By following this guide, you will have what is, in my opinion, the optimal setup to daily drive a rolling release distribution, such as Arch Linux.

If you have any kind of problem after updating your system, you’ll be able to boot into a previous snapshot and rollback to it, as if nothing had happened.

It’s basically like time travel for your computer, but better. Because you’ll be able to choose what gets rolled back and what doesn’t, so that you don’t lose any of your recent work.

Before we start

Requirements

• Your device must meet the system requirements of Arch Linux.
• You must have a bootable Arch Linux installation medium  , such as a USB Pendrive.

Suggestions

• Attempt to follow this guide first in a Virtual Machine should anything go wrong.
• Avoid using an old Arch ISO, especially on bleeding edge hardware.
• You should have a basic understanding of Linux distributions, Arch and btrfs, although I will make my best attempt of explaining everything like you’re 5.

Assumptions

• Your PC doesn’t have Secure Boot  enabled.
• Your PC is connected to the Internet via Ethernet.

Boot the live environment

Load your preferred keyboard layout if the default (en) is not appropriate for you. You can list them all by running the following command:

localectl list-keymaps

For example: As a Spaniard myself, I should choose es.

My keyboard layout is

loadkeys es

Drive partitioning

Identify your target drive

Use this command to list all connected drives to your computer:

fdisk -l

In every screenshot of this post I will be referring to your target drive as /dev/sda, however, you can change the text from the post here to match yours if you’d like.

My target drive is

Create system partitions

I prefer using gdisk instead of fdisk as the Arch Wiki suggests because it’s designed for GPT table partitions and has a few Quailty-of-Life features that fdisk lacks. Feel free to use your preferred partition manager if you have a preference of your own.

gdisk /dev/sda

Partition 1

Create a new partition: n

Accept the default partition number (1) by pressing ENTER.

Accept the default first sector by pressing ENTER.

Set the last sector (aka: partition size) to 1 Gibibyte. Strictly speaking, this is overkill. Although it’s better to over-allocate than to cut it short, it won’t be easy to grow this partition later: 1G

Change partition type to EFI system partition using this hex code: ef00 (these are two zeros, not two letters “O”).

Partition 2

Create a new partition: n

Accept the default partition number (2) by pressing ENTER.

Accept the default first sector by pressing ENTER.

Accept the default last sector (the remainder of the disk) by pressing ENTER.

Change partition type to Linux x86-64 root (/) using this hex code: 8304 (again, that is a zero, not an “O”).

Print the current partition layout

You can print the current partition layout with: p

Write the changes to the disk

Press w to write all changes made this far to the disk.

Formatting

Format the first partition: EFI system partition

mkfs.fat -F 32 /dev/sda1

Format the second partition: Linux x86-64 root (/)

mkfs.btrfs /dev/sda2

Mount the partitions

Mount the root partition

mount /dev/sda2 /mnt

Mount the boot partition

mount --mkdir /dev/sda1 /mnt/boot/efi

Partition results

Your partition table and mount points should look like this if you have followed every step correctly up until this point. You can verify this yourself using:

lsblk -f /dev/sda

Create Btrfs subvolumes

What are they and how do they differ from partitions

In a nutshell, partitions are logical sections of a drive that have independent file systems, mount points and files/directories.

Subvolumes are similar to partitions in the sense that are also logical sections of a drive with independent mount points. However, the file system (btrfs) is shared between all subvolumes created on that partition and files/directories may be shared between different subvolumes.

For example, if you create the following partition layout, each partition will have different file systems and the storage will be specific to each one.

Whereas a functionally identical layout using subvolumes will share the storage of the same btrfs filesystem between all mount points.

Suggested layout

Btrfs subvolume names are not required to match their mountpoint, although by doing so it’s much easier to know what goes where.

Now comes the “fun” part, creating and mounting all the subvolumes individually. Which can be done by executing the following commands for each subvolume.

# Don't copy this snippet just yet
btrfs sub create --parents /mnt/home
mount -t btrfs /dev/sda2 /mnt/home -o subvol=/home

Luckily, this task can be easily automated with a bash script.

SUBVOLUMES=(
    "/home"
    "/var/log"
    "/var/tmp"
    "/var/cache"
    "/var/crash"
    "/var/lib/AccountsService"
    "/var/lib/sddm" # you can skip this if you won't use sddm
    "/var/lib/portables"
    "/var/lib/machines"
)

for SUBVOLUME in "${SUBVOLUMES[@]}"
do
    btrfs sub create --parents "/mnt$SUBVOLUME"
    mount -t btrfs /dev/sda2 "/mnt$SUBVOLUME" -o rw,relatime,compress=zstd:1,ssd,discard=async,space_cache=v2,subvol="$SUBVOLUME"
done

Subvolumes results

You can verify if everything went smoothly by running findmnt -t btrfs /dev/sda2

Configure the system from outside

These commands must be executed from the live environment itself. You shouldn’t need to do anything unless you return here from the next section.

