Bcachefs: Difference between revisions

From NixOS Wiki
Onny (talk | contribs)
Note about currently broken bcachefs mounting on NixOS, and add link to github issue.
Line 194: Line 194:
</syntaxhighlight>
</syntaxhighlight>
And using it like <code>UUID=<UUID></code> in place of <code>/dev/sda1:/dev/sdb1</code> or even just <code>/dev/sda</code>.
And using it like <code>UUID=<UUID></code> in place of <code>/dev/sda1:/dev/sdb1</code> or even just <code>/dev/sda</code>.
Note: this is currently broken as per this [https://github.com/NixOS/nixpkgs/issues/317901 github issue]. As a workaround, use `/dev/disk/by-uuid/<UUID1>:/dev/disk/by-uuid/<UUID2>` instead.


Continue installation as recommended by the [https://nixos.org/manual/nixos/stable/index.html#ch-installation NixOS manual].
Continue installation as recommended by the [https://nixos.org/manual/nixos/stable/index.html#ch-installation NixOS manual].

Revision as of 16:40, 14 September 2024

Bcachefs is a next-generation CoW filesystem that aims to provide features from Btrfs and ZFS with a cleaner codebase, more stability, greater speed and a GPL-compatible license. It is built upon Bcache and is mainly developed by Kent Overstreet.

Installation

To enable filesystem support and availability of user-space utils, add following line to the system configuration

/etc/nixos/configuration.nix
boot.supportedFilesystems = [ "bcachefs" ];

Usage

Format and mount a single device

# bcachefs format /dev/sda
# mount -t bcachefs /dev/sda /mnt

Or, format and mount multiple devices

# bcachefs format /dev/sda:/dev/sdb:/dev/sdc
# mount -t bcachefs /dev/sda:/dev/sdb:/dev/sdc

The same works with partitions, which is probably better for future proofing depending on your specific needs

# bcachefs format /dev/sda1:/dev/sdb2:/dev/sdc3
# mount -t bcachefs /dev/sda1:/dev/sdb2:/dev/sdc3


Format drive with encryption enabled, unlock and mount it afterwards. Following bcachefs commands will ask for a password:

# bcachefs format --encrypt /dev/sda
# bcachefs unlock /dev/sda
# mount -t bcachefs /dev/sda /mnt

Format a drive with compression on by default, foreground and background (Available Compression options are gzip, lz4, and zstd)

# bcachefs format --compression=lz4 --background_compression=zstd /dev/sda 
# mount -t bcachefs /dev/sda

Format a multiple devices with storage tiers, so that reads and writes happen on the fastest disks, with data being stored on slower, bigger drives based on usage patterns

# bcachefs format \
    --label=hdd.hdd1 /dev/sdc \
    --label=hdd.hdd2 /dev/sdd \
    --label=hdd.hdd3 /dev/sde \
    --label=ssd.ssd1 /dev/sdf \
    --label=ssd.ssd2 /dev/sdg
    --foreground_target=ssd \
    --promote_target=ssd \
    --background_target=hdd \
# mount -t bcachefs /dev/sdc:/dev/sdd:/dev/sde:/dev/sdf:/dev/sdg

For a better mounting experience in the previous example, use the external UUID that was printed.

# bcachefs format \
    --label=hdd.hdd1 /dev/sdc \
    --label=hdd.hdd2 /dev/sdd \
    --label=hdd.hdd3 /dev/sde \
    --label=ssd.ssd1 /dev/sdf \
    --label=ssd.ssd2 /dev/sdg
    --foreground_target=ssd \
    --promote_target=ssd \
    --background_target=hdd \
# mount -t bcachefs UUID=<UUID>

Create a subvolume of a mounted bcachefs filesystem. The snapshot of the filesystem state is accessible in the directory /mnt/snap1.

# bcachefs subvolume snapshot /mnt /mnt/snap1

Filesystem check, fix errors and corruptions where a Bcachefs filesystem is on /dev/sda:

# bcachefs fsck /dev/sda

Configuration

Every option for the filesystem can be set by editing /sys/fs/bcachefs/<uuid>/options, for example the file background_compression will change the background compression scheme for background compression. These are persisted with the filesystems, so a bcachefs storage device being mounted on a different computer won't need to know what mount options to use to maintain the same compression levels.

