--mem is a per-node option, not per job. So you'd need to keep it at 120000 in your srun command, otherwise you're requesting 240GB on each node which SLURM is telling you is not available.
(See http://schedmd.com/slurmdocs/srun.html) - Tim On 10/09/2012 03:22 AM, Sefa Arslan wrote: > > I have a such a problem, each nodes I send my jobs has 48 cores and > 128GB memory on them. > > I can run the following job: > > srun -N1 -n48 --mem=120000 -p cuda01 hostname > > > also I can run the same job on every node adding --nodelist parameter to > the above line. > > But when I double the size of the job I get an error: > > srun -n96 -N2 --mem=240000 -p cuda01 hostname > > srun: error: Unable to allocate resources: Memory required by task is not > available > > > I paste the configurations below. What may be the problem about them? > > Thanks in advance. -- Tim Wickberg [email protected] Senior System Administrator, Office of Research Rensselaer Polytechnic Institute
