Why the PMIX ERROR, when job runs well?

Discovery Cluster Parallel Computing
1

I have similar errors for each node X tasks-per-node in the SLURM .out file. This seems to be no problem and I am curious as to why this is.

[d05-28.hpc.usc.edu:32722] PMIX ERROR: BAD-PARAM in file base/bfrop_base_unpack.c at line 692
[d05-28.hpc.usc.edu:32722] PMIX ERROR: BAD-PARAM in file dstore_base.c at line 2225

Context SLURM job file:

#!/bin/bash
#SBATCH --nodes=2
#SBATCH --ntasks-per-node=2
#SBATCH --cpus-per-task=1
#SBATCH --mem=50GB
#SBATCH --time=00:04:10
#SBATCH --partition=main
module load gcc/11.3.0
module load openblas/0.3.20
module load openmpi/4.1.4
module load pmix/3.2.3
module load jags/4.3.0
module load r/4.2.1
export OMP_NUM_THREADS=$SLURM_CPUS_PER_TASK
cd $SLURM_SUBMIT_DIR
echo "working directory = "$SLURM_SUBMIT_DIR
# ulimit -s unlimited
srun --mpi=pmix_v2 -n $SLURM_NTASKS Rscript --vanilla /path/C000_01.R

…

2

As far as I know it’s a benign issue. See open issue: PMIX ERROR: BAD-PARAM in file base/bfrop_base_unpack.c at line 692 · Issue #8548 · open-mpi/ompi · GitHub

The messages only occur when using srun, but the issue should not impact performance. You could use mpirun instead:

mpirun -n $SLURM_NTASKS Rscript --vanilla /path/C000_01.R
3

That seems to work better as there are no ERROR (albeit benign) messages. Just as you say:

mpirun -n $SLURM_NTASKS Rscript --vanilla /path/C000_01.R

Thank you!