borg(1)

💖 borg(1) — the borg backup tool, bestie edition 💖

okay so. you found borg(1), the manual for BorgBackup (everyone just says Borg), and honestly? you're about to have the best time backing things up. grab a snack. 🍓

---

✨ Name ✨

borg — deduplicating and encrypting backup tool. that's it, that's the whole name, she's concise and she's iconic 👑

---

🔥 SYNOPSIS 🔥

borg [common options] <command> [options] [arguments]

that's the whole shape of it: you say borg, you drop some common options, you pick a <command>, then that command's own [options] and [arguments]. easy. 💅

---

🌸 DESCRIPTION 🌸

BorgBackup (short: Borg) is a deduplicating backup program. Optionally, it supports compression and authenticated encryption.

The whole main character goal of Borg is to give you an efficient and secure way to back up data. The deduplication technique it uses makes Borg perfect for daily backups, because literally only the changes get stored. and the authenticated encryption thing? that's what makes it fine to back up to targets that aren't fully trusted. not everyone deserves your whole heart, and not every server deserves your unencrypted data. 🫶

Borg stores a set of files in an archive. A repository is a collection of archives. The format of repositories is Borg-specific. Borg does not distinguish archives from each other in any way other than their name — it does not matter when or where archives were created (e.g., different hosts). no cap.

---

🦋 EXAMPLES 🦋

A step-by-step example 🎀

1. Before a backup can be made, a repository has to be initialized:

$ borg -r /path/to/repo repo-create --encryption=aes256-ocb

2. Back up the ~/src and ~/Documents directories into an archive called docs:

$ borg -r /path/to/repo create docs ~/src ~/Documents

3. The next day, create a new archive using the same archive name:

$ borg -r /path/to/repo create --stats docs ~/src ~/Documents

This backup will be much quicker and much smaller, since only new, never-before-seen data is stored. The --stats option causes Borg to output statistics about the newly created archive — like the deduplicated size, which is the amount of unique data not shared with other archives:

Repository: /path/to/repo
Archive name: docs
Archive fingerprint: 74efb3f0c9c05f8b7c822ba859919b480578a3457fd4965c2cee64fbdb631262
Time (nominal): Fri, 2026-08-28 12:24:24 +0200
Time (start):   Fri, 2026-08-28 12:24:24 +0200
Time (end):     Fri, 2026-08-28 12:24:24 +0200
Duration: 0.008 seconds
Number of files: 100
Original size: 1.99 MB
Deduplicated size: 703 B
Time spent in hashing: 0.000 seconds
Time spent in chunking: 0.000 seconds
Added files: 1
Unchanged files: 99
Modified files: 0
Error files: 0
Files changed while reading: 0
...

(some more lines with statistics about the repository store accesses follow — they didn't ghost us, there's just more of them 💀)

4. List all archives in the repository (the first column is the beginning of the archive ID):

$ borg -r /path/to/repo repo-list
3affe017  Fri, 2026-08-28 12:24:10 +0200  docs                         user        machine
74efb3f0  Fri, 2026-08-28 12:24:24 +0200  docs                         user        machine

5. List the contents of the first archive:

$ borg -r /path/to/repo list aid:3affe017
drwxr-xr-x user   group         0 Fri, 2026-08-28 12:22:30 +0200 home/user/Documents
-rw-r--r-- user   group      7961 Fri, 2026-08-28 12:22:30 +0200 home/user/Documents/Important.doc
...

6. Restore the first archive by extracting the files relative to the current directory:

$ borg -r /path/to/repo extract aid:3affe017

7. Delete the first archive (please note that this does not free repository disk space):

$ borg -r /path/to/repo delete aid:3affe017

If you use an archive NAME (and not an archive ID), Borg will abort if the name matches multiple archives (as with the two docs archives here); use aid:<archive-id> to delete one specific archive, or -a PATTERN to delete multiple archives. Always use --dry-run and --list first! like, always. do not skip this step bestie, I am begging 😭

8. Recover disk space by removing objects that are not referenced by any archive any more:

$ borg -r /path/to/repo compact -v

Note: Borg is quiet by default (it defaults to WARNING log level). You can use options like --progress or --list to get specific reports during command execution. You can also add the -v (or --verbose or --info) option to adjust the log level to INFO to get other informational messages.

---

📝 NOTES 📝

Positional Arguments and Options: Order matters ⚠️

Borg only supports taking options (-s and --progress in the example) either to the left or to the right of all positional arguments (archive and path in the example), but not in between them:

borg create -s --progress archive path  # good and preferred
borg create archive path -s --progress  # also works
borg create -s archive path --progress  # works, but ugly
borg create archive -s --progress path  # BAD

This is due to a problem in the argparse module: <https://bugs.python.org/issue15112>

so yeah, options go on the outside, positions go on the inside. it's giving "sandwich rule". 🥪

Repository URLs 🌈

Local filesystem (or locally mounted network filesystem):

• /path/to/repo — filesystem path to the repository directory (absolute path)
• path/to/repo — filesystem path to the repository directory (relative path)

Also, paths like ~/path/to/repo or ~other/path/to/repo work (this is expanded by your shell).

