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

Re: [Condor-users] GPU machines is "mixed" environment (conditional preemption)

Hi Greg

On Thursday 07 October 2010 16:48:11 Greg Thain wrote:
>   On 10/07/2010 09:19 AM, Carsten Aulbert wrote:
> > However, now the big question how to address preemption. Essentially, I
> > want to ensure that the machine works as a standard compute nodes with
> > multiple cores (identically setup as others; sans vanilla universe jobs)
> > in the absence of any idle jobs which have "NeedGpu" set.
> I'm not sure I entirely understand your requirements, but I think you
> can simply
> use the Machine RANK to prefer jobs with NeedGpu set.

Hmm, I need to think about this then.

To clarify:

Imagine a machine with 4 slots and 2 GPUs ("attached to slot1 and slot2 via 
startd_attributes). No jobs with NeedGpu on the cluster, but 4 normal 
(standard) universe are currently running.

Now a user submits three jobs stating NeedGpu in her submit file. Now I want 
two of the jobs to "replace" two jobs currently running (i.e. slot1&2).

Does this make it clearer?