Jump to content

Btrfs: Difference between revisions

From NixOS Wiki
imported>Onny
Add section on subvolume snapshots
Ponder (talk | contribs)
 
(29 intermediate revisions by 16 users not shown)
Line 1: Line 1:
'''btrfs''' is a modern copy on write (CoW) filesystem for Linux aimed at implementing advanced features while also focusing on fault tolerance, repair and easy administration.
[https://btrfs.readthedocs.io/en/latest/ btrfs] is a modern copy on write (CoW) filesystem for Linux aimed at implementing advanced features while also focusing on fault tolerance, repair and easy administration.
 
{{note| Use [https://github.com/nix-community/disko/ disko] to manage your NixOS storage layout declaratively. The following shows a manual approach as seen in traditional Linux distributions.}}


== Installation ==
== Installation ==
Line 12: Line 14:
=== Format partitions and create subvolumes ===
=== Format partitions and create subvolumes ===
<syntaxhighlight lang="console">
<syntaxhighlight lang="console">
# nix-shell -p btrfs-progs
# mkfs.fat -F 32 /dev/sdX1
# mkfs.fat -F 32 /dev/sdX1


Line 37: Line 40:
<syntaxhighlight lang="console">
<syntaxhighlight lang="console">
# nixos-generate-config --root /mnt
# nixos-generate-config --root /mnt
# nano /mnt/etc/nixos/configuration.nix # manually add mount options
# nano /mnt/etc/nixos/configuration.nix # manually add mount options (see Compression below for an example)
# nixos-install
# nixos-install
</syntaxhighlight>
</syntaxhighlight>
Line 43: Line 46:
== Configuration ==
== Configuration ==


=== Compression == =
=== Compression ===


<code>nixos-generate-config --show-hardware-config</code> doesn't detect mount options automatically, so to enable compression, you must specify it and other mount options in a persistent configuration:
<code>nixos-generate-config --show-hardware-config</code> doesn't detect mount options automatically, so to enable compression, you must specify it and other mount options in a persistent configuration:
Line 65: Line 68:
# umount /mnt
# umount /mnt
# mkdir /swap
# mkdir /swap
# mount -o subvol=swap /dev/sdXY /swap
# mount -o noatime,subvol=swap /dev/sdXY /swap
</syntaxhighlight>
</syntaxhighlight>


Then, create the swap file with copy-on-write and compression disabled:
Then, create the swap file and adjust its size as desired:


<syntaxhighlight lang="console">
<syntaxhighlight lang="console">
# truncate -s 0 /swap/swapfile
# btrfs filesystem mkswapfile --size 8g --uuid clear /swap/swapfile
# chattr +C /swap/swapfile
# btrfs property set /swap/swapfile compression none
# dd if=/dev/zero of=/swap/swapfile bs=1M count=4096
# chmod 0600 /swap/swapfile
# mkswap /swap/swapfile
</syntaxhighlight>
</syntaxhighlight>


