It worked out.
srun -N2 --ntasks=25 --pty devtest
actually ran on 25 cores, 16 of the first and 9 on the second
Thanks alot.
On 21/01/2018 08:33, Nadav Toledo
wrote:
Sorry for delay answer.
First thank you for pointing this out I would have missed that
for sure
Second, If i run the command you wrote, i get only 2 cores, is
that how it suppose to work?
meaning :
srun --ntasks=17 --pty bash
~$ nproc
2
also running dd if=/dev/zero of=/dev/null & , 5 times only
takes 2 cores
Am i missing something?
On 18/01/2018 10:16, Loris Bennett
wrote:
Nadav Toledo <nadavtol...@cs.technion.ac.il> writes:
Nadav Toledo <nadavtol...@cs.technion.ac.il> writes:
Hey everyone,
We've just setup a slurm cluster with few nodes each has 16 cores.
Is it possible to submit a job for 17cores or more?
If not, is there a workaround?
Thanks in advance, Nadav
It should be possible. Have you tried? If so, do you get an error?
srun -c17 --pty bash
srun: error: CPU count per node can not be satisfied
srun: error: Unable to allocate resources: Requested node configuration is not available
The option -c is the short form of --cpus-per-task, which really means
"cores per process". As you only have 16 cores per node, you can't run
a single process on 17 cores.
You probably want
srun --ntasks=17 --pty bash
I'd suggest you use the long forms of the options until you get more
familiar with Slurm (although even that can still be confusing,
cf. core/cpu, task/process).
Cheers,
Loris
|
