btrbk.conf Man Page

btrbk.conf(5),version 0.32.6,2023-03-25

NAME

btrbk.conf - btrbk configuration file

SYNOPSIS

/etc/btrbk.conf
/etc/btrbk/btrbk.conf

DESCRIPTION

The btrbk configuration file specifies which btrfs subvolumes on the filesystem are to be processed, what target subvolumes should be used to create the backups, and where the snapshots should be generated. The retention policy, as well as most other options can be defined either globally or within a section.

The options specified always apply to the last section encountered, superseding the values set in upper-level sections. This means that global options must be set before any sections are defined.

Blank lines are ignored. A hash character (#) starts a comment extending until end of line.

SECTIONS

volume <volume-directory>|<url> (optional): Absolute path pointing to a btrfs file system containing the source subvolume(s) to be backed up. Usually the mount point of a btrfs filesystem mounted with the subvolid=5 option.
subvolume <subvolume-name>: Subvolume to be backed up, relative to the <volume-directory> of the volume section, or absolute if the volume section is omitted. Accepts wildcard character "*".

Note that if this subvolume is btrfs root (id=5), it needs to have a valid UUID, which is not the case for file systems created with btrfs-progs < 4.16.
target [send-receive|raw] <target-directory>|<url>: Target directory where the backup subvolumes are to be created. The optional target type defaults to “send-receive”, see TARGET TYPES below for details.

Multiple target sections are allowed, in any context: a target defined in volume or global context will be used for all underlying subvolume sections (hint: run "btrbk list" or "btrbk config print" to see the resulting configuration).

If a <url> is specified, btrbk actions (shell commands) are executed remotely via ssh, using the SSH Options described below. Accepted formats are:

ssh://<hostname>[:<port>]/<directory>
<hostname>:<directory>

Where <hostname> is either a host name, an IPv4 address in dotted-decimal form, or an IP literal encapsulated within square brackets (e.g. "[2001:db8::7]").

If you are connecting to virtual machines, consider configuring several volume sections for a <hostname>, with distinct <port> numbers for each machine.

OPTIONS

The options described here can be specified in global context as well as volume, subvolume and target sections, unless stated otherwise.

Basic Options

timestamp_format short|long|long-iso

Timestamp format used as postfix for new snapshot subvolume names. Defaults to “long”.

short	`YYYYMMDD[_N]` (e.g. "20150825", "20150825_1")
long	`YYYYMMDD<T>hhmm[_N]` (e.g. "20150825T1531")
long-iso	`YYYYMMDD<T>hhmmss±hhmm[_N]` (e.g. "20150825T153123+0200")

Note that a postfix "_N" is appended to the timestamp if a snapshot or backup already exists with the timestamp of current date/time.

Use “long-iso” if you want to make sure that btrbk never creates ambiguous time stamps (which can happen if multiple snapshots are created during a daylight saving time clock change).

Note that using “long-iso” has implications on the scheduling, see Reference Time below.

snapshot_dir <directory>

Directory in which the btrfs snapshots are created, relative to <volume-directory> of the volume section, or absolute if the volume section is omitted. Note that btrbk does not automatically create this directory, and the snapshot creation will fail if it is not present.

snapshot_name <basename>

Base name of the created snapshot (and backup). This option is only valid in the subvolume section. Defaults to <subvolume-name>.

snapshot_create always|onchange|ondemand|no

If set to “always”, snapshots are always created. If set to “onchange”, snapshots are only created if the last snapshot is not up-to-date, i.e. the source subvolume has changed (more precisely: the btrfs generation has been increased) since the last snapshot was created. If set to “ondemand”, snapshots are only created if at least one target subvolume is reachable (useful if you are tight on disk space and you only need btrbk for backups to an external disk which is not always connected). If set to “no”, the snapshots are never created (useful if another instance of btrbk is taking care of snapshot creation). Defaults to “always”.

incremental yes|no|strict

If set, incremental backups are created. If set to “strict”, non-incremental (initial) backups are never created, and incremental backups are restricted to related parents (by parent-uuid relationship). Defaults to “yes”.

Note that even if the parent-uuid chain is broken, snapshots and backups can still share data (which is especially true for backups created with incremental option enabled), and are perfectly suitable as parents for incremental send-receive operations. But as btrbk can not be certain about this, such operations are disallowed in "incremental strict" mode.

noauto yes|no

If set, the context is skipped by all btrbk actions unless explicitly enabled by a matching btrbk <filter> command line argument (e.g. "btrbk run myfilter").

