Condor: Batch/Grid Computing

Batch-style processing is available on select nodes of the Unix cluster via Condor.

Useful commands for submitting jobs are

• condor_run - for quick & easy job submission
• condor_submit - for full control

Useful commands for job and machine status are

• condor_status shows active machines and queues
• condor_status -submitters shows job submitter summary
• condor_q shows jobs in the local job queue
• condor_q -global shows jobs in the global job queue

For a complete introduction to Condor, please read the Condor Manual:

• http://www.cs.wisc.edu/condor/manual/

Nodes are tagged with a CondorGroup job ClassAd, which allows you to submit jobs to different groups of machines based on where you submit them from. The default behavior means that when you submit a condor job from a physlin workstation or spartha, the job will run on x86 Linux nodes. It is possible to override this behavior in your submit file by manipulating the CondorGroup job ClassAd. You should read the manual page for condor_submit for more details.

Available CondorGroups:

• osf - Tru64 Alpha cluster
• physlin - Physlin machines (general-purpose linux workstations)
• sunfire - high-speed Sunfire linux cluster (8 queues)
• tpilin - FTPI linux workstations (FTPI use only!)
• twins - CLEO compute farm (CLEO use only!)
• minos - MINOS workstations (MINOS use only!)
• cms - CMS servers (CMS use only!)
• astro - Astronomy workstations (Astronomy use only!)

It is possible to run jobs on any CondorGroup, but if you are unsure whether you should be using a group of machines please ask first! Anyone may submit jobs to physlin or sunfire machines, so submitting jobs from a physlin machine (or spartha) are safe bets.

We currently support the vanilla, standard, and MPI universes.