Line 83: Line 81:
swapDevices = [ { device = "/swap/swapfile"; } ];
swapDevices = [ { device = "/swap/swapfile"; } ];
</syntaxhighlight>
</syntaxhighlight>
=== Scrubbing ===
Btrfs filesystems by default keep checksums for all files, to monitor if the file has changed due to hardware malfunctions or other external effects.
Scrubbing is the process of checking file consistency, which may use checksums and/or duplicated copies of data, from raid for example. Scrubbing may be done "online", meaning you don't need to unmount a subvolume to scrub it.
You can enable automatic scrubbing withː
<syntaxhighlight lang="nix">
services.btrfs.autoScrub.enable = true;
</syntaxhighlight>
Automatic scrubbing by default is performed once a month, but you can change that withː
<syntaxhighlight lang="nix">
services.btrfs.autoScrub.interval = "weekly";
</syntaxhighlight>
<code>interval</code> syntax is defined by [https://www.freedesktop.org/software/systemd/man/systemd.time.html#Calendar%20Events systemd.timer's Calendar Events]
By default, autoscrub will scrub all detected btrfs mount points. However, in case of mounted nested subvolumes (e.g. the example above where <code>/nix</code> and <code>/home</code> are nested subvolumes under <code>/</code>), you only need to scrub the topmost one. So an example configuration may look like this:
<syntaxhighlight lang="nix">
services.btrfs.autoScrub = {
  enable = true;
  interval = "monthly";
  fileSystems = [ "/" ];
};
</syntaxhighlight>
The result of the periodic auto scrub will be saved to the system journal, and you can check the status of the last scrubː
<syntaxhighlight lang="bash">
btrfs scrub status /
</syntaxhighlight>
You can also start a scrub in the background manuallyː
<syntaxhighlight lang="bash">
btrfs scrub start /
</syntaxhighlight>
You can check the status of the ongoing scrubbing process with the same <code>status</code> command as above.
=== Deduplication ===
Files with (partially) equal contents can be deduplicated using [https://github.com/Zygo/bees bees] or [https://github.com/markfasheh/duperemove duperemove].
bees can be configured in <code>configuration.nix</code>:
<syntaxhighlight lang="nix">
services.beesd.filesystems = {
  root = {
    spec = "LABEL=root";
    hashTableSizeMB = 2048;
    verbosity = "crit";
    extraOptions = [ "--loadavg-target" "5.0" ];
  };
};
</syntaxhighlight>
This will run the daemon in the background. To disable auto-start, use <code>systemd.services."beesd@root".wantedBy = lib.mkForce [ ];</code> for each filesystem.


== Usage ==
== Usage ==


=== Snapshots ====
=== Subvolumes ===
 
Create a subvolume
 
<syntaxhighlight lang="bash">
btrfs subvolume create /mnt/nixos
</syntaxhighlight>


Taking a read-only (<code>-r</code>) snapshot called <code>nixos_snapshot_202302</code> of the subvolume mounted at <code>/</code>
Remove a subvolume


<syntaxhighlight lang="bash">
<syntaxhighlight lang="bash">
btrfs subvolume snapshot -r / /mnt/@nixos_snapshot_202302
btrfs subvolume delete /mnt/nixos
</syntaxhighlight>
</syntaxhighlight>


List snapshots for <code>/</code>
=== Snapshots ===
 
A snapshot in btrfs is simply a subvolume that shares its data (and metadata) with some other subvolume, using btrfs's CoW capabilities. Because of that, there is no special location for snapshots and you can decide where you want to store them. It can be a simple directory inside the root subvolume, or a directory inside a dedicated "snapshots" subvolume.
 
For this example we are going to store snapshots in a directory <code>/snapshots</code>, that has to be created beforehand with <code>sudo mkdir /snapshots</code>
 
To take a read-only (<code>-r</code>) snapshot called <code>home_snapshot_202302</code> of the subvolume mounted at <code>/home</code>


<syntaxhighlight lang="bash">
<syntaxhighlight lang="bash">
sudo btrfs subvolume list /
btrfs subvolume snapshot -r /home /snapshots/home_snapshot_202302
</syntaxhighlight>
</syntaxhighlight>


