Have you had a look at the HTC documentation? http://schedmd.com/slurmdocs/high_throughput.html
Without knowing what your real objective is it is hard to prescribe a real solution. >From your description it seems strange you would have the script sbatch is >calling call sbatch once again. What are you trying to accomplish there? Wouldn't it just be easier to run this script outside of an allocation? Danny On Wednesday September 14 2011 7:15:04 PM Yuri D'Elia wrote: > On Wed, 14 Sep 2011 13:03:01 -0400 > Christian Hudon <[email protected]> wrote: > > > On 09/14/11 12:41, Yuri D'Elia wrote: > > > On Wed, 14 Sep 2011 11:38:31 -0400 > > > "[email protected]" <[email protected]> wrote: > > > > > >> Upgrade to slum v2.2 or higher for more jobs or steps. The steps are > > >> running within the job's allocation that goes away when the job script ends, so adding "wait" to the end of the script would probably be your simplest solution. > > > Since "sbatch" exits immediately, I cannot do that. > > > > I use "srun foo &" when writing scripts with job steps. Then the "wait" > > at the end works (and is essential). > > I cannot run 100k instances of srun from a single node either.
