Making updates a little safer on RHEL?

[io7m]

Hello!

I run a single RHEL 9 server (Rocky, to be precise) for a small business. We used to run Fedora, but found it too fast moving, and almost every dnf update would end up breaking zfs due to version incompatibility. Moving to Rocky/RHEL obviously helped matters a lot as it's much slower moving, and we'd largely been running dnf update fairly often expecting that we'd only be receiving security updates. This seemed to be fine, nothing ever changed to the point that zfs would stop being available after a reboot.

Then, of course, this morning we upgraded, rebooted, and found zfs no longer working. A workaround in that ticket (dnf update --enablerepo=zfs-testing zfs) got us running again.

Is there some way we can ensure that we don't perform an update if the resulting system won't have working zfs? I was less vigilantly watching dnf update as I wasn't expecting a zfs breakage, but if there was any kind of warning/error, it apparently passed by unnoticed. We really only run dnf update for security patches.

[mskarbek]

https://github.com/zeet-cc/zfs-rpms/blob/main/kernel-zfs.spec - I'm using this as a safeguard. It's an empty package that locks in kmod-zfs and kernel-core together, which prevents undesirable kernel updates. Works perfectly fine on both Fedora and RHEL.