Install initial packages

pacstrap -K /mnt base linux-zen linux-zen-headers linux-firmware btrfs-progs sudo networkmanager nano grub grub-btrfs inotify-tools snapper

• linux-zen, linux-zen-headers: This is my preferred kernel due to being tuned for desktop linux  and making Waydroid  easier to run. There are other kernels  available to choose.
• btrfs-progs: Btrfs filesystem utilities (You’ve been able to use btrfs commands because the Arch ISO has this package installed by default).
• grub-btrfs: grub extension to support booting from btrfs snapshots.
• inotify-tools: Used by grub-btrfs to regenerate grub.cfg automatically when new snapshots are created.
• snapper: CLI utility to manage btrfs snapshots more easily

Generate fstab

This command will scan every current mountpoint of /mnt and save it to /etc/fstab relative to the new root filesystem (/mnt).

genfstab -U /mnt >> /mnt/etc/fstab

Configure the system from inside

Every command inside this section must be executed from within the newly created filesystem. To do that, run this command:

arch-chroot /mnt

Users and Groups

(Optional) Set a root password

Run passwd as root to set your password. This step is not mandatory, as you can always elevate to a root shell as a regular user using sudo.

Enable sudo for users other than root

When you install sudo on Arch Linux no user besides root can use it by default (It probably is that way to avoid unintended behavior).

To change that, edit /etc/sudoers using EDITOR=nano visudo and add the following line:

%wheel ALL=(ALL:ALL) ALL

Which means: Allow any user in group wheel to run any command as root using sudo as long as they provide their password.

Save the file using Ctrl+X and then press Y to confirm.

Create an unprivileged user

My username is

The first command creates a user, its home directory (/home/zlendy) and adds them to the wheel group. The second one sets their password.

useradd -m -G wheel zlendy
passwd zlendy

Time and Localization

If you don’t know your timezone,  you can run the following command to get it based on your IP address:

curl https://ipapi.co/timezone

Set the local time of your system.

ln -sf /usr/share/zoneinfo/{CONTINENT}/{CITY} /etc/localtime

As an example, I would set Continent to “Europe” and City to “Madrid”.

ln -sf /usr/share/zoneinfo/Europe/Madrid /etc/localtime

Run hwclock to generate /etc/adjtime

hwclock --systohc

Run nano /etc/locale.gen and uncomment at least en_US.UTF-8 UTF-8. Uncomment any other locales  as needed. (Tip: You can search in nano using Ctrl+W).

After you finish save the file and run locale-gen.

Run nano /etc/locale.conf and set the LANG variable to your preferred locale from before. Set it to en_US.UTF-8 if you don’t have any preference.

Run nano /etc/vconsole.conf and set the KEYMAP variable to es. This change makes it permanent so you won’t need to set it again.

Network configuration

You’ve had network connectivity throughout this guide thanks to the Arch ISO, but that won’t be the case when you boot natively into your system unless you execute this:

systemctl enable NetworkManager

Run nano /etc/hostname and type your desired hostname,  if you don’t do this it will be assigned to “archiso” by default.

Grub

Install grub to EFI system partition. Mark it as removable to create a .efi file and therefore make the system easier to boot into in case you change your motherboard.

grub-install --target=x86_64-efi --efi-directory=/boot/efi --removable
grub-mkconfig -o /boot/grub/grub.cfg

First boot to your new Arch install

Get out of the chroot using exit.

Unmount every partition in /mnt recursively using umount -R /mnt.

Shutdown the live environment using poweroff and unplug its installation media to avoid booting into it again accidentally. Then power on your computer again.

Log back in as root, if you didn’t set up a root password earlier you can elevate to a root shell using:

sudo -i

Snapper

This is a CLI tool to create and manage btrfs snapshot created by SUSE.

Create snapper configs

These commands will create a .snapshots subvolume in / and /home respectively.

snapper -c root create-config /
snapper -c home create-config /home

These newly created subvolumes won’t be mounted by default unless you add them to /etc/fstab. Replace {ROOT_UUID} with the UUID of the other btrfs subvolumes.

To avoid typing all of that by hand you can open the file with nano /etc/fstab, copy a line using Alt+6 and paste it using Ctrl+U. Don’t forget to change the mountpoint and subvol to /.snapshots and /home/.snapshots respectively.

