Domanda

I'm trying to distribute process in a hpc with 8 cores per node, I also have a partition with 2 nodes.

I have done this csh test script:

#!/bin/tcsh
foreach i (`seq 30`)
    srun csh -c "echo 'running${i} into:'; hostname; sleep 10;echo 'end ${i}'" &
end
echo "waiting for jobs completion"
wait

And I want to run it with salloc:

salloc -N2 -p mypartition testsalloc.sh

The script launch the 30 processes simultaneously when I expected that 16 were launched and 14 were queued.

Is this behaviour posible to do with salloc and srun?

È stato utile?

Soluzione

You could have used sbatch rather than salloc and your original script would have most probably given the expected result

Altri suggerimenti

Finally I have found a solution only adding this params to srun command:

srun --ntasks=1 --exclusive ....
Autorizzato sotto: CC-BY-SA insieme a attribuzione
Non affiliato a StackOverflow
scroll top