1

I am trying to run irace on Compute Canada, and when I used openmpi module, it always gave me this error message below:

mpirun was unable to launch the specified application as it could not access or execute an executable:

Executable: /scratch/irace/test.R

Node: niaXXXX

while attempting to start process rank 0.

My bash script is written

#!/bin/sh
#SBATCH --nodes=5
#SBATCH --ntasks-per-node=40
#SBATCH --mem=0
#SBATCH --mail-type=ALL
#SBATCH --output=/scratch/irace/irace-%j.out
#SBATCH --error=/scratch/irace/irace-%j.err

module load NiaEnv/2019b
module load gcc
module load r
module load openmpi

cd /scratch/irace
mpirun -np 5 /scratch/irace/test.R --parallel 4

When I submitted this job script, the run failed. However, when I replaced the last statement

mpirun -np 5 /scratch/irace/test.R --parallel 4

with the statement

Rscript ./test.R 

, it would run all good and give me the results. What are the causes of this and how can I fix this bug and get it run on Compute Canada?

Phoebe
  • 53
  • 5
  • Is `/scratch/irace/test.R` on a shared filesystem accessible by **all** the compute nodes? if you have any doubts, you'd rather ask your sysadmin/helpdesk. – Gilles Gouaillardet Mar 15 '23 at 00:17
  • How can I see if the test.R is on a shared filesystem? I transfer it from the local machine to the Linux environment. If it's not on a shared filesystem, should I contact the admin working at Compute Canada? Thanks! – Phoebe Mar 15 '23 at 02:35
  • you can `df /scratch/irace/test.R`, both on your frontend and job script, then compare the outputs. If you cannot find this information in the doc, then yes, ask sysadmin/helpdesk how to use the shared filesystem for MPI jobs. – Gilles Gouaillardet Mar 15 '23 at 03:00

0 Answers0