Grouping Options

group <group-name> [<group-name>]…: Add the current section (volume, subvolume or target) to user-defined groups, which can be used as filter for most btrbk commands (see btrbk(1) FILTER STATEMENTS). This option can be set multiple times within the same context.

Retention Policy Options

preserve_day_of_week monday|tuesday|…|sunday: Defines on what day a snapshot/backup is considered to be a "weekly" backup. Weekly, monthly and yearly backups are preserved on this day of week (see RETENTION POLICY below). Defaults to “sunday”.
preserve_hour_of_day [0..23]: Defines after what time (in full hours since midnight) a snapshot/backup is considered to be a "daily" backup. Daily, weekly, monthly and yearly backups are preserved on this hour (see RETENTION POLICY below). Ignored on snapshots or backups without time information (timestamp_format short). Defaults to “0”.
snapshot_preserve no|<retention_policy>: Set retention policy for snapshots (see RETENTION POLICY below). If set to “no”, preserve snapshots according to snapshot_preserve_min only. Defaults to “no”.

Note that snapshot_preserve has no effect if snapshot_preserve_min is set to “all” (the default).
snapshot_preserve_min all|latest|<number>{h,d,w,m,y}: Preserve all snapshots for a minimum amount of hours (h), days (d), weeks (w), months (m) or years (y), regardless of how many there are. If set to “all”, preserve all snapshots forever. If set to “latest”, preserve latest snapshot. Defaults to “all”.
target_preserve no|<retention_policy>: Set retention policy for backups (see RETENTION POLICY below). If set to “no”, preserve backups according to target_preserve_min only. Defaults to “no”.

Note that target_preserve has no effect if target_preserve_min is set to “all” (the default).
target_preserve_min all|latest|no|<number>{h,d,w,m,y}: Preserve all backups for a minimum amount of hours (h), days (d), weeks (w), months (m) or years (y), regardless of how many there are. If set to “all”, preserve all backups forever. If set to “latest”, always preserve the latest backup (useful in conjunction with "target_preserve no", if you want to keep the latest backup only). If set to “no”, only the backups following the target_preserve policy are created. Defaults to “all”.
archive_preserve no|<retention_policy>: Set retention policy for archives ("btrbk archive" command), with same semantics as target_preserve.
archive_preserve_min all|latest|no|<number>{h,d,w,m,y}: Set retention policy for archives ("btrbk archive" command), with same semantics as target_preserve_min.
archive_exclude <pattern>: Exclude subvolumes matching <pattern> from archiving. The pattern accepts wildcard character "*", and is matched against the end of the pathname.

SSH Options

ssh_identity <file>|no: Absolute path to a ssh identity file (private key). If not set, the ssh default is used (see ssh(1), "-i identity_file"). Note that if the identity key is password protected and no authentication agent is used, btrbk will prompt for user input on every connection attempt.
ssh_user <username>|no: Remote username for ssh. Defaults to “root”. Make sure the remote user is able to run "btrfs" with root privileges (see option backend for details). If set to “no”, the ssh default is used.
ssh_compression yes|no: Enables or disables the compression of ssh connections. Defaults to “no”. Note that if stream_compress is enabled, ssh compression will always be disabled for send/receive operations.
ssh_cipher_spec default|<cipher_spec>: Selects the cipher specification for encrypting the session (comma-separated list of ciphers in order of preference). See the "-c cipher_spec" option in ssh(1) for more information. Defaults to “default” (the ciphers specified in ssh_config(5)).

Data Stream Options

stream_compress <compress_command>|no

Compress the btrfs send stream before transferring it from/to remote locations. Defaults to “no”. If enabled, make sure that <compress_command> is available on the source and target hosts. Supported <compress_command>: gzip, pigz, bzip2, pbzip2, bzip3, xz, lzo, lz4, zstd.

stream_compress_level default|<number>

Compression level for the specified <compress_command>. Refer to the related man-page for details (usually [1..9], where 1 means fastest compression). Defaults to “default” (the default compression level of <compress_command>).

stream_compress_long default|<number>

Enable long distance matching for the specified <compress_command>. Refer to the related man-page for details. Only supported for "zstd".

stream_compress_threads default|<number>

Number of threads to use for <compress_command>. Only supported for "pigz", "pbzip2", "bzip3", "zstd" and recent versions of "xz".

stream_compress_adapt default|<number>

Enable adaptive compression for <compress_command>. Only supported for "zstd" (version >= 1.3.6).

stream_buffer <size>|no

Add a buffer to the btrfs send stream (locally, on uncompressed data), with a maximum size of <size>. This can give a speed improvement (measured up to 20%) on both local or remote operations, but also increases system load. A suffix of "k", "m", "g", or "%" can be added to <size> to denote kilobytes (*1024), megabytes, gigabytes, or a percentage of total physical memory. Defaults to “no”.

