I’ve been running Rocky Linux on all my servers since 8.4 was published. A few weeks ago, I decided to also move my desktop clients to a mix of Rocky Linux and KDE from EPEL, since I was weary of OpenSUSE Tumbleweed’s weekly upgrade tsunami (and Leap’s future is more than uncertain).
So my workstation was almost complete. Last thing to setup was KVM,
virt-manager to manage virtual machines, and… uh oh. Trying to setup a VM using
virt-manager results in a hard freeze.
Hard reboot, try again, same result: freeze.
So let’s upgrade everything first. Before doing that, I’ll give it a spin on my sandbox PC, which is running the same setup. Uh oh, upgrade conflicts, so I do a
dnf remove qt5-* to get rid of KDE, we’ll deal with that later. Upgrade goes fine, so now I have to reinstall the
KDE Plasma Workspaces package group. There’s a bunch of conflicts, but since the new KDE is in EPEL Testing, I enable this repository. This time the install goes fine, but I have to reinstall the
nvidia-x11-drv-470xx driver. Reboot to SDDM and… uh oh. No login manager, just a black screen with my mouse pointer, and no SDDM. Looks like something shit the bed again with KDE vs. RHEL like in 8.5.
So now I have an unusable system and another one I can’t update.
The whole point of running an enterprise class system (because that’s what I’ve been doing since the days of CentOS 4.x) was precisely to avoid this kind of nasty surprises.
This is one of these days where I seriously consider moving to Debian and deal with their shorter support cycle.
Edit: looks like KDE from EPEL only works on even point releases (8.4, 8.6, etc.) but not on uneven point releases (8.5, 8.7, etc.)
Edit 2: Why do minor upgrades break basic things like variables in
/etc/os-release ?!? Why ?!?