[slurm-users] problem: cannot srun with jobid and nodelist to run paraview server as additional job step on existing allocated nodes

[程迪]

Hi everyone,

In my practice, I usually reserve a lot of nodes using `salloc` command and
run tasks using `mpirun`. And now I want to do it in slurm's way. I saw in
the document that I can add job steps using `srun` using `--jobid` and
`--nodelist` option. However, I found that I cannot run paraview server as
additional job step on existing allocated nodes. Here is the commands:

```
srun -n 56 --jobid=23456 --nodelist=cn333,cn444 pvserver --mpi --mesa
--use-offscreen-rendering -sp=12345 &
```

And it will show errors saying I cannot run 56 programs listening on the
same port 12345.

It seems the 56 jobs are running separately. How can I diagnostic this
problem further and solve it?

p.s. The mpi I use is OpenMPI 1.10 and the GCC is 5.1. The HPC is tianhe-1A.

Di Cheng
Engineer of Research Center
China Academy of Aerospace Aerodynamics
Phone: +86-l58Ol5949ll

Address: No.17, YunGang West Road, Fengtai District, Beijing, China

Zip Code：100074
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.schedmd.com/pipermail/slurm-users/attachments/20171120/a688ef83/attachment.html>