[slurm-users] gpu state in scontrol show node

[Thomas Schäfer]

Hi,

I have some identical machines with identical configuration. But the 
state in scontrol is slightly different:

scontrol show node | grep Gres
    Gres=gpu:geforce:8(S:0-1)
    Gres=gpu:geforce:8(S:0-1)
    Gres=gpu:geforce:8


When (dis)appears (S:0-1) ?

I am new to slurm and gres/gpu.


gres.conf

AutoDetect=nvml
Name=gpu Type=geforce File=/dev/nvidia[0] Links=-1,0,0,0,0,0,0,0 Cores=0-19
Name=gpu Type=geforce File=/dev/nvidia[1] Links=0,-1,0,0,0,0,0,0 Cores=0-19
Name=gpu Type=geforce File=/dev/nvidia[2] Links=0,0,-1,0,0,0,0,0 Cores=0-19
Name=gpu Type=geforce File=/dev/nvidia[3] Links=0,0,0,-1,0,0,0,0 Cores=0-19
Name=gpu Type=geforce File=/dev/nvidia[4] Links=0,0,0,0,-1,0,0,0 Cores=20-39
Name=gpu Type=geforce File=/dev/nvidia[5] Links=0,0,0,0,0,-1,0,0 Cores=20-39
Name=gpu Type=geforce File=/dev/nvidia[6] Links=0,0,0,0,0,0,-1,0 Cores=20-39
Name=gpu Type=geforce File=/dev/nvidia[7] Links=0,0,0,0,0,0,0,-1 Cores=20-39

slurm.conf

NodeName=w1 CPUs=40 Boards=1 SocketsPerBoard=2 CoresPerSocket=10 
ThreadsPerCore=2 RealMemory=256000 Gres=gpu:geforce:8 State=UNKNOWN
NodeName=w2 CPUs=40 Boards=1 SocketsPerBoard=2 CoresPerSocket=10 
ThreadsPerCore=2 RealMemory=128000 Gres=gpu:geforce:8 State=UNKNOWN
NodeName=w3 CPUs=40 Boards=1 SocketsPerBoard=2 CoresPerSocket=10 
ThreadsPerCore=2 RealMemory=128000 Gres=gpu:geforce:8 State=UNKNOWN

GresTypes=gpu


Thank you,

Thomas


-- 

There’s no place like ::1

Thomas Schäfer (Systemverwaltung)
Ludwig-Maximilians-Universität
Centrum für Informations- und Sprachverarbeitung
Oettingenstraße 67 Raum C109
80538 München ☎ +49/89/2180-9706  ℻ +49/89/2180-9701