[cons013]

I'm having trouble getting fluent to accept multiple nodes on a hpc cluster (128 core nodes). I tried this solution:

https://forum.ansys.com/forums/topic...c-under-slurm/

for 2 nodes, total of 256 processes, but instead of fluent opening 128 processes on two nodes, it spawns 256 individual instances of fluent. Has anyone managed to get fluent working with SLURM? I tried using srun and not using srun, no difference.

[cilliant]

Any update here? I am facing the exact same issue

I believe using this command in your slurm batch script can work, but is not working in my case. (insert it after module loaded)


scontrol show hostnames $SLURM_JOB_NODELIST > hostNamesFile.txt
## Task execution
fluent 3ddp -t240 -g -cnf=hostNamesFile.txt -ssh -i TUI_bfer_star_fine.txt -platform=intel -mpi=intel

[cons013]

Originally Posted by cilliant

Any update here? I am facing the exact same issue

I believe using this command in your slurm batch script can work, but is not working in my case. (insert it after module loaded)


scontrol show hostnames $SLURM_JOB_NODELIST > hostNamesFile.txt
## Task execution
fluent 3ddp -t240 -g -cnf=hostNamesFile.txt -ssh -i TUI_bfer_star_fine.txt -platform=intel -mpi=intel

Using -mpi=Cray (or cray? not sure) it is working, but the HPC team at the cluster I use sorted this all for me. I am not sure what they changed, if I ever find out I will let you know.