SLURM 18.08 now default on RHEL/CentOS 7 clusters

14 views
Skip to first unread message

Riccardo Murri

unread,
Apr 15, 2019, 3:54:54 PM4/15/19
to elasticluster
Dear all,

Thanks to @verdurin and his work on providing COPR repositories with
SLURM packages, SLURM 18.08 is now the default version of SLURM
installed by ElastiCluster on RHEL/CentOS 7.
Older versions are selectable with the `slurm_version` global variable.

Clusters based on Debian or Ubuntu continue to install the version of
SLURM provided by OS packages.

As usual, please let me know of any issue with this (or other) updates.

Ciao,
R

Maiken Pedersen

unread,
May 7, 2019, 8:12:35 AM5/7/19
to elasticluster
Hi,

I am using elasticluster 1.3.dev0 - and at the moment want to keep to that version since the resize works here with paramiko 2.1.2.

However, I have updated my cluster to slurm 18.08.

Can I copy over the changes wrt slurm from 1.3.dev13 to my 1.3.dev0  version?
If so, what files need updating?

Thanks!
Maiken

Riccardo Murri

unread,
May 7, 2019, 10:10:21 AM5/7/19
to Maiken Pedersen, elasticluster
Hello Maiken,

> However, I have updated my cluster to slurm 18.08.
> Can I copy over the changes wrt slurm from 1.3.dev13 to my 1.3.dev0 version?

In principle, you would only need to apply the patch in this commit:
https://github.com/gc3-uzh-ch/elasticluster/commit/20eca47

However, it's totally possible that some changes happened between
1.3.dev0 and 1.3.dev13 that the commit depends on implicitly.

Ciao,
R

Maiken Pedersen

unread,
May 7, 2019, 10:21:14 AM5/7/19
to Riccardo Murri, elasticluster
Hi,
Thanks, yes, that was what I was “worried” about.

I have applied the patch, but have not yet managed to setup the cluster successfully. Getting issues at the slurm step. Need to do some more testing to figure it out.
Might get back to you if I need some help :)

Thanks!
Maiken

Riccardo Murri

unread,
May 7, 2019, 10:29:02 AM5/7/19
to Maiken Pedersen, elasticluster
Hello Maiken,

> I have applied the patch, but have not yet managed to setup the cluster successfully. Getting issues at the slurm step. Need to do some more testing to figure it out.
> Might get back to you if I need some help :)

I am not sure, why can't you just install the current ElastiCluster in
a new virtualenv?
You can still keep the old one around, to be sure. But there should be no need
to try backporting random commits...

Ciao,
R

Maiken Pedersen

unread,
May 7, 2019, 10:31:48 AM5/7/19
to elasticluster
This is because I am getting the paramiko problems with the newer version of elasticluster. But I should give it another try :)

Thanks!

Riccardo Murri

unread,
May 7, 2019, 10:38:29 AM5/7/19
to Maiken Pedersen, elasticluster
Hello Maiken,

> This is because I am getting the paramiko problems with the newer version of elasticluster. But I should give it another try :)

I doubt that error is due to the newer ElastiCluster; sounds like a
library conflict due to (possibly) old system libraries that cannot be
updated. What OS are you using for running the `elasticluster`
program on?

Ciao,
R

Maiken Pedersen

unread,
May 9, 2019, 5:20:39 AM5/9/19
to elasticluster
Hi,
sorry I thought I had answered this.

I decided to uninstall the obsolete slurm packages in order not to get that yum complaint at least, and have also just added ignore_errors: yes in order to keep 18.08.

About paramiko issues: I finally got the right combination working with the latest elasticluster. Thanks
Maiken
Reply all
Reply to author
Forward
0 new messages