Change encryption password for Bcachefs formatted device /dev/sda1

# bcachefs set-passphrase /dev/sda1

Enable zstd compression for device /dev/sda1 at mount time

/etc/nixos/hardware-configuration.nix
fileSystems."/" =
{ device = "/dev/sda1";
  fsType = "bcachefs";
  options = [ "compression=zstd" ];
};

Tips and tricks

Generate bcachefs enabled installation media

Use following Nix Flake-expression to generate a ISO installation image with a bcachefs enabled kernel

flake.nix
{
  description = "Bcachefs enabled installation media";
  inputs.nixos.url = "nixpkgs/nixos-24.05";
  outputs = { self, nixos }: {
    nixosConfigurations = {
      exampleIso = nixos.lib.nixosSystem {
        system = "x86_64-linux";
        modules = [
          "${nixos}/nixos/modules/installer/cd-dvd/installation-cd-minimal-new-kernel-no-zfs.nix"
          ({ lib, pkgs, ... }: {
            # Might be required as a workaround for bug
            # https://github.com/NixOS/nixpkgs/issues/32279
            environment.systemPackages = [ pkgs.keyutils ];
            boot.supportedFilesystems = [ "bcachefs" ];
          })
        ];
      };
    };
  };
}

The following commands will generate the iso-image which will be available in the directory ./result/iso

# git init
# git add flake.nix
# nix build .#nixosConfigurations.exampleIso.config.system.build.isoImage

NixOS installation on bcachefs

Using the installation media generated above, continue the installation as usual following the instructions of the NixOS manual.

For a UEFI installation, the partitioning needs to be adjusted as following

Note: Be sure on which disk you'll perform these filesystem operations. All existing data on it will be erased.
# parted /dev/sda -- mklabel gpt
# parted /dev/sda -- mkpart ESP fat32 1MB 512MB
# parted /dev/sda -- set 1 esp on
# parted /dev/sda -- mkpart primary 512MB 100%

Formatting the boot partition /dev/sda1 and the root filesystem /dev/sda2

# mkfs.fat -F 32 -n boot /dev/sda1
# mkfs.bcachefs -L nixos /dev/sda2

In case you want to enable filesystem encryption, there's a workaround for a bug affecting NixOS 23.11. Formatting and unlocking the encrypted partition would look like this

# nix-env -iA nixos.keyutils
# keyctl link @u @s
# bcachefs format --encrypted /dev/sda2
# bcachefs unlock /dev/sda2

Mount filesystems

There are two ways to mount a bcachefs filesystem. First, by device:

# mount /dev/sda2 /mnt
# mkdir /mnt/boot
# mount /dev/disk/by-label/boot /mnt/boot

For better reliability as with any other filesystem, one should mount the filesystem by UUID, by replacing any drive specifiers with it. When you formatted, there was an external UUID shown, but if you lost it, you can retrieve it using

# lsblk -o +uuid,fsType | grep bcachefs

And using it like UUID=<UUID> in place of /dev/sda1:/dev/sdb1 or even just /dev/sda.

Note: this is currently broken as per this github issue. As a workaround, use `/dev/disk/by-uuid/<UUID1>:/dev/disk/by-uuid/<UUID2>` instead.

Continue installation as recommended by the NixOS manual.

Before nixos-install you'll have to adapt the generated NixOS configuration to also use the latest Linux kernel supporting Bcachefs. Add the following line to the configuration.nix in your installation root (/mnt):

/mnt/etc/nixos/configuration.nix
boot.supportedFilesystems = [ "bcachefs" ];
boot.kernelPackages = pkgs.linuxPackages_latest;

Remote encrypted disk unlocking

See article on remote disk unlocking for a guide on how to enable SSH decryption of Bcachefs enabled systems.

Automatically mount encrypted device on boot

Since the Bcachefs mount options do not support supplying a key file yet, we could use the bcachefs command and run it on boot using a Systemd unit:

systemd.services."bcachefs-mount" = {
  after = [ "local-fs.target" ];
  wantedBy = [ "multi-user.target" ];
  script = ''
    #!${pkgs.runtimeShell} -e
    ${pkgs.keyutils}/bin/keyctl link @u @s
    while [ ! -b /dev/sda1 ]; do
      echo "Waiting for /dev/sda1 to become available..."
      sleep 5
    done
    ${pkgs.bcachefs-tools}/bin/bcachefs mount -f /etc/keyfile_test /dev/sda1 /mnt
  '';
  serviceConfig = {
    Type = "oneshot";
    User = "root";
  };
};

This example unit mounts the Bcachefs encrypted partition /dev/sda1 to the target /mnt by using the key file /etc/keyfile_test.