Sbatch running error with public-cpu

HPC Technical
1

Dear HCP Team,

On Baobab (public-cpu), I’m trying to run a MATLAB script with parallel processing via sbatch, but I’m having some issues, and I am wondering if the configuration parameters are valid or not.

When I try to launch the script, I get this message: sbatch: error: Batch job submission failed: Requested node configuration is not available

Please find below the parameters:

#!/bin/bash
#SBATCH --job-name=split_half
#SBATCH --output=split_half_%j.out
#SBATCH --error=split_half_%j.err
#SBATCH --partition=public-cpu
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=16 # how many cores MATLAB should use
#SBATCH --mem-per-cpu=10G # memory per core
#SBATCH --time=4-00:00:00 # wall‑time hh:mm:ss
#SBATCH --mail-type=BEGIN,END,FAIL

echo “=== JOB START: $(date) ===”
module load MATLAB/2022a

srun matlab -nodisplay -r “run(‘/home/users/a/awada/consensus_clustering_CAPs_deactivated_GVA_T1/e_Script_SplitHalf_ consensus_clustering_PARALLEL_SLURM.m’); exit;”

echo “=== JOB END: $(date) ===”

Can you please help or advice on how to best configure the parameters for running this?

many thanks in advance,
Jade

2

Dear @Jade.Awada on Baobab cluster, the public-cpu partition is very small yet and the nodes are old: they have 20 cores with 96GB of ram.

(baobab)-[root@admin1 ~]$ sinfo -p public-cpu -N -o "%n %c %m"
HOSTNAMES CPUS MEMORY
cpu237 20 96000
cpu238 20 96000
cpu239 20 96000
cpu240 20 96000
cpu241 20 96000
cpu242 20 96000
cpu243 20 96000
cpu244 20 96000

You are requesting 16 cores with 10G per core => 160GB which is more than the compute nodes have.

Your options are:

  • use less ram
  • use shared-cpu partition which is bigger and have more powerfull nodes
  • use Bamboo cluster which is newer, public-cpu have 128 cores and 512GB ram.