I have the ff sbatch script:
#SBATCH --job-name=nrm3
#SBATCH --partition=batch
#SBATCH --nodes=8
#SBATCH --ntasks=40
#SBATCH --ntasks-per-node=5
#SBATCH --output=OUTPUTlognrm.3.log
for ((i = 0; i < 40; ++i)); do
#prepare config file $i
echo "config_params" > "config_${i}"
srun -N1 -n1 java main "config_${i}" &
done
wait
I’m trying to run the Java program with their config files in parallel on 8 nodes; each node gets 5 tasks. I checked sstat and htop, and the job steps seem to be running sequentially on one node. How can I fix this?