So I took the plunge and installed Fedora Silverblue because of all that immutable buzz. And it’s the most frustrating change I have made in almost 20 years of my distrohopping.
After installing Silverblue I configured it as usual. I installed necessary flatpaks, played with toolbox and distrobox, installed codecs, configured my bluetooth keyboard and other stuff in /etc and /var. Applied some useful tweaks I found on the web and… well… everything works. Nothing to do anymore. No issues. Nothing breaks, no dependency hell, everything runs smooth. I have nothing to tweak, tinker or configure anymore. So frustrating.
Every update is just… meh. Smooth, new, fresh system not affected by my stupid tweaking and breaking. Booooring.
I don’t have to distrohop anymore. If I want other distros I can just install them in distrobox. Other versions of apps? Something from AUR perhaps…? No problem. What’s the point of distrohopping now? Other DEs? I just rebase my system to other images with almost any DE or WM I want without losing data or messing everything up (damn you, UBlue!).
I don’t even have to reinstall the damn thing cause every time I update the system or rebase it to another image it’s like reinstalling it.
Silverblue killed distrohopping for me. Really frustrating.
Oh man. I’m so sorry for your loss. May your system break at some vague point in the future in a way that is nigh impossible to diagnose and that no one else seems to have experienced. Godspeed, you unwillingly content penguin!
that the thing, if it breaks, the roolback is there or simply rebase without merging /etc, so basically a factory reset
deleted by creator
I’ve been running Bazzite based on silverblue on my desktop for remote gaming and dockering. Everything was amazing until I started doing some mid-level docker stuff because of the rigidity of the distro.
Podman largely works but since it’s rootless it won’t have access to mounted drives easily due to SELinux.
Mounting a drive automatically wasn’t intuitive either and I ended up editing the /etc/fstab manually.
Setting up a swapfile was also tedious, I needed more than 8GB so I made a 32GB swapfile but I still had to run a sudo command on startup since I’m not really confident with creating a systemd service on an immutable distro.
All in all I should have just gone for Nobara or a regular Fedora but that’s because I have a really edge use-case.
That being said I still highly recommend it. It’s stable, easy to “rebase-hop” and everything just works well and it’s very stable. I’d recommend it for pretty much anyone unless you’re going to do some heavy self hosting with multiple HDs.
Wow I was so confused while reading this haha, got me good there! Happy to hear its working as expected :P
Silverblue killed distrohopping for me. Really frustrating.
Then kill it. Distrohop again.
Only thing I haven’t figured out, yet, is how to install the Private Internet Access client. It uses a
.run
install script, and it fails when installing viarpm-ostree
(tries to write to/etc
) and doesn’t like being installed in a Distrobox (needs systemd).But yeah, I’m currently looking at some other options for my main system to drop Windows, and I’m always comparing to Fedora Atomics, now.
Any program with an install script makes assumptions about your system, if it doesn’t work it just isn’t compatible.
Either modify the script, package the software for your distro or find out if someone else has done it.
My first instinct would be to look if it’s in the AUR and install it inside an Arch Toolbox.Yeah, third-party Linux VPN clients are pretty screwed on silverblue, and probably always will be. Especially since when installed in a container, they require being ran in a rootful container with selinux labeling disabled to enable direct access to /dev/net/tun, and as you’ve quickly found out, most of those weird bash based installers haven’t adapted. It’s best to use generic VPN configs through your DE atm.
you can unlock your /usr with rpm-ostree usroverlay
you can unlock the file system, don’t remember how tho
why not use fedora’s built-in openvpn client and just add the pia info? That should likely work. https://helpdesk.privateinternetaccess.com/guides/linux/linux-installing-openvpn-through-the-terminal
or built-in wireguard client? https://helpdesk.privateinternetaccess.com/guides/linux/alternative-setups-4/linux-manual-connection-scripts
I don’t use PIA, but /opt and /etc are both r/w in Silverblue/Kionite
Can you still install extensions in GNOME? I hate the defaults
Yes but only from Gnome directly with an app called extensions manager. You came install them from the Fedora repo.
Thank you!
I’m a bit behind on these immutable distros and have a small question. People keep saying you can just switch to another image if you want to switch desktop environments. But how does this solve the problem of the config files of the various DEs (GTK rc files or other theme stuff) messing with each other in the home directory? Because this was always a pain in the ass in normal distros
I guess you just have to use home manager on NixOS
But how does this solve the problem of the config files of the various DEs (GTK rc files or other theme stuff) messing with each other in the home directory?
It does not. Your dotfiles will be a bit wrecked when you rebase. See: https://universal-blue.discourse.group/t/why-is-rebasing-between-desktop-environments-bad/690/4 It’ll also cause random issues like: https://discussion.fedoraproject.org/t/flatpak-apps-crashing-after-rebasing-from-silverblue-to-kinoite/83623/2
It’s mostly plasma fighting gnome, though. I haven’t seen any conflicts with say, sway.
