Fedora Workstation Brainstorming A Possible GUI-Based Linux Recovery Environment

[phoronix]

Phoronix: Fedora Workstation Brainstorming A Possible GUI-Based Linux Recovery Environment

When it comes to system recovery on Linux, users are most often only left with a command-line for trying to recover from a failed kernel boot, borked boot loader configuration, or other show-stopping problems. With Fedora Workstation right now they have only their CLI-based Linux recovery process but are eyeing the possibility of creating a complementary GUI-based recovery environment...

https://www.phoronix.com/scan.php?page=news_item&px=Fedora-GUI-Based-Linux-Recovery

[⤐⤐⤐⤐]

Create artificially cripled installation scenarios to test your software against, procedural thrownage of problems your recovery software should pass and you will get a fine software, just don't transform into a "home project", document all bonking procedures and all the fixing approches and the resoaning around that fixing way taken, why think to create something like Mach when you can get something like even better coming from you and all the Linux community supporting you. Good Linuxing for you.

[amdevereux]

One of my SBCs (an Odroid HC4), uses petitboot as a bootloader, which is Linux. It has the option to connect to the internet, download an iso to RAM, and boot it. This means so long as I have internet, I have a recovery environment on hand. It's really helpful and something like that would be wonderful to have for any distro. I think Fedora is on the right track with this, especially because a simple Wayland compositor can be put into an initram using very little disk space.

[Waethorn]

It's a shame that there are too many hurdles to overcome with having a static OS image in Linux desktops. If there was a static OS image (preferably compressed), a second, older version copy that had a previously-verified full boot could be used to facilitate an OS rollback, ala ChromeOS. Silverblue does this already, sort of, but there are still a lot of issues to be worked out before this becomes a norm.

I don't like that user applications are similarly shoehorned into being containerized on atomic host OS's and standard packages make it so difficult to keep the host OS stable when it's atomic. I've run into numerous issues trying to install applications that want to write to somewhere else other than just /var on Silverblue. Anywhere else, and the OS will often become corrupted by future OStree updates to where it becomes unstable or fails to boot. Customizing the OS image after install is also trouble. More needs to be done to have bundled applications separated from the OS core.

[Waethorn]

One of my SBCs (an Odroid HC4), uses petitboot as a bootloader, which is Linux. It has the option to connect to the internet, download an iso to RAM, and boot it. This means so long as I have internet, I have a recovery environment on hand. It's really helpful and something like that would be wonderful to have for any distro. I think Fedora is on the right track with this, especially because a simple Wayland compositor can be put into an initram using very little disk space.

There needs to be better ARM GPU functionality (like, any) before you worry about having to recover an OS that doesn't support it.

[c117152]

NixOS retains the previous system states in-between updates / configuration changes so if something went wrong, you just hit the up arrow once or twice during the boot to switch to the previous state's configuration and you've recovered the system.

What Fedora could do to imitate this is to put the whole of /etc in a git repository and commit between reboots. Then all they need is a basic recovery environment that can do basic rollbacks over the git tree with some optional git and meld style tools that will let more advance users narrow down what recent change broke their boot-up.

[scottishduck]

Something like the macOS recovery environment would be very welcome. Current gold standard.

[uid313]

One time my GNOME desktop was messed up, so I had to install Weston.
Another time my packages were messed up, so I had to use Synaptic Package Manager.
Another time my network was messed up, I had to write weird commands to fix it.

One time when I shutdown Windows 10 it had put my network card into some deep ACPI sleep mode, so when I restarted to Linux my network card was still in sleep mode and wouldn't work. It would be great with some network diagnostics that check if I have an IP address, if it can resolve DNS, if the Ethernet cable is plugged in, if the network card is in some weird ACPI deep sleep mode.

[ruthan]

I also hope that Linux will once reach Windows 95 quality (first version with safe mode).. but it would only work, if would be same or better than some Live USB session, this should be connected with something like Windows System recovery, where system changes would be recorded and small incremental backups created.. and it would be actually working.

[ezst036]

Yes, please! Thank you. Good grief. This is sorely needed and has been overlooked for far too long.

There is nothing that should be left to the CLI like this. It's off-putting for users with a modicum of base-level tech skills. "If I just want to pick an old image, can I just please click the radio button? Why isn't there a list of available bak images here? Don't you value my time? Why do I have to spend ten minutes Googling the correct command? What is the correct command to Google anyways? The CLI doesn't always tell me and I wasn't born knowing the correct command. Oh wait, what was with this -s switch? There were three switches I needed to type?"

RM -RF that!!!!!! Just let me click it. Why is this so difficult? "I'm already frustrated that something broke, my patience is running thin from all of this Googling. I just want to fix it quick and go. I just want to get back to work." And all this assumes correctly typing the command. Accidental mistypes increases frustration even more cause guess what? "I can't cut and paste a command from one computer(or phone) to the other. Hopefully this isn't a fat finger moment. It didn't work, and now I have to closely examine this command to ensure I typed it correctly. Why?"

All of this is preventable. The CLI is a great option, but it should always be optional.

I'm sure someone might reply and say "yes but exceptional issues...." ok, well then let them remain exceptional and the CLI is great for exceptions. For the rest of us normal users, we need GUI.

/rant