name of the job script | specify the | ||
| no, default is the normal partition | Submit jobs to the | ||
| not if you are only member of one compute-time project | specify the compute-time project | ||
| no, default is the default QoS of your project | Use the QoS | ||
| no, default value is | specify at which event you want a mail notification. | ||
| no | specify your mail that should receive the mail notifications |
...