You can also snapshot the root subvolume. But keep in mind that nested subvolumes are '''not''' part of a snapshot. So if you have subvolumes <code>/nix /home</code>, taking a snapshot of <code>/</code> will not include them.
<syntaxhighlight lang="bash">
btrfs subvolume snapshot -r / /snapshots/nixos_snapshot_202302
</syntaxhighlight>
Make snapshot read-write againː
<syntaxhighlight lang="bash">
btrfs property set -ts /snapshots/home_snapshot_202302 ro false
</syntaxhighlight>
However, changing read-only property of a snapshot in-place may [//lore.kernel.org/linux-btrfs/06e92a0b-e71b-eb21-edb5-9d2a5513b718@gmail.com/ causes issues] with any future incremental send/receive. Instead, a read-only snapshot itself (being a simple subvolume) can be snapshotted again as a read-write snapshot like this:<syntaxhighlight lang="bash">
btrfs subvolume snapshot /snapshots/home_snapshot_202302 /snapshots/home_snapshot_202302_rw
</syntaxhighlight>
Or it can be restored directly to <code>/home</code> straight away like this:
{{warning|1=this will delete the current <code>/home</code> and restore the snapshot! <code>/home</code> must be unmounted for this operation}}
<syntaxhighlight lang="bash">
btrfs subvolume delete /home
btrfs subvolume snapshot /snapshots/home_snapshot_202302 /home
</syntaxhighlight>
After this you can mount <code>/home</code> again.
=== Transfer snapshot ===
Sending the snapshot <code>/snapshots/nixos_snapshot_202302</code> compressed to a remote host via ssh at <code>root@192.168.178.110</code> and saving it to a subvolume mounted on a directory at <code>/mnt/nixos</code>
<syntaxhighlight lang="bash">
sudo btrfs send /snapshots/nixos_snapshot_202302 | zstd | ssh root@192.168.178.110 'zstd -d | btrfs receive /mnt/nixos'
</syntaxhighlight>
== Tips and tricks ==
=== Installation with encryption ===
Using [https://en.wikipedia.org/wiki/Linux_Unified_Key_Setup Luks2]:
<syntaxhighlight lang="bash">
cryptsetup --verify-passphrase -v luksFormat "$DISK"p2
cryptsetup open "$DISK"p2 enc
</syntaxhighlight>
You can use any device partition for your bootloader. Note that this bootloader is unencrypted by default:
<code>mkfs.vfat -n BOOT "$DISK"p1</code>
==== Creating subvolumes ====
<syntaxhighlight lang="bash">
mkfs.btrfs /dev/mapper/enc # Creating btrfs partition
mount -t btrfs /dev/mapper/enc /mnt
# Create the subvolumes
btrfs subvolume create /mnt/root # The subvolume for /
btrfs subvolume create /mnt/home # The subvolume for /home, which should be backed up
btrfs subvolume create /mnt/nix # The subvolume for /nix, which needs to be persistent but is not worth backing up, as it’s trivial to reconstruct
btrfs subvolume create /mnt/log # The subvolume for /var/log.
</syntaxhighlight>
Unmount to mount on the subvolumes for the next steps:
<code>umount /mnt</code>
Once the subvolumes have been created, mount them with the desired options.
Example with [https://facebook.github.io/zstd/ Zstandard compression] and noatime:
<syntaxhighlight lang="bash">
mount -o subvol=root,compress=zstd,noatime /dev/mapper/enc /mnt
mkdir /mnt/home
mount -o subvol=home,compress=zstd,noatime /dev/mapper/enc /mnt/home
mkdir /mnt/nix
mount -o subvol=nix,compress=zstd,noatime /dev/mapper/enc /mnt/nix
mkdir -p /mnt/var/log
mount -o subvol=log,compress=zstd,noatime /dev/mapper/enc /mnt/var/log
# do not forget to create and mount the bootloader
mkdir /mnt/boot
mount "$DISK"p1 /mnt/boot
</syntaxhighlight>
Configure <code>hardware-configuration.nix</code>
<syntaxhighlight  lang="nix">
# enable btrfs support
boot.supportedFilesystems = [ "btrfs" ];
fileSystems."/var/log" =
    { device = "/dev/disk/by-uuid/X";
      fsType = "btrfs";
      # enable noatime and zstd to the other subvolumes aswell
      options = [ "subvol=log" "compress=zstd" "noatime" ];
      # to have a correct log order
      neededForBoot = true;
    };
</syntaxhighlight >
Generate Nixconfig:
<syntaxhighlight lang="bash">
nixos-generate-config --root /mnt
</syntaxhighlight >
=== Convert Ext3/Ext4 system partition to Btrfs ===
<div style="border: 1px solid #D33; background: #FFEBEB; padding: 30px; border-radius: 5px; margin: 10px 0px; display: flex; align-items: center;">
    <div style="color: #D33; font-size: 40px; margin-right: 15px; background: #FFEBEB; display: flex; line-height: 0;  align-items: center;">⚠</div>
    <div style="color: #D33; font-size: 15px; font-style: normal; font-weight: 400; line-height: normal; text-align: left;">Note that migrating your existing root filesystem can cause data loss or make your system unbootable. Make sure to backup the partition or your files. Proceed only if you know what you're doing!</div>
</div>
To convert the existing filesystem (Ext3/4) to Btrfs, boot into a NixOS live system and run the following commandː<syntaxhighlight lang="sh">
fsck -f /dev/sdXY
btrfs-convert /dev/sdXY
</syntaxhighlight>Replace the device path with the target partition. Converting larger filesystems can take a long time.
Next, mount the converted filesystem and chroot into itː<syntaxhighlight lang="sh">
mount /dev/sdXY /mnt
nixos-enter --root /mnt
</syntaxhighlight>Replace the partition UUID with the new one, which can be obtained using the command <code>blkid</code>, in <code>/etc/nixos/hardware-configuration.nix</code> and also change the filesystem to <code>btrfs</code>.<syntaxhighlight lang="nix">
  fileSystems."/" =
    { device = "/dev/disk/by-uuid/44444444-4444-4444-8888-888888888888";
      fsType = "btrfs";
    };
</syntaxhighlight>Apply the changes<syntaxhighlight lang="sh">
nixos-rebuild boot
</syntaxhighlight>If the Grub bootloader is used and it doesn't get reinstalled correctly, you can run the following command inside chrootː<syntaxhighlight lang="sh">
grub-install /dev/sdX
</syntaxhighlight>If the conversion was successful and no rollback is required, the backup image which was stored by btrfs-convert can be removed withː<syntaxhighlight lang="sh">
btrfs subvolume delete /btrfs/ext2_saved
</syntaxhighlight>
[[Category: Configuration]]
[[Category: Configuration]]
[[Category:Filesystem]]

