mpi job runs fine when submitted via scheduler, but fails when run by hand

[Faraz H]

I figured it out by using the full path to the cp2k executable and passing 
the LD_LIBRARY_PATH environment variable. I had to do mpirun -x 
LD_LIBRARY_PATH . For some reason I do not have to do this in my submission 
script. Perhaps the scheduler takes care of exporting paths and env. 
variables for you?

On Friday, July 21, 2017 at 9:56:06 AM UTC-4, Faraz H wrote:
>
> I have a simple job submission script which runs fine. However, when I try 
> and run the same script by hand ( outside the scheduler), it fails with the 
> error below. By hand I mean that I take the nodes offline, login to the 
> first node, load the cp2k/open mpi modules, keep same machine file, then 
> run the command:
>
> mpirun -machinefile machines -np 40 --mca btl openib,self cp2k.popt 
> test.inp
>
>
> mpirun was unable to find the specified executable file, and therefore
> did not launch the job.  This error was first reported for process
> rank 20; it may have occurred for other processes as well.
>
> NOTE: A common cause for this error is misspelling a mpirun command
>       line parameter option (remember that mpirun interprets the first
>       unrecognized command line token as the executable).
>
> Node:       lustwzb5
> Executable: /gpfs/app/software/cp2k/3.0.0/exe/Linux_x86-64-intel/cp2k.popt
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20170802/19d1fa74/attachment.htm>