[torqueusers] Submitting jobs to multi-cpu nodes
carheden at cira.colostate.edu
Thu Feb 9 16:08:47 MST 2006
I've got Torque 2.0.0 running on an 8-node linux cluster. Each node has
2 CPUs. I've configured my nodes with np=2. I've got two problems/questions.
1) Everything works more or less as expected for jobs requiring less
than 8 nodes. For example, I can submit 4 jobs that each require 4 nodes
or 2 jobs that require 8 nodes and they will all run in parallel. When I
submit a 16-node job, however, I get the error message:
qsub: Job exceeds queue resource limits
I've just upgraded from Torque 1.2.0 and I could swear that it let me
submit 16-node jobs.
I imagine this is just a matter of telling torque that I need 16
processors as opposed to exclusive access to 16 nodes. Any clues on how
to do that?
2) When I submit two 4-node jobs, they end up running in parallel on the
first four nodes. How do I have torque fill all nodes with one job
before submitting jobs to multi-cpu nodes?
More information about the torqueusers