If enabled, make sure that the "mbuffer" command (at least version 20180505) is available on the host running btrbk. As of btrbk-0.29.0, mbuffer(1) is used for both rate_limit and stream_buffer options:

mbuffer [-m <stream_buffer>] [-r <rate_limit>]

Note that mbuffer(1) always reads defaults from "/etc/mbuffer.rc" and "~/.mbuffer.rc".

Leave this option disabled if your main concern is a stable backup process: while recent versions of mbuffer have proven reliable, it is often desirable to keep things simple rather than adding an additional, multi-threaded process to the command pipe.

stream_buffer_remote <size>|no

Add a buffer on remote hosts (either source or target). Defaults to “no”.

Enable this if you prefer buffering on the remote side, or even on both sides: reasons for this depend on available memory, disk and cpu performance (btrfs send/receive, compression), as well as networking constraints.

rate_limit <rate>|no

Limit the read rate of the btrfs send stream to <rate> bytes per second (locally, on uncompressed send stream). A suffix of "k", "m", "g", or "t" can be added to denote kilobytes (*1024), megabytes, and so on. Defaults to “no”. Note that rate_limit implicitly adds a stream buffer (see stream_buffer option above).

rate_limit_remote <rate>|no

Add rate limit on remote hosts (either source or target). Defaults to “no”. Note that it usually does not make much sense to enable both rate_limit and rate_limit_remote.

System Options

transaction_log <file>|no

If set, all transactions (snapshot create, subvolume send-receive, subvolume delete) as well as abort messages are logged to <file>, in a space-separated table format: "localtime type status target_url source_url parent_url message".

transaction_syslog <facility>|no

If set, all transactions (as described in transaction_log above) are logged to syslog. The program name used in the messages is "btrbk". Accepted parameters for <facility>: user, mail, daemon, auth, lpr, news, cron, authpriv, local0..local7.

lockfile <file>|no

Create lockfile <file> on startup; checks lockfile before running any btrfs commands (using perl "flock"), and exits if the lock is held by another btrbk instance. Ignored on dryrun (-n, --dry-run). See also --lockfile command-line option.

backend <backend>

Backend filesystem utilities to be used for btrfs specific operations. Available backends:

btrfs-progs: Default backend, btrfs commands are called as specified in btrfs(8) (e.g. "btrfs subvolume show").
btrfs-progs-btrbk: btrfs commands are separated by a dash instead of a whitespace (e.g. "btrfs-subvolume-show" instead of "btrfs subvolume show"). Useful for setting suid or file capabilities (setcap) on specific btrfs commands, as implemented in https://github.com/digint/btrfs-progs-btrbk.
btrfs-progs-sudo: btrfs commands are prefixed with "sudo -n" (e.g. "sudo -n btrfs subvolume show" instead of "btrfs subvolume show"). Make sure to have appropriate (root) permissions for the "btrfs" command groups as well as the "readlink" and "test" commands in /etc/sudoers.
btrfs-progs-doas: Similar to btrfs-progs-sudo, using prefix "doas -n".

If you want to set this option for local or remote hosts only, set backend_local or backend_remote (e.g. "backend_remote btrfs-progs-btrbk").

If you want to set this option for regular (non-root) user only, set backend_local_user.

compat <compat-option>…

Enable compatibility options. Available compat-option:

busybox: Use busybox compatible commands, at the expense of slight overhead while reading filesystem information.
ignore_receive_errors *experimental*: Tell btrfs-receive(8) to not terminate on errors by setting "--max-errors=0" option. Print warnings instead.

A known use case for this are target hosts lacking xattr support (e.g. some Synology NAS), while the send-stream contains "lsetxattr" commands. Another case is targets failing to set otime, complaining with "ERROR: attribute 12 requested but not present".

Note that there is no guarantee that backups created with this option enabled can be restored at all.

If you want to set this option for local or remote hosts only, set compat_local or compat_remote (e.g. "compat_remote busybox").

cache_dir <directory>

If set, cache extent maps for the "btrbk extents" command.

