[erwincoumans]

It would be great if Linux had some automatic checkpoints so you can revert to a working system after an update. So many times updates leave a broken system and re-install Ubuntu from scratch is quicker than fixing the problem.

[usr1106]

OpenSUSE had this 7 years ago when I last used it. I think they still have it, it's called snapper.

Not that I would really have turned any Linux significantly non-functional by updating during the last 15 years. But it's all software, so everything can happen one day...

[boomboomsubban]

Bootable snapshots are a thing, also known as boot environments. You need to use ZFS or btrfs, or possibly bcachefs or LVM offer them. And they never have seemed to have a great implementation on Linux compared FreeBSD, but they do exist.

[tredre3]

> but they do exist.

And here's the crux of it, isn't it? Unless it's accessible and enabled by default, it might as well not exist from the PoV of the end user.

Windows' restore points happen automatically before updates or driver updates. Then you're prompted automatically to restore if Windows fails to boot after an update.

As usual Linux has all the pieces but no desire to create UX to go with it. Which is a shame because ZFS with automatic snapshotting daily or before major events combined with a simple UI that can run in the preboot environment would be a game changer and make Windows look like a toy.

[dizhn]

OpenSUSE Tumbleweed has this. The installer defaults to btrfs and snapshots using snapper. Snapshots are automatic before and after every package manager operation. Save for something that breaks grub, you can always recover.

[bravetraveler]

In that spirit, if you have an EFI system, you don't need GRUB

[boomboomsubban]

You sort of do if you want to boot kernels on btrfs. Refind is an option, but not most of the other EFI bootloaders.

[bravetraveler]

Ah, good point - I neglected the BTRFS involvement/need to find kernels... that one would ideally include with these snapshots, of course

[account42]

You put the kernel (built with efistub) directly onto the EFI partition - no boot loaded needed at all.

[vladvasiliu]

> And here's the crux of it, isn't it? Unless it's accessible and enabled by default, it might as well not exist from the PoV of the end user.

For what it's worth, at one point Ubuntu supported installing on ZFS through its regular installer and installed the boot environment thingy, with a snapshot before every update. I have one such machine which is now running 23.04 and that still works. I've never had to use in practice, though.

[hef19898]

Oh, so this! Having had the option last Friday to just reboot into whatever it was before said update, and just skip the next couple of updates, man, that would have been great!

[michaelmrose]

To be pedantic It's impossible for "Linux" to have this as this exists way higher than the kernel or even the filesystem layer.

Linux mint has timeshift.

The now defunct Project Trident installer set up a zfs root with zfsbootmenu and an update script that automatically sets up a boot environment you can revert and boot from with the built in feature of zfsbootmenu.

Both let you achieve your goal. You can set up either system even if they aren't built in.

[rkagerer]

A lot of comments here say Timeshift is crap - do you find it helpful yourself and can you elaborate on what works / doesn't?

https://community.linuxmint.com/software/view/timeshift

[michaelmrose]

I have not personally used timeshift. I have used zfsbootmenu and syncoid found it useful and fit for purpose. Personally I think its more reasonable to separate boot environments from backing up as they are two separate if related concerns. EG boot environments are for when something went wrong with an update and backups are for when your drive failed. If both went wrong you just do both operations in sequence restore from backup and THEN pick the prior boot environment and make it the default.

Insofar as timeshift consider the review source, circumstances, and nature of complaints. Users who don't have difficulties rarely post anything and yet many reviews are positive. Those that aren't seem to focus on people who didn't realize they could provide a secondary drive and just kept storing data until their OS drive filled up.

Mint leans heavily towards technically incompetent users and yet many people were able to use it successfully. Meeting every users needs no matter how incapable is probably a poor benchmark to define if somethings is "crap" or not even if those users problems are are a great guideline to a path forward to helping all users better.

For instance it might be better if timeshift only supported a fs with snapshots instead of rsync and required a secondary disk and warned of space issues long before the disk fills up but it seems quite possible to competently use it right now.

[justinclift]

Some Linux distros actually do this. :)

[joshspankit]

Once NixOS is desktop-ready you’ll have a way more powerful checkpoint system

[tracker1]

I haven't had that happen for a few years now... I will say the RX 5700 XT had a really rough few months after launch on Linux. Other than that, it's been relatively smooth.

Mostly been using PopOS and Ubuntu-Budgie.

[SAI_Peregrinus]

NixOS has this by default, on any filesystem. You can also do it with ZFS or BTRFS filesystem snapshots.

[aednichols]

Check out immutable distros like Fedora Silverblue & Kinoite.