|Posted on Thursday, December 22, 2016 - 4:23 pm: |
My department recently purchased WebMO Enterprise for one of our researchers. I'm trying to get it to cooperate nicely with our cluster, which runs Torque 3.0.1 and Maui 3.2.6p21 on Red Hat Enterprise Linux 6. I'm using internal authentication at this point, and used the admin interface to configure several queues and point to our g09 and molpro installations.
I can now create a molecule with the editor and request a computation via WebMO (I'm not a chemist so I have no idea what I'm doing; I'm just leaving the defaults in place as I hit next). However, the submitted job kills the pbs_mom process on the client nodes. It can't even be restarted without clearing out the job information in /var/spool/torque/mom_priv/jobs/$jobnum.*
I'm working to determine what's causing the failure. It looks like Maui is assigning the job and handing the task off to Torque, which is then sending the run-script off to the compute node all correctly. That's where things break. I'm reasonably sure I can figure this out (WebMO has proven very challenging to get working correctly and has required several structural modifications to our cluster, but those are separate issues). Any help will be appreciated.
|Posted on Thursday, December 22, 2016 - 7:29 pm: |
Okay, figured it out: I had neglected to create the webmo user on the compute nodes, which must have made torque lose its little mind.