Hi,
Just installed the CS 3.0.1 in a PBS cluster. When I use just one node with 16 cores, it works quite well. But when try to run on 2 nodes with 16 cores each, I got a problem and the cs_solver cannot be started. I got the following message in the .o file.
Any suggestions will be appreciated!
Code_Saturne is running
***********************
Version: 3.0.1
Path: /apps/code-saturne/3.0
Result directory:
/export72/work/xhan1/test/case/RESU/20131202-1110
Parallel code_saturne on 32 processes.
****************************************
Compiling user subroutines and linking
****************************************
****************************
Preparing calculation data
****************************
***************************
Preprocessing calculation
***************************
**********************
Starting calculation
**********************
Defaulting to ethernet
No MPI process pinning
The specified executable [-n] cannot be run. -n doesn't exist. Either filename is wrong, or it's not in your PATH.
PATH is currently /apps/code-saturne/3.0/bin/:/apps/python/2.7.3//bin:/usr/lib64/qt-3.3/bin:/bin:/usr/bin:/usr/local/sbin:/usr/sbin:/sbin:/opt/pbs/default/bin:/home/xhan1/bin:/opt/dell/srvadmin/bin:/apps/code-saturne/3.0/bin:/apps/code-saturne/3.0/:/apps/intel/2013sp1/bin:/apps/intel/2013sp1/composer_xe_2013/pkg_bin/intel64:/apps/intel/ict/mpi/3.1.038/bin:/export72/work/xhan1/test/case/RESU/20131202-1110
****************************
Saving calculation results
****************************
problem when start cs 3.0 running on multi nodes
Forum rules
Please read the forum usage recommendations before posting.
Please read the forum usage recommendations before posting.
-
- Posts: 4220
- Joined: Mon Feb 20, 2012 3:25 pm
Re: problem when start cs 3.0 running on multi nodes
Hello,
It seems for some reason, the mpiexec or mpirun executable is not found when you are using multiple nodes.
Could you try using the "--with-mpi-exec=<PATH>" configure option when installing the code to specify the prefix for MPI executables ? It may help.
If this is not enough, could you both:
- post the run_solver.sh script from the execution directory
- add a line containing: "env"
near the beginning of your runcase script, and post the main output of you PBS job (it should the contain your whole environment, which we can source to try to reproduce the issue even with no local install of PBS).
Regards,
Yvan
It seems for some reason, the mpiexec or mpirun executable is not found when you are using multiple nodes.
Could you try using the "--with-mpi-exec=<PATH>" configure option when installing the code to specify the prefix for MPI executables ? It may help.
If this is not enough, could you both:
- post the run_solver.sh script from the execution directory
- add a line containing: "env"
near the beginning of your runcase script, and post the main output of you PBS job (it should the contain your whole environment, which we can source to try to reproduce the issue even with no local install of PBS).
Regards,
Yvan