Btrbk is a backup tool for btrfs subvolumes, taking advantage of btrfs specific capabilities to create atomic snapshots and transfer them incrementally to your backup locations.

The source and target locations are specified in a config file, which allows to easily configure simple scenarios like "laptop with locally attached backup disks", as well as more complex ones, e.g. "server receiving backups from several hosts via ssh, with different retention policies".

Key Features:

• Atomic snapshots
• Incremental backups
• Flexible retention policy
• Backups to multiple destinations
• Transfer via ssh
• Robust recovery from interrupted backups (for removable and mobile devices)
• Archive to offline storage
• Encrypted backups to non-btrfs storage
• Wildcard subvolumes (useful for docker and lxc containers)
• Transaction log
• Comprehensive list and statistics output
• Resolve and trace btrfs parent-child and received-from relationships
• List file changes between backups
• Calculate accurate disk space usage based on block regions

Btrbk is designed to run as a cron job for triggering periodic snapshots and backups, as well as from the command line (e.g. for instantly creating additional snapshots).

Btrbk is a single perl script, and does not require any special installation procedures or libraries. Download the latest btrbk source tarball, or try latest master:

wget https://raw.githubusercontent.com/digint/btrbk/master/btrbk
chmod +x btrbk
sudo ./btrbk ls /

For more information, read the installation documentation.

• btrfs-progs: Btrfs filesystem utilities >= v4.12
• Perl interpreter: Probably already installed on your system
• OpenSSH: If you want to transfer backups from/to remote locations
• mbuffer: If you want rate limiting and progress bars

Please consult the btrbk(1) man-page provided with this package for a full description of the command line options.

Before running btrbk, you will need to create a configuration file. You might want to take a look at btrbk.conf.example provided with this package. For a detailed description, please consult the btrbk.conf(5) man-page.

After a configuration change, it is highly recommended to check it by running btrbk with the -n,--dry-run option:

# btrbk -c /path/to/myconfig -v -n run

This will read all btrfs information on the source/target filesystems and show what actions would be performed (without writing anything to the disks).

The examples below assume that the btrfs subvolume containing home and rootfs is mounted at /mnt/btr_pool. This is usually the btrfs root subvolume, which always has subvolid=5.

Mounting subvolid=5 is recommended (mandatory for btrbk < v0.32.0) if you want to backup your root filesystem /.

/etc/fstab:

/dev/sda1  /mnt/btr_pool  btrfs  subvolid=5,noatime  0 0

Note that some default btrfs installations (e.g. Ubuntu) use subvolume names @ for rootfs (mounted at /) and @home for /home, as a naming convention. If this is the case on your file system, replace the subvolume declarations in the examples accordingly.

The simplest use case is to only create snapshots of your data. This will obviously not protect it against hardware failure, but can be useful for:

• protection against inadvertent changes or deletions
• keeping past states of copies from rsync or similar tools

Let's assume you need regular snapshots of your home directory, which is located in the subvolume home of the volume /mnt/btr_pool. The snapshots are to be stored in btrbk_snapshots (on the same volume).

/etc/btrbk/btrbk.conf:

timestamp_format        long
snapshot_preserve_min   18h
snapshot_preserve       48h

volume /mnt/btr_pool
  snapshot_dir btrbk_snapshots
  subvolume home

Notice that the target option is not provided, and btrbk will only manage snapshots located on the same volume in snapshot_dir. Btrbk does not create subdirs by default, the snapshot directory must first be created manually:

# mkdir /mnt/btr_pool/btrbk_snapshots

The "volume" section is merely used as a specifier for a base directory, and can be skipped if you prefer to configure everything using absolute paths. The above configuration can also be written as:

snapshot_dir /mnt/btr_pool/btrbk_snapshots
subvolume    /mnt/btr_pool/home

If you don't want to mount the btrfs root filesystem to /mnt/btr_pool, you might as well configure it like this:

snapshot_dir /btrbk_snapshots
subvolume    /home

Start a dry run (-n, --dry-run):

# btrbk run -n

Create the first snapshot:

# btrbk run

Print schedule (-S, --print-schedule):

# btrbk run -n -S

If it works as expected, configure a cron job to run btrbk hourly:

/etc/cron.hourly/btrbk:

#!/bin/sh
exec /usr/bin/btrbk -q run

Snapshots will now be created every hour. All snapshots are preserved for at least 18 hours (snapshot_preserve_min), whether they are created by the cron job or manually by calling sudo btrbk run on the command line. Additionally, 48 hourly snapshots are preserved (snapshot_preserve).

