Condor FAQ

1. Overview.
   • How do i get an account?

2. Using the Condor Pool.
   • How do I submit jobs

Overview.

Condor is a distributed computing environment that allows users to utilise spare computing cycles on commodity machines to run jobs. Within Informatics we have two polls running, one based in the Forum and another based in the Appleton tower. The pools are configured to flock which allows jobs to be passed between pools when there are free nodes in the other pool. Unfortuntely the condor_ commands do not show information about the other pool. Please note that condor is not run as a guaranteed service and it may be withdrawn at little or no notice if it impacts on the nodes primary role.

How do I get an account or use condor?

Condor uses standard accounts however you have to be authorised to submit jobs from a machine. Generally the easiest way to do this is to volunteer your desktop to be in one of the pools and you will automatically be able to submit jobs. Undergraduates and MSc students who need to use condor should ask their project supervisor or the academic running the course to contact support about getting access.

Using the condor pool

How do I submit jobs

You should read the condor user manual section on submitting jobs, the manual is pretty complete on this however computers in the labs running condor cannot be used to submit jobs, staff/research students can have condor enabled on their own desktops.

Issues running condor with AFS

Because we use AFS homedirectories and condor is not AFS aware it's not possible to submit jobs from your AFS homedir, you need to either use condors file transfer mode and for complex scripts package your jobs up in the submit directory so that they can be unpackaged and run on the targer directory on the remote machine. Alternatively it ought to be possible to use some of the techniques described in the AFS long running jobs top tips file. However great care should be taken to ensure that keytab material is not left on the remote node after the job has finished running. Alternatively if you have space on an NFS filesystem you can run your jobs from there.