[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Condor-users] Problem of policy with SMP cluster


I am a newbie in Condor. This is my first post in this mailing list. Please
give me some suggestions for my following questions.

We have a SMP cluster, about 40 nodes. Each node has 4 CPUs. We have
setup Condor 6.8.2 successfully to run both vallina and parallel jobs in
that cluster.

Now the problems are:

1. Could I run the parallel job which is prefered to occupy only 2 CPUs for
   each node? For example, now I want to run a job with 8 CPUs, I hope that
   the following rules could be applied:       are available.

2. I want to limit the total number of CPUs for each user to run his jobs.
   I mean, the sum of the CPUs of all of his running jobs. For example, if
   I set the CPU limit to be 40, and current he has 3 jobs, with 10, 8, 6 CPUs
   occupied respectively. Now he want to submit another new job, and the job
   will run only if it require lesser than 16 CPUs, otherwise this job will
   be queued and wait. How could I do that?

Thank you very much for your kindly help :)


   (a) If possible, run the job in 4 nodes, with each node only occupy 2 CPUs.
   (b) If (a) failed, try to run the job in 2 nodes, with each node only
       occupy 4 CPUs.
   (c) If (b) failed too, try to run the job freely, as long as totally 8 CPUs