|Posted on Monday, March 21, 2016 - 4:02 am: |
I installed parallel NWChem 6.6 on my WebMO 15.0.003 server running CentOS 6.5. The webmo user can run NWC jobs directly with an "mpirun -np 2 /usr/local/nwchem/bin/nwchem input.in" command. But when the job is run through the WebMO interface, the job fails with a message "Cannot execute /usr/local/nwchem/bin/nwchem: No such file or directory". (Just to simply things, I had temporarily removed the -machinefile terms from run_nwchem.cgi. When they are left in, the job still fails.) SELINUX is disabled on this machine. Any hints on what I should try next?
Post Number: 498
|Posted on Monday, March 21, 2016 - 8:25 am: |
You may very well need the full path to mpirun (and maybe even adding some MPI libraries to LD_LIBRRARY_PATH) since these environment that WebMO is seeing is likely different than what you get on the command line.
Check the 'run_log' and other files in the job output directory for error messages.
|Posted on Monday, March 21, 2016 - 8:06 pm: |
JR, Thanks! That did it. I edited line 65 of run_nwchem.cgi to read:
$exec_command = "/usr/lib64/openmpi/bin/mpirun -np $nproc $nwchemBinary input.inp";
That is, added the full path of mpirun. The webmo daemon must not see the same environment variables and values as the webmo user. The PATH variable of this user includes /usr/lib64/openmpi/bin as the first entry, which should allow WebMO to find the executable.