# /etc/fstab
# Append these lines

UUID={ROOT_UUID}    /.snapshots    btrfs    rw,relatime,compress=zstd:1,ssd,discard=async,space_cache=v2,subvol=/.snapshots
UUID={ROOT_UUID}    /home/.snapshots    btrfs    rw,relatime,compress=zstd:1,ssd,discard=async,space_cache=v2,subvol=/home/.snapshots

After you’re done run systemctl daemon-reload and then mount -va.

Configure user permissions

Run these commands to be able to run snapper commands as your unprivileged user without using sudo.

snapper -c root set-config ALLOW_USERS=zlendy SYNC_ACL=yes
snapper -c home set-config ALLOW_USERS=zlendy SYNC_ACL=yes

Create an initial subvolume

This will allow you to perform your first rollback should you ever need it.

snapper create -d "Initial subvolume" --read-write

snapper create by default makes snapshots, in other words, read-only subvolumes.

Just this once, you will need to add a flag to create it in read-write mode. Do not do this regularly, it is a bad practice to modify btrfs snapshots.

Change the default subvolume of / to /.snapshots/1/snapshot

SNAPSHOT_ID="$(btrfs inspect-internal rootid /.snapshots/1/snapshot)"
btrfs subvolume set-default $SNAPSHOT_ID /

Create snapshots automatically

Snapshots not only give you the means to restore system data, they also allow you to restore your data.

Therefore, it is advisable to create snapshots automatically every now an then, so in the unfortunate scenario that you delete data you shouldn’t have deleted, you can easily recover it.

Enable the following services to create snapshots at certain moments of the day (snapper-timeline) and every time you turn on your computer (snapper-boot).

grub-btrfs is a daemon which will regenerate GRUB every time a snapshot is created or removed.

systemctl enable --now snapper-timeline.timer snapper-boot.timer grub-btrfsd

You may want to create snapshots every time you install or uninstall a package. By installing snap-pac, a hook will be added to pacman which creates pre/post snapshots on every pacman transaction.

These are a special kind of snapshot that should always be created in pairs. They represent the state before and after a change is made respectively.

pacman -S snap-pac

Testing snapshots

At this point, the system should be perfectly suitable for daily usage. But don’t stop reading just yet! You should test if booting into a snapshot and rolling back works before finding it out the hard way.

Recover data from a snapshot

Browse the contents of .snapshots/{SNAPPER_ID}/snapshot and copy any file or directory found there to another writable location. Substitute {SNAPPER_ID} with the numerical ID of the snapshot you wish to access.

If the data you wish to recover is contained in /home you should go to /home/.snapshots, otherwise go to /.snapshots.

Remember that / and /home are different subvolumes and as such have different snapshots each (you can check that using snapper -c root ls and snapper -c home ls).

The image below shows an example of data recovery using btrfs subvolumes, feel free to experiment and tinker with alternative methods.

Boot into a snapshot

Make any change that is easily noticeable in / (not /home), such as installing a package before going on. In this example I installed cowsay.

Reboot the system. You will notice there’s an additional GRUB menu named “Arch Linux Snapshots” just below “Reboot into UEFI”. Select any entry created prior to your change in that list and boot into it.

When you do, that filesystem will be mounted in read-only mode. Meaning that no packages can be installed, removed, upgraded, etc.

After I boot into snapshot number 4, cowsay is gone.

What do those symbols beside snapshot IDs mean in snapper ls?

Roll back to a previous snapshot

snapper rollback

That command will do the following:

1. Create a read-only snapshot of the subvolume currently in use.
2. Create a read-write snapshot of the previous one.
3. Set the default subvolume of / to the read-write snapshot created in step 2.

Select “Arch Linux” and press e.

Manually edit the number of the subvolume to the newly created one and boot it using F10 or Ctrl+X.

Deleting old snapshots

There are timers that delete snapshots automatically for you, but sometimes you will need to manually delete a snapshot to recover storage ASAP.

Fortunately, that’s an easy task. You can just run:

snapper -c {CONFIG} rm {SNAPPER_ID}
snapper -c {CONFIG} rm {SNAPPER_ID_1}-{SNAPPER_ID_2}

Just don’t delete the snapshot that GRUB is currently using to load the configuration files (the one marked in snapper ls with a * or +), if you want to have a bootable system, that is.

Acknowledgements

This guide was created with the assistance of:

• Arch Wiki: Installation guide 
• Arch Wiki: Btrfs 
• Arch Wiki: GRUB 
• SysGuides: How to Install Fedora 41 with Snapshot and Rollback Support 

Afterword

Note that btrfs snapshots ARE NOT backups by themselves, you’d need to send them to a different storage media using, for example, btrfs send.

And that concludes this guide, the first one I make, actually. If you have any suggestions or find any errors please let me know in the comments section below or contact me directly on social media so that I can fix it. Thank you for reading!