Switching DEs is not recommended by devs so I assume the configs are still conflicting. Home dir doesn’t get affected by an image rebase most likely.
I’ve switched between Plasma, Cosmic, Sway, and Hyprland without any conflicts. For the Plasma 5->6 transition it did change my config in a way that broke Plasma 5 when I rolled back, so problems are possible.
Basically your mileage may vary.
yeah, home directory is mutable, but you can simply create another user, the /etc is also mutable(the system do a diff of it every update) but you can see every file that changed there(compared with the remote image) using ostree, or create another deploy where you discart your /etc, so, if you discart your /etc, and create another user, you have fresh install, without needing to reinstall using a pendrive etc
Your files are a mutable part, they stick around for rebase and rollback. (I believe /etc also.) If it’s only files in a home directory you could try a different DE by making a new user. But yeah I don’t think it has a built-in solution for something like that.
deleted by creator
I had an entirely different experience with Bazzite. It would not boot to Wayland after an update, so I had to boot to xorg, reboot, and then wayland would work, until the last update where Wayland just wouldn’t work anymore, so I ended up going back to Fedora Workstation.
Bazzite has been smooth sailing about 80% of the time for me. The rest of the 20% were due to either plasma or runner crashing, requiring me to perform a hard reset using the power button. And then it magically atarted working again. I’ve also had my home folder become read-only on occasion. Very strange.
11 months later …
NixOS looks interesting whoosh sucked into a warp
Love the irony, but this is painting a little too good a picture
Every update is just… meh. Smooth, new, fresh system not affected by my stupid tweaking and breaking
Most times yes, but major updates usually cause some trouble, like from 39 to 40, you couldn’t do it without uninstalling the codecs for Firefox. Firefox that is installed by default as an RPM, because the Flatpak Firefox doesn’t yet have 100% compatibility with all the features that work with the RPM, so as a user you’re pretty much led to get yourself stuck in this hole, not too difficult to fix in the end, but still a pain to find out and fix.
Everything else is 100% true! And I think it will be always hard to beat as an implementation of immutability (second place only to NixOS imo), A/B partitioning doesn’t hold a candle to OSTree
you couldn’t do it without uninstalling the codecs for Firefox
what happened is rpm-fusion was lagging behind the official fedora repos, so, you could have just waited, or enabled the automatic update and forget about it
Is that so? From the issue I read there was no way around it because the two images are fundamentally incompatible once you layer that package, you had to remove the layered package, it seemed from the discussion that they might have “fixed” the base image at some point as a pull request was opened on Pagure. I waited a bit for it to go upstream, but nothing happened for a long time and just went thorugh with the manual intervention, and actually, now that I check it again, the maintainer siosm commented that they can’t accept the PR
oh, i never had that issue, only the rpm-fusion lag, never thought that the codecs needed a different approach
Weird. I use Bazzite which is off of Kinoite and the upgrade from 39 -> 40 was seamless.
Doesn’t Bazzite have the base image modified to have the codecs included already? I think that’s probably why you didn’t experience any disruption there
- Full hardware accelerated codec support for H264 decoding.
link: https://github.com/ublue-os/bazzite/?tab=readme-ov-file#about--features
If installing the surface kernel (kind of necessary for my Surface Go 1) and installing a few appimages didn’t look so difficult, I guess I would already be on Silverblue.
I’m kind of the opposite of OP and just having nightmares about breaking my system 😅
That’s why I’m doing clonezilla backup but I think the custom kernel would be a problem if I reinstall on another non-Surface computer. Maybe I should just go back to the normal kernel before doing a backup…
You can make your own silverblue image with your custom kernel ;)
If the kernel is available in a COPR or another third party repo, you can just do a little swapping with rpm-ostree: https://github.com/openshift/os/blob/master/docs/faq.md#q-how-do-i-replace-the-current-kernel-with-kernel-rt-or-a-new-kernel-version-in-rhcos
Edit: Just in case this is the project you’re using, here’s specific install instructions for Fedora Silverblue: https://github.com/linux-surface/linux-surface/wiki/Installation-and-Setup#fedora-silverblue
Seems a bit too complicated for me, even if it probably ain’t.
But I’d probably use it if one day I break my Fedora workstation install.
Oh it’s definitely over-complicated, and contrary to what others say here, Silverblue can definitely have some very difficult to troubleshoot problems (especially when using things outside the direct Fedora ecosystem), which are greatly worsened by rpm-ostree taking 15 years to do anything despite sharing code with the supposedly lighting-quick dnf5. For servers, rpm-ostree is great (it’s in all of RH k8s offerings, see RHCOS), but on desktops, there’s definitely a good reason why RH has to apparent offering and Fedora calls theirs “emerging”. Still miles better than having an unbootable system after updating.
I’ve had a similar experience with Guix.
Oh, you!