Latest revision as of 14:22, 19 February 2025

btrfs is a modern copy on write (CoW) filesystem for Linux aimed at implementing advanced features while also focusing on fault tolerance, repair and easy administration.

Note: Use disko to manage your NixOS storage layout declaratively. The following shows a manual approach as seen in traditional Linux distributions.

Installation

Note: The following example is for EFI enabled systems. Adjust commands accordingly for a BIOS installation.

Partition the disk

# printf "label: gpt\n,550M,U\n,,L\n" | sfdisk /dev/sdX

Format partitions and create subvolumes

# nix-shell -p btrfs-progs
# mkfs.fat -F 32 /dev/sdX1

# mkfs.btrfs /dev/sdX2
# mkdir -p /mnt
# mount /dev/sdX2 /mnt
# btrfs subvolume create /mnt/root
# btrfs subvolume create /mnt/home
# btrfs subvolume create /mnt/nix
# umount /mnt

Mount the partitions and subvolumes

# mount -o compress=zstd,subvol=root /dev/sdX2 /mnt
# mkdir /mnt/{home,nix}
# mount -o compress=zstd,subvol=home /dev/sdX2 /mnt/home
# mount -o compress=zstd,noatime,subvol=nix /dev/sdX2 /mnt/nix

# mkdir /mnt/boot
# mount /dev/sdX1 /mnt/boot

Install NixOS

# nixos-generate-config --root /mnt
# nano /mnt/etc/nixos/configuration.nix # manually add mount options (see Compression below for an example)
# nixos-install

Configuration

Compression

nixos-generate-config --show-hardware-config doesn't detect mount options automatically, so to enable compression, you must specify it and other mount options in a persistent configuration:

fileSystems = {
  "/".options = [ "compress=zstd" ];
  "/home".options = [ "compress=zstd" ];
  "/nix".options = [ "compress=zstd" "noatime" ];
  "/swap".options = [ "noatime" ];
};

Swap file

Optionally, create a separate subvolume for the swap file. Be sure to regenerate your hardware-configuration.nix if you choose to do this.

