Error assigning CPUs to worker

Install, Configure and Update
1

Hello,

we are having issues with assigning CPUs to a worker. We have 96 CPUs and 8 GPUs, and we want to assign 24 CPUs and 4 GPUs to one worker. We found the following error:

cryosparcw connect --master u035809.lgp.ehu.es -n -worker bremlive-worker --port 39000 --ssdpath /home/cryosparcuser/scratch/cryosparc_cache --newlane --lane bremlive --cpus 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24 --gpus 0,1,2,3

usage: connect.py [-h] [–worker WORKER] [–master MASTER] [–port PORT]
[–sshstr SSHSTR] [–update] [–nogpu] [–gpus GPUS]
[–nossd] [–ssdpath SSDPATH] [–ssdquota SSDQUOTA]
[–ssdreserve SSDRESERVE] [–lane LANE] [–newlane]
[–rams RAMS] [–cpus CPUS] [–monitor_port MONITOR_PORT]
connect.py: error: argument --cpus: invalid int value: ‘0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24’

Without the CPU part, everything works well. We can create a worker with 4 GPUs, but we cannot modify the CPUs or the RAM. Are we doing something wrong?

Thanks for your help!
Jorge

2

Hi Jorge,

I’m not sure if it’s the intended behaviour, but having added the worker node, I believe you can modify the --cpus and --rams parameters using the --update flag. For example, --cpus 24 and --rams 8 will update it to use 24 CPU slots and 64GB worth of memory.

cryosparcw connect --master u035809.lgp.ehu.es --worker bremlive-worker --port 39000 --cpus 24 --rams 8 --update

Cheers,
Yang

3

Yes. The --cpu parameter should be specified with a single integer. The guide has been updated accordingly.

4

Hi @wtempel,

I think the underlying issue may be that the --cpus and --rams parameters seem to be ignored when adding a new worker node, either to a new lane or an existing lane. They do take subsequently when applied with the --update flag. --gpus is respected in both scenarios. It seems like a bug to me.

Cheers,
Yang

5

Confirmed. We have taken note.
Thank you for reporting the problem @jplopez @leetleyang .