Hi, I read the problem differently. Might also want to look at heterogeneous jobs. Slurm Workload Manager - Heterogeneous Job Support (schedmd.com) <https://slurm.schedmd.com/heterogeneous_jobs.html>
Doug On Tue, Feb 28, 2023 at 3:27 PM Jake Jellinek <jakejelli...@outlook.com> wrote: > Hi Brian > > Thanks for your response > > > I am guessing you are using srun to get an interactive session on a > node. That approach is being deprecated and you get a shell by default with > salloc > This is exactly what I'm trying to do .... I didn’t know about the salloc > thing > > Let me do some more testing and I'll see if you've just resolved my issue. > > Will be in touch very soon > Jake > > -----Original Message----- > From: slurm-users <slurm-users-boun...@lists.schedmd.com> On Behalf Of > Brian Andrus > Sent: 28 February 2023 18:47 > To: slurm-users@lists.schedmd.com > Subject: Re: [slurm-users] Chaining srun commands > > Jake, > > It may help more to understand what you are trying to do accomplish rather > than find out how to do it the way you expect. > > I am guessing you are using srun to get an interactive session on a node. > That approach is being deprecated and you get a shell by default with salloc > > If you are trying to start new jobs on other nodes, you would want to use > salloc/sbatch to launch them. > If you are wanting to have multiple nodes on a single job, IIRC, you would > request them with the initial salloc and then use options to srun to launch > appropriately. > > What specifically do you want to get (resource-wise) and how do you want > to use them? > > Brian Andrus > > On 2/28/2023 9:49 AM, Jake Jellinek wrote: > > Hi all > > > > I come from a SGE/UGE background and am used to the convention that I > can qrsh to a node and, from there, start a new qrsh to a different node > with different parameters. > > I've tried this with Slurm and found that this doesn’t work the same. > > > > For example, if I issue an 'srun' command, I get a new node. > > However if I then try to start a new srun session to a different node > type (different resource requirements), it just puts me back on the same > box. > > > > I did find a post from 12 years ago that suggested that this was by > design but am hoping that this has now changed or that there is a config > option which turns off this feature. > > > > Thank you > > Jake > >