In this example, we assume you have a laptop with:

• a disk having a btrfs root subvolume (subvolid=5) mounted on /mnt/btr_pool, containing a subvolume rootfs for the root filesystem (i.e. mounted on /) and a subvolume home for the user data,
• a directory or subvolume /mnt/btr_pool/btrbk_snapshots which will hold the btrbk snapshots,
• a backup disk having a btrfs volume mounted as /mnt/btr_backup, containing a subvolume or directory mylaptop for the incremental backups.

Retention policy:

• keep all snapshots for 2 days, no matter how frequently you (or your cron job) run btrbk
• keep daily snapshots for 14 days (very handy if you are on the road and the backup disk is not attached)
• keep monthly backups forever
• keep weekly backups for 10 weeks
• keep daily backups for 20 days

/etc/btrbk/btrbk-mylaptop.conf:

snapshot_preserve_min   2d
snapshot_preserve      14d

# Create snapshots only if the backup disk is attached
#snapshot_create ondemand

target_preserve_min    no
target_preserve        20d 10w *m

snapshot_dir           btrbk_snapshots

volume /mnt/btr_pool
  target /mnt/btr_backup/mylaptop
  subvolume rootfs
  subvolume home
  [...]

/etc/cron.daily/btrbk:

#!/bin/sh
exec /usr/bin/btrbk -q -c /etc/btrbk/btrbk-mylaptop.conf run

• This will create snapshots on a daily basis:
  • /mnt/btr_pool/btrbk_snapshots/rootfs.YYYYMMDD
  • /mnt/btr_pool/btrbk_snapshots/home.YYYYMMDD
• And create incremental backups in:
  • /mnt/btr_backup/mylaptop/rootfs.YYYYMMDD
  • /mnt/btr_backup/mylaptop/home.YYYYMMDD

If you prefer triggering the backups manually, change the cron command to run the snapshot action instead of run. Start the backups manually by running:

# btrbk resume

For a quick additional snapshot of your home, run:

# btrbk snapshot home

Let's say you have a fileserver at "myserver.example.org" where you want to create backups of your laptop disk. The config could look like this:

ssh_identity /etc/btrbk/ssh/id_rsa

volume /mnt/btr_pool
  subvolume rootfs
    target /mnt/btr_backup/mylaptop
    target ssh://myserver.example.org/mnt/btr_backup/mylaptop

In addition to the backups on your local usb-disk mounted at /mnt/btr_backup/mylaptop, incremental backups would also be pushed to myserver.example.org.

If you're a sysadmin and want to trigger backups directly from your fileserver, the config would be something like:

ssh_identity /etc/btrbk/ssh/id_rsa

volume ssh://alpha.example.org/mnt/btr_pool
  target /mnt/btr_backup/alpha
  subvolume rootfs
  subvolume home

volume ssh://beta.example.org/mnt/btr_pool
  target /mnt/btr_backup/beta
  subvolume rootfs
  subvolume dbdata

This will pull backups from alpha/beta.example.org and locally create:

• /mnt/btr_backup/alpha/rootfs.YYYYMMDD
• /mnt/btr_backup/alpha/home.YYYYMMDD
• /mnt/btr_backup/beta/rootfs.YYYYMMDD
• /mnt/btr_backup/beta/dbdata.YYYYMMDD

Let's say we have a host (at 192.168.0.42) running btrbk with the setup of the time-machine example above, and we need a backup server to only fetch the snapshots.

