Hi, try adding "-N 10" to explicitly ask for ten nodes too.
If you have access to the slurm.conf and don't have to share the cluster, or share it with people with the same needs, you might like SelectTypeParameters=CR_LLN or LLN on a per partition basis. (s. http://slurm.schedmd.com/slurm.conf) In general it should be better to ask for that on a per job level -- as you try to do -- otherwise you could block a job that needs any resource from a whole node (please, ignore my chatter if you already know that). Regards, Benjamin On 11/02/2016 15:32, CB wrote: > Hi, > > I'm running Slurm 15.08.12. > > I've tried the following sbatch command but it doesn't seem to work. > > sbatch --array=1-10 --ntasks-per-node=1 -o env.log-%A.%a > --wrap="/usr/bin/env" > > I'd like to distribute each task on a different node but it looks like > it is using fill up the node before moving on to the next node. > > $ grep SLURMD_NODENAME env.log-74980.* > env.log-74980.1:SLURMD_NODENAME=an-16 > env.log-74980.10:SLURMD_NODENAME=an-39 > env.log-74980.2:SLURMD_NODENAME=an-16 > env.log-74980.3:SLURMD_NODENAME=an-16 > env.log-74980.4:SLURMD_NODENAME=an-16 > env.log-74980.5:SLURMD_NODENAME=an-16 > env.log-74980.6:SLURMD_NODENAME=an-16 > env.log-74980.7:SLURMD_NODENAME=an-16 > env.log-74980.8:SLURMD_NODENAME=an-39 > env.log-74980.9:SLURMD_NODENAME=an-39 > > Any suggestions? > > Thanks, > - Chansup > -- FSU Jena | JULIELab.de/Staff/Benjamin+Redling.html vox: +49 3641 9 44323 | fax: +49 3641 9 44321
