Hello Michael, 


I can not get access to CUDA_MPS_PIPE_DIRECTORY



o_digeroni@sx138:1194145 $ nvidia-cuda-mps-control -d
Warning: Failed writing log files to directory [/var/log/nvidia-mps]. No logs will be available.
Cannot access CUDA_MPS_PIPE_DIRECTORY=/run/mps/control

Even thougt I do: export CUDA_MPS_PIPE_DIRECTORY=/run/mps/


Thanks, Br

From: Fladischer, Michael
Sent: 31 July 2023 12:46:56
To: Di Geronimo Quintero, Bruno; Trajanoski, Slave; medbionode@lists.medunigraz.at
Subject: AW: slurm configuration issue solved
 
nvidia-smi is now again available in the global $PATH.
________________________________________
Von: Di Geronimo Quintero, Bruno <bruno.digeronimo@medunigraz.at>
Gesendet: Montag, 31. Juli 2023 12:35
An: Trajanoski, Slave; medbionode@lists.medunigraz.at
Betreff: [Medbionode] Re: slurm configuration issue solved

Hello Slave,


Thanks for the update. I guess that GPUs from node sx138 are not yet available. If I log to the node I can not monitor them via nvidia-smi neither run jobs:


o_digeroni@sx138:2233 $ nvidia-smi
-bash: nvidia-smi: command not found


sx138
2023-07-31T12:32:38 CEST
cudaGetDeviceCount failed unknown error


I do not know if someone else is facing same issue.


Thanks in advances for the efforts.


Br

________________________________
From: Trajanoski, Slave <slave.trajanoski@medunigraz.at>
Sent: 31 July 2023 11:31:39
To: medbionode@lists.medunigraz.at
Subject: [Medbionode] slurm configuration issue solved

Dear MedBioNode users,

I have good news for you. IT department managed to solve the slurm configuration issue and sbatch is working normally again.

Regards,
Slave

--
Slave Trajanoski, Phd
Senior Scientist Bioinformatics
CF Computational Bioanalytics, Center for medical research
Medical University Graz
Neue Stiftingtalstraße 6-West, Turm/Stiege:P, 4.Stock
8010 Graz
Tel. +43 316 385 73024
E-Mail: slave.trajanoski@medunigraz.at<mailto:slave.trajanoski@medunigraz.at>