/etc/btrbk/btrbk.conf (on backup server):

target_preserve_min        no
target_preserve            0d 10w *m

volume ssh://192.168.0.42/mnt/btr_pool
  target /mnt/btr_backup/my-laptop
  subvolume home
    snapshot_dir           btrbk_snapshots
    snapshot_preserve_min  all
    snapshot_create        no

If the server runs btrbk with this config, 10 weeklies and all monthlies are received from 192.168.0.42. The source filesystem is never altered because of snapshot_preserve_min all.

Common virtual machine setups have multiple volume sections with same host, but distinct port numbers for each machine.

/etc/btrbk/btrbk.conf:

# This propagates to all subvolume sections:
target /mnt/btr_backup/

volume ssh://localhost:2201/mnt/btr_pool
  group vm vm01
  subvolume home
    snapshot_name vm01-home
  subvolume data
    snapshot_name vm01-data

volume ssh://localhost:2202/mnt/btr_pool
  group vm vm02
  subvolume home
    snapshot_name vm02-home

volume ssh://localhost:2203/mnt/btr_pool
  [...]

This will create /mnt/btr_backup/vm[NN]-home, vm[NN]-data, ...

Note that btrbk holds a single reference to every btrfs filesystem tree, regarding UUID's as "globally unique". If the configured subvolumes point to the same filesystem on different machines (ports), you will see log lines like this when running btrbk -v:

Assuming same filesystem: "ssh://localhost:2201/dev/sda1", "ssh://localhost:2202/dev/sda1"

If you want to make backups from a filesystem other than btrfs (e.g. ext4 or reiserfs), you need to create a synchronization subvolume on the backup disk:

# btrfs subvolume create /mnt/btr_backup/myhost_sync

Configure btrbk to use myhost_sync as source subvolume:

volume /mnt/btr_backup
  subvolume myhost_sync
    snapshot_name           myhost

    snapshot_preserve_min   latest
    snapshot_preserve       14d 20w *m

The btrbk package provides the "btrbk-mail" script, which automates the synchronization using rsync, and can be run as cron job or systemd timer unit. For configuration details, see the config section in "/contrib/cron/btrbk-mail".

Alternatively, you can run any synchronization software prior to running btrbk. Something like:

#!/bin/sh
rsync -az --delete \
      --inplace --numeric-ids --acls --xattrs \
      -e 'ssh -i /etc/btrbk/ssh/id_rsa' \
      myhost.example.org:/data/ \
      /mnt/btr_backup/myhost_sync/

exec /usr/bin/btrbk -q run

This will produce snapshots /mnt/btr_backup/myhost.20150101, with retention as defined with the snapshot_preserve option.

If your backup server does not support btrfs, you can send your subvolumes to a raw file.

This is an experimental feature: btrbk supports "raw" targets, meaning that similar to the "send-receive" target the btrfs subvolume is being sent using btrfs send (mirroring filesystem level data), but instead of instantly being received (btrfs receive) by the target filesystem, it is being redirected to a file, optionally compressed and piped through GnuPG.

/etc/btrbk/btrbk.conf:

raw_target_compress   xz
raw_target_encrypt    gpg
gpg_keyring           /etc/btrbk/gpg/pubring.gpg
gpg_recipient         btrbk@example.org

volume /mnt/btr_pool
  subvolume home
    target raw ssh://cloud.example.com/backup
      ssh_user  btrbk
      # incremental  no

This will create a GnuPG encrypted, compressed files on the target host. For each backup, two files are created:

• /backup/home.YYYYMMDD.btrfs.xz.gpg: main data file containing the btrfs send-stream,
• /backup/home.YYYYMMDD.btrfs.xz.gpg.info: sidecar file containing metadata used by btrbk.

If you are using raw incremental backups, please make sure you understand the implications (see btrbk.conf(5), TARGET TYPES).

Since btrbk needs root access, it is very advisable to take all the security precautions you can. In most cases backups are generated periodically without user interaction, so it is not possible to protect your ssh key with a password. The steps below will give you hints on how to secure your ssh server for a backup scenario. Note that the btrbk package is not required on the remote side, but you will need the btrfs executable from the btrfs-progs package.

