[slurm-users] Segmentation fault when launching mpi jobs using Intel MPI

[Bob Smith]

Hello all,

I am having an issue submitting mpi jobs via sbatch using Intel MPI 2019
Update 1. The job ends with a segmentation fault immediately:

[user at head mpihello]$ cat mpihello-intelmpi.out.62
srun: error: node003: task 0: Segmentation fault
srun: error: node004: task 0: Segmentation fault
[mpiexec at node002] main
(../../../../../src/pm/i_hydra/mpiexec/mpiexec.c:2024): downstream was
killed by signal 11 (Segmentation fault)


Here is what the sbatch script looks like:

#!/bin/bash

#SBATCH --job-name=mpihello-intelmpi
#SBATCH --output=mpihello-intelmpi.out.%j
#SBATCH --error=mpihello-intelmpi.err.%j

#SBATCH -N 3 --tasks-per-node=4
#SBATCH --time=10:00
#SBATCH --mem-per-cpu=100

/shared/intel/compilers_and_libraries/linux/mpi/intel64/bin/mpirun
/home/usr/mpihello/mpihello-intelmpi


To simplify things, I have tried launching a simple 'hostname' command in
place of mpihello-intelmpi, but I get the same segmentation fault issue.
Addtionally, I can directly launch the simple mpihello program as well as a
'hostname' command on the command line without issue.

Any ideas on what is going on?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.schedmd.com/pipermail/slurm-users/attachments/20190206/c08c0eb9/attachment.html>