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

Re: [Condor-users] Limiting jobs per machine

On Aug 9, 2010, at 12:46 PM, Tomer Altman wrote:

> I am a new Condor cluster admin. Our group is interested in learning how to limit the number of Condor jobs per machine.
> Our condor machines are set up such that for any one physical machine ("Schnapps"), we have each core from the physical machine show up as a separate processor that Condor can allocate jobs to:
> vm1@Schnapps. LINUX       X86_64 Unclaimed  Idle       0.000  4019  1+07:20:50
> vm2@Schnapps. LINUX       X86_64 Unclaimed  Idle       0.000  4019 27+22:51:05
> vm3@Schnapps. LINUX       X86_64 Unclaimed  Idle       0.000  4019  0+02:50:06
> vm4@Schnapps. LINUX       X86_64 Unclaimed  Idle       0.000  4019  1+07:25:53
> For small jobs, we benefit from the fine-grained scheduling. For some memory-intensive jobs, we don't want four instantiations of the job being allocated to one physical machine. I've read through the Condor manual, but I can't seem to figure out how to create a job description file that will limit the jobs of the batch to one job per physical machine. Any insight would be greatly appreciated.

Checkout out this recipe: https://condor-wiki.cs.wisc.edu/index.cgi/wiki?p=WholeMachineSlots
It allows a machine to accept one large-memory job or several small-memory jobs.

|           Jaime Frey           | I used to be a heavy gambler.     |
|       jfrey@xxxxxxxxxxx        | But now I just make mental bets.  |
|                                | That's how I lost my mind.        |