On the client side, create a ssh key dedicated to btrbk, without password protection:

# ssh-keygen -t rsa -b 4096 -f /etc/btrbk/ssh/id_rsa -C btrbk@example.org -N ""

The content of the public key (/etc/btrbk/ssh/id_rsa.pub) is used for authentication in "authorized_keys" on the server side (see sshd(8) for details).

The most straight forward setup is to allow root login on the remote host. If this is not an option for you, refer to the more complex "Dedicated Btrbk User Login" section below.

/etc/ssh/sshd_config:

PermitRootLogin prohibit-password

Add your btrbk public key to "/root/.ssh/authorized_keys" on the server, and you are good to go.

Restrict ssh access to a static IP address within your network. On the remote host, either add a "Match" block in:

/etc/ssh/sshd_config:

Match Address 192.168.0.42

Or restrict in authorized_keys:

from="192.168.0.42" <pubkey>...

Consult the sshd_config(5) man-page for a detailed explanation and more options.

If allowing root login is not an option for you, there are several ways to restrict SSH access to a regular user.

On the client side, configure btrbk use the sudo backend. This changes the ssh calls to btrfs commands to sudo btrfs <subcommand> <options>.

/etc/btrbk/btrbk.conf:

backend_remote btrfs-progs-sudo

On the remote host, grant root permissions for the "btrfs" command groups (subcommands) in "/etc/sudoers". If you are using ssh_filter_btrbk(1), also add the ssh_filter_btrbk.sh --sudo option in "authorized_keys" (see below).

Instead of using the all-inclusive btrfs command, "btrfs-progs-btrbk" allows you to restrict privileges to its subcommands using linux capabilities(7) or setuid.

Note that the "btrfs-progs-btrbk" package is not available on all linux distributions, you might need to build and install it on your own (refer to btrfs-progs-btrbk on GitHub for more details).

/etc/btrbk/btrbk.conf:

backend_remote btrfs-progs-btrbk

Make sure that only the required binaries with elevated privileges can be called by the btrbk user. For example, on a server acting as "btrbk source", allow only the following binaries for the "btrbk" group:

# getcap /usr/bin/btrfs-*
/usr/bin/btrfs-send cap_dac_read_search,cap_fowner,cap_sys_admin=ep
/usr/bin/btrfs-subvolume-delete cap_dac_override,cap_sys_admin=ep
/usr/bin/btrfs-subvolume-list cap_dac_read_search,cap_fowner,cap_sys_admin=ep
/usr/bin/btrfs-subvolume-show cap_dac_read_search,cap_fowner,cap_sys_admin=ep
/usr/bin/btrfs-subvolume-snapshot cap_dac_override,cap_dac_read_search,cap_fowner,cap_sys_admin=ep

# ls -l /usr/bin/btrfs-*
-rwx--x--- 1 root btrbk  /usr/bin/btrfs-send
-rwx--x--- 1 root btrbk  /usr/bin/btrfs-subvolume-delete
-rwx--x--- 1 root btrbk  /usr/bin/btrfs-subvolume-list
-rwx--x--- 1 root btrbk  /usr/bin/btrfs-subvolume-show
-rwx--x--- 1 root btrbk  /usr/bin/btrfs-subvolume-snapshot

Btrbk comes with a shell script "ssh_filter_btrbk.sh", which restricts ssh access to sane calls to the "btrfs" command needed for snapshot creation and send/receive operations (see ssh_filter_btrbk(1)).

Copy "ssh_filter_btrbk.sh" to "/backup/scripts/", and configure sshd to run it whenever the key is used for authentication. Example "/root/.ssh/authorized_keys":

# example backup source (also allowing deletion of old snapshots)
command="/backup/scripts/ssh_filter_btrbk.sh -l --source --delete",restrict <pubkey>...

# example backup target (also allowing deletion of old snapshots)
command="/backup/scripts/ssh_filter_btrbk.sh -l --target --delete",restrict <pubkey>...

