I use SLURM as a job scheduler and queue for a small cluster (single node with 64 cores). To submit a batch job I use:
> sbatch run.sh
Where run.sh looks like:
#! /bin/bash
#SBATCH --ntasks=4
export OMP_THREAD_LIMIT=4
/home/Benchmarks/Graph500/omp-csr/omp-csr -s 23
However, when I submit 2 batch jobs one after the other I get:
> squeue
JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON)
29 debug run.sh anonymou PD 0:00 1 (Resources)
28 debug run.sh anonymou R 1:13 1 localhost
Each job needs only 4 cores, so both jobs should run. Maybe I misconfigured the slurm controller, the relevant lines from /etc/slurm.conf are:
# COMPUTE NODES
NodeName=localhost CPUs=64 Sockets=4 CoresPerSocket=8 ThreadsPerCore=2 State=UNKNOWN
PartitionName=debug Nodes=localhost Default=YES MaxTime=INFINITE State=UP
I would be thankful for any help/hint.