[Beowulf] Random PBS Question
Glen Beane
glen.beane at jax.org
Tue Nov 27 16:12:21 PST 2007
--- Original Message ---
> For those reading the list that run PBS, we have an interesting
> situation... After a fairly substantial crash, we've had to
> rebuild our
> PBS configuration. We've also been asked to reset to the PBS
> job id
> counter to a position near where it left off [for reasons of
> accounting]. Has anyone ever done this??
>
> cheers
> john
PBS Pro, Open PBS, or TORQUE?
One easy way, but a bit of a hack, is to write a script that submits a bunch of dummy jobs and qdel's them.
With Open PBS and older versions of torque you can edit the code that creates the server database to initialize the job counter to whatever number you want, then recompile the server, and start it with the -t create flag.
Newer versions of TORQUE include a qmgr hidden server attribute (documented, but not printed if you run "print server") called next_job_number. qmgr -c "s s next_job_number=N"
With OpenPBS or TORQUE you can look at the source and figure out the format of the binary server database file and change the job id counter with a hex editor. You might be able to reverse engineer the location of this in the server database for PBSPro by submitting a job and then looking to see what got incremented in the file - it is probably in the native byte order of your system.
Glen L. Beane
The Jackson Laboratory
Software Engineer III
Phone 207-288-6153
More information about the Beowulf
mailing list