MPI fail on multiprocess run

[fracenvi]

Hello,

I'm trying to run a case on a multiprocess setup, but whatever the amount of processors i provide it always gives me back:

--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD
with errorcode 1.

NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.
--------------------------------------------------------------------------

I built that versione of the code with:

Code: Select all

./code_saturne/configure --prefix=/home/cenvinzf@coria.fr/cs_openmpi --with-hdf5=/home/cenvinzf@coria.fr/Desktop/CS_4COUPLING/CS_LIB/hdf5/install_mpi --with-med=/home/cenvinzf@coria.fr/Desktop/CS_4COUPLING/CS_LIB/med/install_mpi --with-mpi=/home/cenvinzf@coria.fr/Desktop/OpenMPI --enable-mpi --enable-debug CC=/home/cenvinzf@coria.fr/Desktop/OpenMPI/bin/mpicc CXX=/home/cenvinzf@coria.fr/Desktop/OpenMPI/bin/mpicxx FC=/home/cenvinzf@coria.fr/Desktop/OpenMPI/bin/mpif90

And i made sure the MPI paths are always visible through the .bashrc file. Where might be the problem then?

Thanks to anyone who can help,
Francesco

[StandardRANSUser001]

Hi Francesco,

Sometimes I find it handy to run the install_saturne.py script, and then if something crashes, copy the configure command that is generated in this script. I am unsure whether I needed the CXX and FC commands were needed in my install. I have found that it is also important to get the order of the packages correct (The script handles this).


Best regards,
Sean Hanrahan

[Yvan Fournier]

Hello,

Don't forget the usual troubleshooting recommendations, as described in the BPG's (https://www.code-saturne.org/documentat ... oting.html).

What do the run_solver.log or error* files say ?

Regards,

Yvan

[fracenvi]

Hello,

I actually managed to sort the bug out, as the command i was using to launch the code was not passing the correct number of processors needed to perform the computation.

Regards,
Francesco