[slurm-users] Upgrading a slurm on a cluster, 17.02 --> 18.08

[Baker D.J.]

Hello,


I wondered if I could compare notes with other community members who have upgraded slurm on their cluster. We are currently running slurm v17.02 and I understand that the rpm mix/structure changed at v17.11. We are, by the way, planning to upgrade to v18.08.


I gather that I should upgrade slurmdbd first and ideally the slurmctld should not be touched/killed at point. On our cluster we run both the slurmdbd and the slurmctld on a single host. I have been advised to just upgrade slurmdbd initially -- upgrading the rpms in a piecemeal fashion. That is, just installing the slurm and slurmdbd rpms in the first instance. Due to the rpm structure changes at v17.11 that strategy doesn't work -- the yum updates fail with dependency issues.


I guess that the only solution is to upgrade all the slurm at once. That means that the slurmctld will be killed (unless it has been stopped first). Is there anyone who has done an upgrade who would be willing to share their experiences, please? In other words, is it valid to kill both the slurmdbd and slurmcltd processes at the start of the upgrade and, if so, how does the loss of the slurmctld affect the cluster re running jobs, user activity, etc?


Best regards,

David
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.schedmd.com/pipermail/slurm-users/attachments/20180925/b5cfb2ff/attachment.html>