Hello everybody,

we have a shared memory cluster, which has 512 cores (64 sockets * 8 cores) and 
we had set the following in the Slurm configuration:
NodeName=venus Sockets=64 CoresPerSocket=8 ThreadsPerCore=1 State=UNKNOWN 
CPUSpecList=0-23 RealMemory=8000000

When I create a reservation over 368 cores, I only can use 344 cores within 
that reservation.
That’s the reason why I guess, the CPUSpecList causes the problem.
Also, I removed the CPUSpecList and everything works fine!
We currently run Slurm 16.05.2, but I think, we still had the problem in older 
versions too.

Could you please tell me, what I’m doing wrong?

Kind regards,
Danny

Attachment: smime.p7s
Description: S/MIME cryptographic signature

Reply via email to