Re: SLURM -- Upgrade from stretch to bullseye
Hi Julien,
On Wed, Aug 25, 2021 at 12:27:50PM +0200, Julien Tailleur wrote:
> I noticed that, since stretch, the name of the packages have changed and I
> wanted to clean up my system. For instance, running ls -d /etc/slur* returns
>
> /etc/slurm /etc/slurm-llnl
>
> but the package slurm-llnl is not there anymore. Should I remove this ?
> (Shouldn't it have been removed by the dist-upgrade?).
you can remove the /etc/slurm-llnl if you don't include files under it
inside the slurm.conf file that should have been copied under
/etc/slurm. That's basically the reason why it is not automatically
removed by the upgrade process.
> I am also a bit
> confused about what packages should be installed on the master and computing
> nodes. I tried running the examples from
>
> https://slurm.schedmd.com/quickstart.html
>
> (The one with my.script). But it fails with the following outcome:
>
> cat my.stdout
> FX31
> /tmp/slurmd/job06775/slurm_script: 4: srun: not found
> /tmp/slurmd/job06775/slurm_script: 5: srun: not found
>
> Indeed, srun is not available on the computing nodes. Is this the desired
> installation ? Running dpkg -l | grep slurm on master nodes and computing
> nodes yields:
>
> Master:
> ii slurm-client 20.11.7+really20.11.4-2 amd64
> SLURM client side commands
> ii slurm-wlm-basic-plugins 20.11.7+really20.11.4-2 amd64
> SLURM basic plugins
> ii slurm-wlm-doc 20.11.7+really20.11.4-2 all
> SLURM documentation
> ii slurmctld 20.11.7+really20.11.4-2 amd64
> SLURM central management daemon
> ii slurmdbd 20.11.7+really20.11.4-2 amd64
> Secure enterprise-wide interface to a database for SLURM
>
> Computing nodes:
> ii slurm-wlm-basic-plugins 20.11.7+really20.11.4-2 amd64
> SLURM basic plugins
> ii slurmd 20.11.7+really20.11.4-2 amd64
> SLURM compute node daemon
In this particular case you need to install the slurm-client package
also on the computing nodes.
Best regards,
--
Gennaro Oliva
Reply to: