Re: [HTCondor-users] Use CPUs on GPU nodes

["Beyer, Christoph" <christoph.beyer@xxxxxxx>]

Hi,

the '+GPUJob = true' part is rather a local config and not default condor I guess ?

You can always define different slots with individual attributes on the gpu machines and run cpu jobs at the same time a gpu job is running:

#####################################################
## slot definition
### ##################################################
# # define GPU slots here
@use feature : GPUs
SLOT_TYPE_1 = GPUs=1, CPUs=2
SLOT_TYPE_1_PARTITIONABLE = FALSE
NUM_SLOTS_TYPE_1 = 1
START = (NODE_IS_HEALTHY =?= True) && (StartJobs =?= True) && TARGET.RequestGpus
## define an additional partitionable slot with the rest of the cpus

SLOT_TYPE_2 = CPUs=10

NUM_SLOTS_TYPE_2 = 1

SLOT_TYPE_2_PARTITIONABLE = True
#######################################################
#

If you want to make the usage of the non-gpu slot depending on the usage of the gpu slot maybe concurrency_limits are a way to go, see:

https://htcondor.readthedocs.io/en/stable/admin-manual/setting-up-special-environments.html?highlight=concurrency#concurrency-limits

Best

christoph

--
Christoph Beyer
DESY Hamburg
IT-Department

Notkestr. 85
Building 02b, Room 009
22607 Hamburg

phone:+49-(0)40-8998-2317
mail: christoph.beyer@xxxxxxx

---

Von: "Bo-Ru Roy Lu" <roylu@xxxxxx>
An: "htcondor-users" <htcondor-users@xxxxxxxxxxx>
Gesendet: Samstag, 9. Mai 2020 16:06:42
Betreff: [HTCondor-users] Use CPUs on GPU nodes

Hi,

In order to fully use our CPUs on our server, I'd like to use CPU nodes on our GPU machines when people are not using GPUs. 

I found that if I want to submit a CPU job to a GPU node, I have to set 2 flags, "request_gpus = 1" and "+GPUJob = true" in job description file. If I don't set the flags, the job will be idle forever. However, I don't want to let condor mark I am using GPU because people will be unable to use GPU machines.

Do you have a good solution for this?

Best,

Roy

_______________________________________________
HTCondor-users mailing list
To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
subject: Unsubscribe
You can also unsubscribe by visiting
https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users

The archives can be found at:
https://lists.cs.wisc.edu/archive/htcondor-users/