Btrfs Specific Options

incremental_prefs <list-spec>[:<amount>]…

Specify the preferences to determine the best common (correlated) parent and clone sources for incremental backups, by choosing from predefined candidate lists.

The list-spec defines from what candidate list the next parent/clone-src should be appended to a result list; amount defines how many (e.g. "sro:1 sro:1" is identical to "sro:2"), or all if omitted. Any candidate which is already in the results is dropped.

The resulting list of subvolumes is then used as parameters for the btrfs-send(8) command: the first for "-p <parent>", all others for "-c <clone-src>".

Available list-spec (candidate lists = filtered subsets of correlated subvolumes):

sro,srn: All from snapshot_dir matching snapshot_name, with parent_uuid relationship, sorted by btrbk timestamp (o=older n=newer).
sao,san: All from snapshot_dir matching snapshot_name, sorted by btrbk timestamp (o=older n=newer).
aro,arn: All from incremental_resolve, with parent_uuid relationship, sorted by cgen (o=older n=newer).

Defaults to "sro:1 srn:1 sao:1 san:1 aro:1 arn:1". Note that for most operations the default resolves a single parent, as there usually are no newer snapshots, and all "sro:1 sao:1 aro:1" resolve to the same snapshot.

Example: "defaults,sao,san,aro,arn" takes the defaults, and adds clone sources for all (!) known candidates on the filesystem.

incremental_clones yes|no

If enabled, btrbk adds "-c <clone-src>" to the btrfs-send(8) command for all correlated subvolumes resolved by incremental_prefs. If disabled, only "-p <parent>" is used. Defaults to “yes”.

incremental_resolve mountpoint|directory

Specifies where to search for the best common parent for incremental backups. If set to “mountpoint”, use parents in the filesystem tree below the mount point of the snapshot and target directory. If set to “directory”, use parents strictly below snapshot/target directories. Set this to “directory” if you get access problems (when not running btrbk as root). Defaults to “mountpoint”.

btrfs_commit_delete yes|no

If set, wait for the transaction commit at the end of each snapshot or backup deletion (sets --commit-each option for "btrfs subvolume delete"). Defaults to “no”.

send_protocol <number>|no *experimental*

Use btrfs send protocol version N. If enabled on target, btrbk adds "--proto <number>" to the btrfs-send(8) command. Defaults to “no” (btrfs default).

send_compressed_data yes|no *experimental*

Send data that is compressed on the filesystem directly without decompressing it. This requires protocol version 2 or higher (btrfs-progs >= 5.19), and implies "send_protocol 2". If enabled on target, btrbk adds "--compressed-data" to the btrfs-send(8) command. Defaults to “no” (btrfs default).

snapshot_qgroup_destroy yes|no *experimental*

target_qgroup_destroy yes|no *experimental*

archive_qgroup_destroy yes|no *experimental*

Whenever a subvolume is deleted, also destroy corresponding default qgroup "0/<subvol-id>". Only useful if you have enabled btrfs quota support. See also: https://bugzilla.kernel.org/show_bug.cgi?id=91751

Informative Options

warn_unknown_targets yes|no: If set, prints a warning if btrbk encounters a target subvolume at a unknown location (i.e. not following btrbk naming scheme, or outside the target directory). Defaults to “no”.

RETENTION POLICY

Retention policies are defined individually for snapshots, backups and archives (summarized as "backups" in the following text), using a combination of:

*_preserve_min all|latest|no|<number>{h,d,w,m,y}: Amount of time (duration) in which all backups are preserved.
*_preserve no|<retention_policy>: Schedule (single points in time) for which individual backups are preserved.

Note that if "preserve_min" is set to “all” (the default), any setting of "preserve" obviously has no effect.

The format for <retention_policy> is:

[<hourly>h] [<daily>d] [<weekly>w] [<monthly>m] [<yearly>y]

hourly: Defines how many hours back hourly backups should be preserved. The first backup of an hour is considered an hourly backup.
daily: Defines how many days back daily backups should be preserved. The first backup of a day (starting at preserve_hour_of_day) is considered a daily backup.
weekly: Defines how many weeks back weekly backups should be preserved. The first daily backup created at preserve_day_of_week (or the first backup in this week if none was made on the exact day) is considered as a weekly backup.
monthly: Defines how many months back monthly backups should be preserved. Every first weekly backup in a month is considered a monthly backup.
yearly: Defines for how many years back yearly backups should be preserved. Every first monthly backup in a year is considered a yearly backup.