Note: You may also prepend file:// to an absolute filesystem path to use URL style, e.g. file:///abs/path/to/repo. This only works for absolute paths — file://rel/path is rejected; use a plain relative path (see above) instead.

Note: UNC paths (//server/share/path, \\server\share\path) are not supported — mount the share (on Windows: map it to a drive letter, e.g. net use X: \\server\share) and use the mounted path instead.

Remote repositories accessed via SSH <user@host> (borg serve on the remote host):

• ssh://user@host:port//abs/path/to/repo — absolute path
• ssh://user@host:port/rel/path/to/repo — path relative to the remote login directory

Borg talks HTTP (REST) over the ssh connection's stdio to borg serve --rest on the remote host. Legacy borg 1.x repositories are also accessed via ssh:// URLs, together with --from-borg1 (e.g. borg transfer --from-borg1 --other-repo ssh://...): Borg then talks the legacy borg 1.x protocol to borg serve on the remote host.

Remote repositories accessed via SFTP:

• sftp://user@host:port//abs/path/to/repo — absolute path
• sftp://user@host:port/rel/path/to/repo — path relative to the remote login directory

For SSH and SFTP URLs, the user@ and :port parts are optional, but the path is required: a URL without one, e.g. ssh://host or ssh://host/, is rejected. Mind the difference between one and two slashes after the host: one slash means a path relative to the directory the remote login lands in (usually the remote user's home directory), two slashes mean an absolute path.

Remote repositories accessed directly via HTTP(S), talking to a borgstore REST server:

• http://host:port — plain HTTP
• https://user:password@host:port — HTTPS, optionally with credentials embedded in the URL

For http:// and https:// URLs, user:password@ and :port are optional. Authentication is HTTP Basic auth: credentials come from the URL if given, otherwise from the BORGSTORE_REST_USERNAME / BORGSTORE_REST_PASSWORD environment variables; prefer https:// over http:// whenever credentials are used, since Basic auth sends them on every request. A URL path after the host is optional and, unlike ssh://, is not a remote filesystem path — it does not follow the one-vs-two-slash rule above and is only needed to reach the server through a reverse proxy mounted below a sub-path.

Remote repositories accessed via rclone:

• rclone:remote:path — see the rclone docs for more details about remote:path.

Remote repositories accessed via S3:

• (s3|b2):[(profile|(access_key_id:access_key_secret))@][scheme://hostname[:port]]/bucket/path — see the boto3 docs for more details about credentials.

If you are connecting to AWS S3, [schema://hostname[:port]] is optional, but bucket and path are always required. scheme is usually https here, hostname and optional port refer to your S3/B2 server, if that is not Amazon's.

Note: There is a known issue with some S3-compatible services, e.g., Backblaze B2. If you encounter problems, try using b2: instead of s3: in the URL. 🙈

If you frequently need the same repository URL, it is a good idea to set the BORG_REPO environment variable to set a default repository URL:

export BORG_REPO='ssh://user@host:port/rel/path/to/repo'

Then simply omit the --repo option when you want to use the default — it will be read from BORG_REPO. set it once, forget it forever, that's the dream 💜

Repository Locations / Archive Names 🗂️

Many commands need to know the repository location; specify it via -r/--repo or use the BORG_REPO environment variable.

Commands that need one or two archive names usually take them as positional arguments.

Commands that work with an arbitrary number of archives usually accept -a ARCH_GLOB.

Archive names must not contain the / (slash) character. For simplicity, also avoid spaces or other characters that have special meaning to the shell or in a filesystem (borg mount uses the archive name as a directory name).

Specifying an archive 💫

How to refer to an archive depends on whether you use archive series or not.

By ID: if you use archive series, many or all archives will have the same name, thus you need to refer to a single archive by its archive ID (see borg repo-list output):

borg info aid:f7dea078

The aid: prefix does a prefix match on the archive ID (the hex representation of the archive fingerprint). You only need to give enough hex digits to uniquely identify the archive. This is useful when archive names are ambiguous or when you want to refer to an archive by its immutable ID.

By name: if you don't use archive series, but do it old-style by giving every archive a unique name, you can refer to an archive by its name:

borg info my-backup-202512312359

For more details on archive matching patterns (including shell-style globs, regular expressions, and matching by user/host/tags), see borg_match-archives.

