Best way running OMA using SLURM

Question

[dimar.gonzalez.soto]

Hi everyone, I'm executing OMA through a cluster using a bash script mentioned in the documentation of OMA standalone. For parallelization I'm using SLURM with this script:

sbatch --array=1-100 -N1 <<EOF
#!/bin/sh
export NR_PROCESSES=100
/usr/local/bin/OMA/bin/OMA
EOF

SLURM is running only one OMA proccess per node, I want to execute all 100 proccess at same time. How can I achieve that?

[GenoMax]

Is this all there is to the script? What is your actual OMA command line with options/inputs/outputs?

[dimar.gonzalez.soto]

I have not put input/output to the OMA executable

[GenoMax]

How will you actually use OMA for analysis of your data if you don't tell it what to analyze.

You seem to have copied the section about SLURM parallelization from OMA Standalone help page but are not actually constructed the OMA command line for running the analysis.

[Adrian Altenhoff]

There is a default parameter file that is used in case there are no arguments

[GenoMax]

I see. @adrian: Thanks for the clarification.

[dimar.gonzalez.soto]

Ok I will construct a command properly

Answer 1

[Adrian Altenhoff]

Hi

your command should indeed start 100 jobs of oma standalone in parallel. however the cluster configuration specifies how much resources a job should be allocated by default if not otherwise specified. I assume that your configuration allows to use a lot of resources and hence it will only start one job per node. This should not be a problem as long as you have enough nodes in your cluster... However, you will block others from using the cluster if you don't specify tight resource limits (see man sbatch; in particular --mem option)

Best wishes Adrian

[dimar.gonzalez.soto]

if start one process per node, should be it about slurm configuration?