Use an asterisk for “all” (e.g. "target_preserve 60d *m" states: "preserve daily backups for 60 days back, and all monthly backups").

Hint: Run btrbk with the -S, --print-schedule option to get a comprehensive output of the scheduler results.

Reference Time

The local time on the host running btrbk defines the reference time for all date/time calculations, especially for "beginning of a day", and as a consequence for the first daily, weekly, monthly or yearly backups. The local time on remote hosts (ssh source/target) is never used.

Unless "timestamp_format long-iso" is set, daily backups are preserved at "preserve_hour_of_day" (defaults to midnight) of the respective time zone (and not for "00:00 UTC", which would be "14:00" in Honolulu). This becomes relevant for setups with multiple btrbk instances, e.g. many snapshot-only instances (spread around the world), and a fetch-only instance on the backup server.

Caveat:

If "timestamp_format long-iso" is set, each btrbk instance on has a different interpretation of "first in day". Make sure to run btrbk with the same time zone on every host, e.g. by setting the TZ environment variable (see tzset(3)).

TARGET TYPES

send-receive

Backup to a btrfs filesystem, using "btrfs send/receive". This is the recommended (standard) target type. The <target-directory> must be an absolute path and point to a subvolume or directory within a btrfs file system. See btrfs-send(8), btrfs-receive(8).

raw *experimental*

Backup to a raw (filesystem independent) file from the output of btrfs-send(8), with optional compression and encryption.

Note that the target preserve mechanism is currently disabled for incremental raw backups (btrbk does not delete any incremental raw files)!

Raw backups consist of two files: the main data file containing the btrfs send stream, and a sidecar file ".info" containing metadata:

<snapshot-name>.<timestamp>[_N].btrfs[.gz|.bz2|...][.gpg]
<snapshot-name>.<timestamp>[_N].btrfs[.gz|.bz2|...][.gpg].info

For incremental backups ("incremental yes"), please note that:

As soon as a single incremental backup file is lost or corrupted, all later incremental backups become invalid, as there is no common parent for the subsequent incremental images anymore. This might be a good compromise for a vacation backup plan, but for the long term make sure that a non-incremental backup is triggered from time to time.
There is currently no support for rotation of incremental backups: if incremental is set, a full backup must be triggered manually from time to time in order to be able to delete old backups.

Additional options for raw targets:

raw_target_compress <compress_command>|no: Compression algorithm to use for raw backup target. Supported <compress_command>: gzip, pigz, bzip2, pbzip2, bzip3, xz, lzo, lz4, zstd.
raw_target_compress_level default|<number>: Compression level for the specified <compress_command>.
raw_target_compress_long default|<number>: Enable long distance matching for the specified <compress_command>.
raw_target_compress_threads default|<number>: Number of threads to use for <compress_command>.
raw_target_split <size>|no: Split the raw backup file into pieces of size <size>.
raw_target_block_size <number>: Block size to use for writing the raw backup file. Defaults to “128K”.
raw_target_encrypt gpg|openssl_enc|no: If enabled, encrypt the target raw file using gpg or openssl_enc.

Additional options for "raw_target_encrypt gpg":

gpg_keyring <file>: Keyring to use for gpg, e.g. "/etc/btrbk/gpg/pubring.kbx".
gpg_recipient <name>…: Encrypt for user id <name> (email address).

Additional options for "raw_target_encrypt openssl_enc" (very experimental):

openssl_ciphername <name>

Defaults to “aes-256-cbc”.

openssl_iv_size <size-in-bytes>|no

Depends on selected cipher.

openssl_keyfile <file>|no

Point to a key file in hex (absolute path). Example key file creation (256bit key):

  # dd if=/dev/urandom bs=1 count=32 \
    | od -x -A n \
    | tr -d "[:space:]" > /path/to/keyfile

kdf_backend <file>|no

KDF backend to be executed, e.g. "/usr/share/btrbk/scripts/kdf_pbkdf2.py".

kdf_keysize <size-in-bytes>

Defaults to “32”.

kdf_keygen once|each

Defaults to “once”.

AVAILABILITY

Please refer to the btrbk project page https://digint.ch/btrbk/ for further details.

AUTHOR

Axel Burri <axel@tty0.ch>

Document Archive: btrbk-0.32.6

btrbk-0.32.6