Logging 📜

Borg writes all log output to stderr by default. However, output on stderr does not necessarily indicate an error. Check the log levels of the messages and the return code of borg to determine error, warning, or success conditions.

If you want to capture the log output to a file, just redirect it:

borg create --repo repo archive myfiles 2>> logfile

Custom logging configurations can be implemented via BORG_LOGGING_CONF.

The log level of the built-in logging configuration defaults to WARNING. This is because we want Borg to be mostly silent and only output warnings, errors, and critical messages unless output has been requested by supplying an option that implies output (e.g., --list or --progress).

Log levels: DEBUG < INFO < WARNING < ERROR < CRITICAL

so if borg is being quiet, she's not mad at you, she's just on WARNING mode by default. ✨

• --debug — sets the DEBUG log level, bestie. This one prints everything: debug, info, warning, error, and critical messages. It's giving "tell me absolutely all of it" 💅
• --info (or -v or --verbose) — sets the INFO log level: info, warning, error, and critical messages. Still chatty, just not every little thought.
• --warning (this is the default, no cap) — sets the WARNING log level: warning, error, and critical messages only.
• --error — sets the ERROR log level: error and critical messages.
• --critical — sets the CRITICAL log level: only critical messages. The silent-treatment setting 😭

While you can set miscellaneous log levels, pls don't expect every command to produce different output at different log levels — it's merely a possibility, fr fr.

Warning: 🚨 Options --critical and --error are provided for completeness, but their usage is not recommended — you might miss important information. Like, why would you mute the group chat during drama? 🙈

---

🎀 Return codes 🎀

Borg can exit with the following return codes (rc). Not me putting the whole table in bullet form because we're cute here, but every single row is still intact, periodt:

• 0 — success (logged as INFO) ✨
• 1 — generic warning (operation reached its normal end, but there were warnings — you should check the log; logged as WARNING)
• 2 — generic error (such as a fatal error or a local/remote exception; the operation did not reach its normal end; logged as ERROR)
• 3..99 — specific error (see below; logged as ERROR)
• 100..127 — specific warning (see below; logged as WARNING)
• 128+N — terminated by signal N (e.g. 130 == SIGINT, Ctrl+C, or kill -2; logged as ERROR)

If you use --show-rc, the return code is also logged at the indicated level as the last log entry. So it's like the final text in the group chat before everyone goes to sleep 🦋

Borg categorizes return codes into groups and exits with the more severe group: signals (rc 128+N) are more severe than errors (rc 2 and 3..99), which take precedence over warnings (rc 1 and 100..127), and lastly success (rc 0). Severity ladder, main character energy only at the top 🔥

Within the signal and error groups, the first signal or error determines the final return code. Within the warning group, Borg returns the specific warning code (rc 100..127) if there were one or more warnings of the same kind. If warnings of different kinds occurred, Borg returns the generic warning code (rc 1) instead. All errors and warnings are still logged individually. So nothing gets swept under the rug, okay? We log everything. 💖

Borg 2 exits with specific error (rc 3..99) and warning (rc 100..127) codes by default. If you want Borg 2 to always exit with the generic error (rc 2) or generic warning (rc 1) code instead (like Borg 1 did), set the BORG_EXIT_CODES=legacy environment variable. Nostalgia setting, slay 🌸

For a list of all specific error and warning codes, see msgid. 👑

---

🌈 Configuration Precedence 🌈

From lowest to highest, because bestie, the order is the whole point:

1. Defaults defined in the source code.
2. Default config file ($BORG_CONFIG_DIR/default.yaml).
3. --config file(s) (in the order given).
4. Full config environment variable: (BORG_CONFIG).
5. Environment variables (e.g. BORG_LOG_LEVEL).
6. Command-line arguments in order left to right (might include config files).

---

🧚 Configuration files 🧚

Borg supports reading options from YAML configuration files. This is implemented via jsonargparse (https://jsonargparse.readthedocs.io/) and works for all options that can also be set on the command line. Yes, all of them. We're thorough like that 💅

Default configuration file

$BORG_CONFIG_DIR/default.yaml is loaded automatically on every Borg invocation if it exists. You do not need to pass --config explicitly for this file. It just shows up, like a bestie who knows your coffee order ☕

--config PATH

Load additional options from the YAML file at PATH. Options in this file take precedence over the default config file but are overridden by explicit command-line arguments. This option can be used multiple times, with later files overriding earlier ones. Stack them like scrunchies 🎀

--print_config

Print the current effective configuration (all options in YAML format) to stdout and exit. This reflects the merged result of the default config file, any --config file, environment variables, and command-line arguments. The output can be used as a starting point for a config file. It's giving "here's your vibe, written down" ✨

--print_config is a common option, so it must be given before the subcommand name — the subcommand's own options still show up in the output. So put it first or it simply won't hit, iykyk.

