[torqueusers] submitted jobs not running all nodes

Mark Moorcroft mark.w.moorcroft at nasa.gov
Fri Jun 21 12:56:11 MDT 2013



It turns out the version of openmpi shipping with Rocks does not support 
Torque. The previous Torque package for Rocks had a "hack" that allowed 
it to work without rebuilding it. Apparently whatever they were doing 
before is not working now. I rebuilt openmpi and things are working as 
expected again.




> I have been trying to test a new "build" of a torque/maui package for Rocks. I am running a head node and 2 test nodes in a VMWare environment under CentOS6. The issue I'm seeing that also seems to have been reported on the torque list by someone as well is that everything appears to go as expected but all the work units end up on 1 node. Maui reports the load on that node is too high. I am pasting a variety of log and command outputs here in case someone has any useful observations or suggestions.





More information about the torqueusers mailing list