# mkdir -p /mnt
# mount /dev/sdXY /mnt
# btrfs subvolume create /mnt/swap
# umount /mnt
# mkdir /swap
# mount -o noatime,subvol=swap /dev/sdXY /swap

Then, create the swap file and adjust its size as desired:

# btrfs filesystem mkswapfile --size 8g --uuid clear /swap/swapfile

Finally, add the swap file to your configuration and nixos-rebuild switch:

swapDevices = [ { device = "/swap/swapfile"; } ];

Scrubbing

Btrfs filesystems by default keep checksums for all files, to monitor if the file has changed due to hardware malfunctions or other external effects.

Scrubbing is the process of checking file consistency, which may use checksums and/or duplicated copies of data, from raid for example. Scrubbing may be done "online", meaning you don't need to unmount a subvolume to scrub it.

You can enable automatic scrubbing withː

services.btrfs.autoScrub.enable = true;

Automatic scrubbing by default is performed once a month, but you can change that withː

services.btrfs.autoScrub.interval = "weekly";

interval syntax is defined by systemd.timer's Calendar Events

By default, autoscrub will scrub all detected btrfs mount points. However, in case of mounted nested subvolumes (e.g. the example above where /nix and /home are nested subvolumes under /), you only need to scrub the topmost one. So an example configuration may look like this:

services.btrfs.autoScrub = {
  enable = true;
  interval = "monthly";
  fileSystems = [ "/" ];
};

The result of the periodic auto scrub will be saved to the system journal, and you can check the status of the last scrubː

btrfs scrub status /

You can also start a scrub in the background manuallyː

btrfs scrub start /

You can check the status of the ongoing scrubbing process with the same status command as above.

Deduplication

Files with (partially) equal contents can be deduplicated using bees or duperemove.

bees can be configured in configuration.nix:

services.beesd.filesystems = {
  root = {
    spec = "LABEL=root";
    hashTableSizeMB = 2048;
    verbosity = "crit";
    extraOptions = [ "--loadavg-target" "5.0" ];
  };
};

This will run the daemon in the background. To disable auto-start, use systemd.services."beesd@root".wantedBy = lib.mkForce [ ]; for each filesystem.

Usage

Subvolumes

Create a subvolume

btrfs subvolume create /mnt/nixos

Remove a subvolume

btrfs subvolume delete /mnt/nixos

Snapshots

A snapshot in btrfs is simply a subvolume that shares its data (and metadata) with some other subvolume, using btrfs's CoW capabilities. Because of that, there is no special location for snapshots and you can decide where you want to store them. It can be a simple directory inside the root subvolume, or a directory inside a dedicated "snapshots" subvolume.

For this example we are going to store snapshots in a directory /snapshots, that has to be created beforehand with sudo mkdir /snapshots

To take a read-only (-r) snapshot called home_snapshot_202302 of the subvolume mounted at /home

btrfs subvolume snapshot -r /home /snapshots/home_snapshot_202302

You can also snapshot the root subvolume. But keep in mind that nested subvolumes are not part of a snapshot. So if you have subvolumes /nix /home, taking a snapshot of / will not include them.

btrfs subvolume snapshot -r / /snapshots/nixos_snapshot_202302

Make snapshot read-write againː

btrfs property set -ts /snapshots/home_snapshot_202302 ro false

However, changing read-only property of a snapshot in-place may causes issues with any future incremental send/receive. Instead, a read-only snapshot itself (being a simple subvolume) can be snapshotted again as a read-write snapshot like this:

btrfs subvolume snapshot /snapshots/home_snapshot_202302 /snapshots/home_snapshot_202302_rw

Or it can be restored directly to /home straight away like this:

Warning: this will delete the current /home and restore the snapshot! /home must be unmounted for this operation
btrfs subvolume delete /home
btrfs subvolume snapshot /snapshots/home_snapshot_202302 /home

After this you can mount /home again.