# example fetch-only backup source (snapshot_preserve_min=all, snapshot_create=no),
# restricted to subvolumes within /home or /data
command="/backup/scripts/ssh_filter_btrbk.sh -l --send -p /home -p /data",restrict <pubkey>...

Btrbk does not provide any mechanism to restore your backups, this has to be done manually. In the instructions below, we assume that you have a btrfs volume mounted at /mnt/btr_pool, and the subvolume you want to restore is at /mnt/btr_pool/data.

Important: don't use btrfs property set to make a subvolume read-write after restoring. This is a low-level command, and leaves "Received UUID" in a false state which causes btrbk to fail on subsequent incremental backups. Instead, use btrfs subvolume snapshot (without -r flag) as described below.

# list snapshots managed by btrbk
btrbk list snapshots

# alternative: list all subvolumes
btrbk ls /
btrbk ls -L /

From the list, identify the snapshot you want to restore. Let's say it's /mnt/btr_pool/_btrbk_snap/data.20150101.

(skip this step if you restore from a snapshot)

# locally mounted backup disk
btrfs send /mnt/btr_backup/data.20150101 | btrfs receive /mnt/btr_pool/

# from / to remote host
ssh root@remote btrfs send /mnt/btr_backup/data.20150101 | btrfs receive /mnt/btr_pool/
btrfs send /mnt/btr_backup/data.20150101 | ssh root@remote btrfs receive /mnt/btr_pool/

Hint: Try to send-receive backups incrementally if possible. In case you still have common snapshot / backup pairs (i.e. both "snapshot_subvol" and "target_subvol" are listed above), use btrfs send -p <parent>.

From this point on, data.20150101 on both disks can be used as parents for subsequent send-receive operations, and a received_uuid relationship is established (see below).

# if still present, move broken subvolume away
mv /mnt/btr_pool/data /mnt/btr_pool/data.BROKEN

# create read-write subvolume
btrfs subvolume snapshot /mnt/btr_pool/data.20150101 /mnt/btr_pool/data

Your data subvolume is restored, you can carry on with incremental backups to /mnt/btr_backup.

# if everything went fine, delete the broken subvolume
btrfs subvolume delete /mnt/btr_pool/data.BROKEN

Make sure to keep data.20150101 subvolumes on both disks at least until you created a new backup using btrbk, in order to keep the incremental chain alive.

btrbk origin -t /mnt/btr_backup/data.20150101
btrbk ls -L /mnt/btr_pool /mnt/btr_backup

• received_uuid relationship: correlated, identical read-only subvolumes, cross-filesystem.

  a.received_uuid = b.received_uuid
  a.received_uuid = b.uuid
  

  • Required for subvolumes used as parent (or clone-src) of send-receive operations.
  • Present on subvolumes created by btrfs send | btrfs receive.
  • /mnt/btr_pool/data.20150101 === /mnt/btr_backup/data.20150101

• parent_uuid relationship: "is-snapshot-of"

  a.parent_uuid = b.uuid
  

  • Present on subvolumes created by btrfs subvolume snapshot or btrfs send -p | btrfs receive.
  • Used by btrbk to determine best parent.
  • /mnt/btr_pool/data.20150101 <-- /mnt/btr_pool/data

Make sure to also read the btrbk FAQ page. Help improve it by asking!

So btrbk saved your day?

I will definitively continue to develop btrbk for free. If you want to support my hard work with a donation, you are welcome to do so!

The source code for btrbk is managed using Git.

Official repository:

git clone https://dev.tty0.ch/btrbk.git

Mirror on GitHub:

git clone https://github.com/digint/btrbk.git

Your contributions are welcome!

If you would like to contribute or have found bugs:

• Visit the btrbk project page on GitHub and use the issues tracker there.
• Talk to us on Libera.Chat in #btrbk.
• Contact the author via email (the email address can be found in the sources).

Any feedback is appreciated!

btrbk is free software, available under the GNU General Public License, Version 3 or later.