Occasionally we need to restart the pbs_server daemon on our clusters to implement config changes, such as adding text tags in our previous post. There is a bugfeature in PBS such that after issuing a restart of pbs_server, CPU times are set to 00:00:00 and no longer increment. In order to work around this we have written a nqstat script which synthetically computes the cpu time and displays it in the same format as qstat.
Caveats: currently nqstat does not accept any arguments and only displays data in the format of the original qstat command in a single list. The original qstat command can still be called with -f, -t arguments etc and the data displayed is stil correct, except for the missing cpu time in the case of a recent pbs_server restart.
Q: should I use qstat or nqstat?
A: if you run qstat and you see that your jobs show zero cpu hours then run nqstat.
Q: what if I want to see detailed information on my job?
A: run qstat -f.