Elasticluster v1.3.dev13 - centos 7 - slurm 18.08.5 - but have 18.08.8 installed on cluster

19 views
Skip to first unread message

Maiken Pedersen

unread,
May 7, 2019, 1:00:43 PM5/7/19
to elasticluster
Working on resizing my cluster, where I have slurm 18.08.8 version installed.

The task "common:Upgrade all installed packages to latest version" fails with
fatal: [frontend001]: FAILED! => {"changed": true, "msg": "Error: Package: slurm-contribs-18.08.5-2.el7.x86_64 (verdurin-slurm)\n           Requires: slurm(x86-64) = 18.08.5-2.el7\n           Installed: slurm-18.08.7-1.el7.x86_64 (installed)\n               slurm(x86-64) = 18.08.7-1.el7\n           Available: slurm-18.08.5-2.el7.x86_64 (verdurin-slurm)\n               slurm(x86-64) = 18.08.5-2.el7\n", "obsoletes": {"slurm-sjobexit": {"dist": "x86_64", "repo": "@verdurin-slurm", "version": "15.08.13-1.el7.centos"}, "slurm-sjstat": {"dist": "x86_64", "repo": "@verdurin-slurm", "version": "15.08.13-1.el7.centos"}}, "rc": 1, "results": ["Loaded plugins: fastestmirror\nLoading mirror speeds from cached hostfile\n * base: ftp.uninett.no\n * elrepo: ftp.nluug.nl\n * epel: fedora.uib.no\n * extras: ftp.uninett.no\n * updates: ftp.uninett.no\nResolving Dependencies\n--> Running transaction check\n---> Package slurm-contribs.x86_64 0:18.08.5-2.el7 will be obsoleting\n--> Processing Dependency: slurm(x86-64) = 18.08.5-2.el7 for package: slurm-contribs-18.08.5-2.el7.x86_64\n---> Package slurm-sjobexit.x86_64 0:15.08.13-1.el7.centos will be obsoleted\n---> Package slurm-sjstat.x86_64 0:15.08.13-1.el7.centos will be obsoleted\n--> Finished Dependency Resolution\n You could try using --skip-broken to work around the problem\n You could try running: rpm -Va --nofiles --nodigest\n"]}

Should I downgrade my slurm version on the cluster to make this work? Or is there a chance that this should work with 18.08.7?

Thanks!
Maiken

Riccardo Murri

unread,
May 7, 2019, 3:15:31 PM5/7/19
to Maiken Pedersen, elasticluster
Hello Maiken,

> The task "common:Upgrade all installed packages to latest version" fails with
> [...]
> Should I downgrade my slurm version on the cluster to make this work? Or is there a chance that this should work with 18.08.7?

My reading of the error message is this:

- you have SLURM 18.08.7 installed from some RPM package which is *not
available from a YUM repository*
- ElastiCluster runs `yum install slurm` so YUM only finds the
`slurm-*` packages coming from @verdurin's repository (which ATM only
contains SLURM 18.08..5)
- so YUM tries to downgrade SLURM packages to 18.08.5
- *but* some of the packages you have installed right now
(`slurm-sjstat`, `slurm-contribs`, `slurm-sjobexit`) are
"obsolete"[1], so they would be removed -- which seems not to be
allowed without additional options to YUM.[2]

I do not know YUM (or CentOS) for the matter to know how exactly to
cope with this situation. Can you try setting `obsoletes=0` in
`/etc/yum.conf` (as explained in [2]) on all nodes and see if this
fixes the issue? You will end up with SLURM 18.08.5 on the cluster,
though.

[1]: https://serverfault.com/questions/814040/install-obsoleting-package-without-removing-obsoleted-package
[2]: https://stackoverflow.com/a/48041045/459543

Ciao,
R

Maiken Pedersen

unread,
May 20, 2019, 7:31:06 AM5/20/19
to elasticluster
Issue was surpassed by removing slurm-sjstat, slurm-sjobexit, slurm-contribs on the frontend and adding ignore_errors on the task.
A bit of a hack, but well, no need to spend more time on it :)
Reply all
Reply to author
Forward
0 new messages