[HTCondor-users] running MPI job on dedicated SLOT_TYPE

[Alessandro Italiano <alessandro.italiano@xxxxxxxxxx>]

Hi,

we have servers [belonging to a HTCondor cluster] with both GPUs and CPUs available.
I would like to dedicate those CPUs to mpi job through a DedicatedScheduler serving parallel universe jobs.
While I would like to let the GPU get allocated to Vanilla jobs.

Is there a way to define on this kind of server two slots, SLOT_TYPE_1 with the CPUs and SLOT_TYPE_2 with all the GPUS
and let the dedicatedScheduler to use only the SLOT_TYPE_1 ?

thanks in advance

Ale