FWIW we use a placeholder node with state set to future to allow submitting jobs to empty partitions:
slurm.conf ... # A placeholder nodes to be included in any otherwise empty partitions to avoid # sbatch: error: Batch job submission failed: More processors requested than permitted NodeName=placeholder procs=2048 state=future PartitionName=2x4x1 Nodes=placeholder Do take note of the race-condition where a bare srun job may fail if you type srun and then update slurm.conf to have new nodes underneath it - this can be avoided with salloc srun or sbatch. On Tue, Feb 17, 2015 at 2:47 PM, Lyn Gerner <[email protected]> wrote: > Hi Taras, > > Not sure what your use case is, but you might check out node states > "CLOUD" and "FUTURE" in slurm.conf. > > Best, > Lyn > > On Tue, Feb 17, 2015 at 3:26 AM, Taras Shapovalov < > [email protected]> wrote: > >> Hi guys, >> >> Do you see any way to allow jobs submission to a queue without any nodes? >> Slurm calculates a number of all CPUs in a queue and rejects a job if it >> requests more then available, but how this constrain can be eliminated? >> >> Thanks, >> >> Taras >> > >