Transfer snapshot

Sending the snapshot /snapshots/nixos_snapshot_202302 compressed to a remote host via ssh at root@192.168.178.110 and saving it to a subvolume mounted on a directory at /mnt/nixos

sudo btrfs send /snapshots/nixos_snapshot_202302 | zstd | ssh root@192.168.178.110 'zstd -d | btrfs receive /mnt/nixos'

Tips and tricks

Installation with encryption

Using Luks2:

cryptsetup --verify-passphrase -v luksFormat "$DISK"p2 

cryptsetup open "$DISK"p2 enc

You can use any device partition for your bootloader. Note that this bootloader is unencrypted by default:

mkfs.vfat -n BOOT "$DISK"p1

Creating subvolumes

mkfs.btrfs /dev/mapper/enc # Creating btrfs partition

mount -t btrfs /dev/mapper/enc /mnt

# Create the subvolumes 

btrfs subvolume create /mnt/root # The subvolume for /

btrfs subvolume create /mnt/home # The subvolume for /home, which should be backed up

btrfs subvolume create /mnt/nix # The subvolume for /nix, which needs to be persistent but is not worth backing up, as it’s trivial to reconstruct

btrfs subvolume create /mnt/log # The subvolume for /var/log.

Unmount to mount on the subvolumes for the next steps:

umount /mnt

Once the subvolumes have been created, mount them with the desired options.

Example with Zstandard compression and noatime:

mount -o subvol=root,compress=zstd,noatime /dev/mapper/enc /mnt 

mkdir /mnt/home

mount -o subvol=home,compress=zstd,noatime /dev/mapper/enc /mnt/home

mkdir /mnt/nix

mount -o subvol=nix,compress=zstd,noatime /dev/mapper/enc /mnt/nix

mkdir -p /mnt/var/log

mount -o subvol=log,compress=zstd,noatime /dev/mapper/enc /mnt/var/log

# do not forget to create and mount the bootloader

mkdir /mnt/boot

mount "$DISK"p1 /mnt/boot

Configure hardware-configuration.nix

 # enable btrfs support
 boot.supportedFilesystems = [ "btrfs" ];

 fileSystems."/var/log" =
    { device = "/dev/disk/by-uuid/X";
      fsType = "btrfs";
      # enable noatime and zstd to the other subvolumes aswell
      options = [ "subvol=log" "compress=zstd" "noatime" ];
      # to have a correct log order
      neededForBoot = true;
    };

Generate Nixconfig:

nixos-generate-config --root /mnt

Convert Ext3/Ext4 system partition to Btrfs

Note that migrating your existing root filesystem can cause data loss or make your system unbootable. Make sure to backup the partition or your files. Proceed only if you know what you're doing!

To convert the existing filesystem (Ext3/4) to Btrfs, boot into a NixOS live system and run the following commandː

fsck -f /dev/sdXY
btrfs-convert /dev/sdXY

Replace the device path with the target partition. Converting larger filesystems can take a long time. Next, mount the converted filesystem and chroot into itː

mount /dev/sdXY /mnt
nixos-enter --root /mnt

Replace the partition UUID with the new one, which can be obtained using the command blkid, in /etc/nixos/hardware-configuration.nix and also change the filesystem to btrfs.

  fileSystems."/" =
    { device = "/dev/disk/by-uuid/44444444-4444-4444-8888-888888888888";
      fsType = "btrfs";
    };

Apply the changes

nixos-rebuild boot

If the Grub bootloader is used and it doesn't get reinstalled correctly, you can run the following command inside chrootː

grub-install /dev/sdX

If the conversion was successful and no rollback is required, the backup image which was stored by btrfs-convert can be removed withː

btrfs subvolume delete /btrfs/ext2_saved