Basic SLURM documentation
Here you will find a basic example job script for the SLURM batchsystem and further links for using the system.
Batch scripts are files that contain the definition of required hardware resources (see #SBATCH lines) and the code you wish to run on an HPC node.
These batch scripts are submitted to Slurm using the a unix console with the command sbatch. Using the console is required to submit batch scripts using sbatch.
Basic MPI job script:
#!/usr/local_rwth/bin/zsh ### Start of Slurm SBATCH definitions # ask for eight tasks #SBATCH --ntasks=8 # ask for less than 4 GB memory per (cpu) task=MPI rank #SBATCH --mem-per-cpu=3900M #M is the default and can therefore be omitted, but could also be K(ilo)|G(iga)|T(era) # name the job #SBATCH --job-name=MPI_JOB # declare file where the STDOUT/STDERR outputs will be written #SBATCH --output=output.%J.txt ### end of Slurm SBATCH definitions ### beginning of executable commands ### your program goes here (a.out is an example) $MPIEXEC $FLAGS_MPI_BATCH ./a.out |
Basic SLURM doumentation on hpc-wiki.info:
- What is a scheduler anyways and why should I bother using one?
- For basic documentation on SLURM please look here or at the official docs.
- Submit Jobs
- Write a job script
- Jobscript examples