The setup works fine if I submit jobs from the machine on which the pbs server is running. But, if I go to a compute node and try to submit a job, I get the following error:
qsub: Job rejected by all possible destinations
According to the torque webpage, I should change a line in qmgr to read:
set queue slf route_destinations = slf_exec@localhost
But after doing this, I am unable to submit any jobs from any host, including cdf30. The solution for now, is to change the line back to slf_exec and submit jobs only from cdf30, which is the machine running the pbs server.