Re: [HTCondor-users] setting up dedicated pool for parallel universe

[Greg Thain <gthain@xxxxxxxxxxx>]

On 12/28/18 10:21 AM, Kodanda Ram Mangipudi wrote:
> Dear Ivo,
>
> Thanks for writing. I started with simple condor and vanilla univ. I 
> have used the attached job submission script and the shell script 
> calling an MPI program vasp.
> The job runs but uses only one slot. What is the right way to run this?


I think you are on the right track. If you want your MPI job to run on 
multiple cores all on the same machine, you want it to run in a single 
slot that has multiple cores. To do this, you'll want to add a

Request_cpus = 12

to your submit file if you want 12 cores, and configure your start to 
allow multiple cpu slots. To do this with partitionable slots, add the 
following to you worker node config:

NUM_SLOTS = 1
NUM_SLOTS_TYPE_1 = 1
SLOT_TYPE_1 = cpus=100%
SLOT_TYPE_1_PARTITIONABLE = true


And then your condor job will be provisioned with 12 cores.


-greg