I have an MPI application written in C++. When I run the application using mpirun or mpiexec on my local machine shell, it works fine. The problem is that the mpi application should be executed after a request from the network (e.g. HTTP request) is received. Assuming that the application server is written in python, how can I execute the mpi application using mpiexec command from a python script? I used subprocesses.run()
command but nothing happens.
My general question: What is the best way to run an MPI application in client/server architecture?
Thank you
I found the solution. In the main python script, MPI should not get imported from mpi4py
package. Otherwise subprocess.run("mpiexec")
does not do anything.