[ncl94]

Dear foarmers,


I am having a problem to execute openfoam solver in parallel using mpirun. In my private laptop, I have a single network with multiple cores (8), hence using the command mpirun is not a problem and the simulation runs in parallel smoothly. However, when I try to run in my university's server which is composed of multiple nodes, communication issues arise and the simulation gets stuck before start printing the output of each step. I believe the problem is because the server has several networks and it becomes an issue for mpirun to handle the parallelization. Any ideas on how to run OF in parallel for such cases?


Best,



PS: different parallelization methods (scotch, simple, etc...) gave me the same problem.

[Yann]

Hello,

One way to deal with multiple nodes would be something like this, lets say running 128 cores on 2 nodes:

mpirun -np 128 -H node1:64,node2:64 yourSolver -parallel

But this really depends on how your university network/server is configured and managed.
You should probably ask to your University IT department, because we don't know how your university server works, and how you are supposed to start a job on it. They might use a job scheduler such as SLURM or other.