Similar to you we host our slurmdbd and slurmctld on the same host. So
when we upgrade we follow the following steps:
1. Stop slurmctld
2. Stop slurmdbd
3. Take backups of the slurm database and slurm spool directory
4. Update rpms (make sure to neutered the autorestart function in the
rpm spec as that can restart things before you are ready)
5. Run slurmdbd -D in a tmux (as the upgrade can take a while and if you
use systemd it will terminate early).
6. Once slurmdbd ugprade is complete stop the command line version of
slurmdbd and restart it using systemd
7. Global restart of slurmctld and slurmd for the cluster.
-Paul Edmon-
On 2/3/26 9:20 AM, Ron Gould via slurm-users wrote:
Hi there.
The SLURM DB service is installed alongside the other SLURM packages on the
node I'm starting with.
slurm-slurmdbd-22.05.3-1.el7.x86_64
slurm-22.05.3-1.el7.x86_64
slurm-contribs-22.05.3-1.el7.x86_64
slurm-devel-22.05.3-1.el7.x86_64
slurm-example-configs-22.05.3-1.el7.x86_64
slurm-libpmi-22.05.3-1.el7.x86_64
slurm-openlava-22.05.3-1.el7.x86_64
slurm-pam_slurm-22.05.3-1.el7.x86_64
slurm-perlapi-22.05.3-1.el7.x86_64
slurm-slurmctld-22.05.3-1.el7.x86_64
slurm-slurmd-22.05.3-1.el7.x86_64
slurm-torque-22.05.3-1.el7.x86_64
It doesn't seem like I am able to upgrade that one package first like the
instructions say to. Is it okay to upgrade all of the SLURM packages in one go?
If so, I would stop the slurmdbd service, back up the appropriate directories,
back up the DB into a self-contained file, upgrade the packages all together,
start the slurmdbd service ( `sudo -u slurm slurmdbd -D` ), and so forth
according to the instructions.
Is that a cogent plan?
--
slurm-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]
