Integrating IntelMPI with PBS scheduler

[Alaa_hassan]

Hello everyone,
When I try to submit a multi-node LAMMPS job on PBS, it launches separate processes (4 independant LAMMPS instance) instead of launching parallel MPI processes. Can you please let us know your insights on that issue?

Job:
====
#PBS -N 5-5-08
#PBS -l select=2:ncpus=2:mpiprocs=2
#PBS -l walltime=1:00:00
#PBS -l place=scatter:excl
#PBS -j oe
cd $PBS_O_WORKDIR

module load intelmpi-2026.0.0/mpi/2021.18
module load lammps-2025.0.7-intel
echo "Nodefile:"
cat $PBS_NODEFILE
NPROCS=$(wc -l < $PBS_NODEFILE)
time mpirun lmp_mpi -in in1.lj

Simplified Output:
==============
Nodefile:
xx-cn-001
xx-cn-001
xx-002
xx-002
No PMIx server was reachable, but a PMI1/2 was detected.
If srun is being used to launch application, 4 singletons will be started.
> The output get replicated four times

[TobiasK]

@Alaa_hassan srun is for SLURM not for PBS. Why are you trying to use srun?

[Alaa_hassan]

Yes I know that. This is the output of the job, I am not using srun as shown in the job but I am facing this issue that it will start 4 single separate processes not MPI processes.

[TobiasK]

ok, other question, why is srun even available on a system that is using PBS?
in any way, please provide the output of 
I_MPI_DEBUG=10 I_MPI_HYDRA_DEBUG=1 mpirun IMB-MPI1 pingpong