Hi: 

Thanks for your reply.

It is not clear to me how the slurm.conf file should be set:

I changed the node04 line to

> NodeName=node04 Feature="gtx580","gtx580","tesla","gtx690","gtx690" Gres=gpu:5
and still I cannot submit the job.

Thanks

Marco

On Jan 27, 2013, at 6:53 PM, "[email protected]" <[email protected]> wrote:

> your feature syntax for node 04 is bad. see "man slurm.conf"
> --
> Sent from my Android phone. Please excuse my brevity and typos.
> 
> Marco A Escobar <[email protected]> wrote:
> 
> Hi All:
> 
> I am new to slurm:
> 
> I am trying to configure it to run in a kind of  "heterogenous" gpu cluster.
> 
> I mean all the nodes have the same number and types of gpus, except one, 
> I am attaching the relevant parts of the slurm.conf
> ____________________________
> # Nodes
> NodeName=node[01-03] Feature="gtx590" Gres=gpu:6
> NodeName=node04 Feature="gtx580:2"&"tesla:1"&"gtx690:2" Gres=gpu:5
> NodeName=node[05-07] Feature="gtx690" Gres=gpu:8
> # Partitions
> PartitionName=defq Nodes=node[01-07] Default=YES MinNodes=1 
> MaxNodes=UNLIMITED MaxTime=UNLIMITED AllowGroups=ALL Priority=1 
> DisableRootJobs=NO RootOnly=NO Hidden=NO Shared=NO
> # Generic resources types
> GresTypes=gpu
> __________________________
> 
> As is evident node04 is the problem, I cannot run the following script
> 
> #!/bin/bash
> #SBATCH --job-name="FastMag_Test_1"
> #SBATCH --nodes=1
> #SBATCH --ntasks=1
> #SBATCH --cpus-per-task=1
> #SBATCH --ntasks-per-node=1
> #SBATCH --gres=gpu:1
> #SBATCH --constraint=tesla
> #SBATCH --mail-type=end
> #SBATCH [email protected]
> echo "The Job was sent to node:"
> echo $SLURM_JOB_NODELIST
> srun FastMagJan_2013.run
> 
> 
> It runs when I use gtx590 or gtx690, but not when I say tesla.
> 
> the output is the following:
> 
> [marco@cobalt example1]$ sbatch ~/runscript.sh
> sbatch: error: Batch job submission failed: Invalid feature specification
> 
> 
> Thanks for your help
> 
> Marco
> 

Reply via email to