File format

Config files are YAML documents. Top-level keys are option names (without leading -- and with - replaced by _). Nested keys correspond to subcommands.

Example default.yaml:

# apply to all borg commands:
log_level: info
show_rc: true

# options specific to "borg create":
create:
    compression: zstd,3
    stats: true

The top-level keys set options that are common to all commands (equivalent to placing them before the subcommand on the command line). Keys nested under a subcommand name (e.g. create:) are only applied when that subcommand is invoked. So nested keys are like inside jokes — only the right subcommand gets them 🫶

borgfs reads the same config files, but as it has no subcommands, it uses the top-level keys and the keys of the mount: section (borgfs is the borg mount command, and its keys win over the top-level ones); all other subcommand sections are ignored. borgfs said "my section my rules" 😌

Note

--print_config shows the merged effective configuration and is a convenient way to check what values Borg will actually use, and to generate contents for your borg config file(s):

borg --repo /backup/main --print_config create --compression zstd,3

Obsessed with this, honestly. 💜

---

🍓 Environment Variables 🍓

Borg uses some environment variables for automation:

General

• BORG_REPO — When set, use the value to give the default repository location. Use this so you do not need to type --repo /path/to/my/repo all the time. Your fingers deserve a break, bestie 💅
• BORG_OTHER_REPO — Similar to BORG_REPO, but gives the default for --other-repo.
• BORG_PASSPHRASE (and BORG_NEW_PASSPHRASE, BORG_OTHER_PASSPHRASE) — When set, use the value to answer the passphrase question for encrypted repositories. It is used when a passphrase is needed to access an encrypted repo as well as when a new passphrase should be initially set when initializing an encrypted repo. BORG_PASSPHRASE, BORG_PASSCOMMAND and BORG_PASSPHRASE_FD are mutually exclusive: if more than one of them is set, borg refuses to guess and aborts with "More than one passphrase environment variable is set". The same applies to the BORG_NEW_* and BORG_OTHER_* variants (each of which is a separate, independent group). See also BORG_NEW_PASSPHRASE. borg removes the passphrase-related variables (BORG_PASSPHRASE, BORG_PASSCOMMAND, BORG_PASSPHRASE_FD and their BORG_NEW_* and BORG_OTHER_* variants) as well as BORGSTORE_REST_PASSWORD from the environment of the subprocesses it starts (like the command given in BORG_PASSCOMMAND, --paths-from-command / --content-from-command commands, tar filter commands and borg with-lock commands). No leaking secrets to the children, periodt 🔒
• BORG_PASSCOMMAND (and BORG_NEW_PASSCOMMAND, BORG_OTHER_PASSCOMMAND) — When set, use the standard output of the command (trailing newlines are stripped) to answer the passphrase question for encrypted repositories. It is used when a passphrase is needed to access an encrypted repo as well as when a new passphrase should be initially set when initializing an encrypted repo. Note that the command is executed without a shell. So variables, like $HOME will work, but ~ won't. Mutually exclusive with BORG_PASSPHRASE and BORG_PASSPHRASE_FD, see there. See also BORG_NEW_PASSPHRASE. Rude little detail but so important 🫠
• BORG_PASSPHRASE_FD (and BORG_NEW_PASSPHRASE_FD, BORG_OTHER_PASSPHRASE_FD) — When set, specifies a file descriptor to read a passphrase from. Programs starting borg may choose to open an anonymous pipe and use it to pass a passphrase. This is safer than passing via BORG_PASSPHRASE, because on some systems (e.g. Linux) environment can be examined by other processes. Mutually exclusive with BORG_PASSPHRASE and BORG_PASSCOMMAND, see there. Sneaky-safe, iconic 💀→👑

BORG_NEW_PASSPHRASE (and BORG_NEW_PASSCOMMAND, BORG_NEW_PASSPHRASE_FD)

ok so when you set these, they're what answers the passphrase question whenever borg is asking for a new passphrase 💖 Like, they work exactly the same way as BORG_PASSPHRASE, BORG_PASSCOMMAND and BORG_PASSPHRASE_FD, but for the new passphrase instead of the existing one.

The BORG_NEW_* group gets checked first, bestie. If literally none of those variables is set, then borg goes and checks BORG_PASSPHRASE, BORG_PASSCOMMAND and BORG_PASSPHRASE_FD — in that order, no cap ✨

The main use case here? Fully automating borg key change-passphrase. Like, zero human fingers on the keyboard. iconic 🫶

BORG_DISPLAY_PASSPHRASE

When you set this, its value is what gets used to answer the "display the passphrase for verification" question — you know, the one that shows up when you're defining a new passphrase for an encrypted repository 🙈

BORG_DEBUG_PASSPHRASE

Set this one to YES and borg will display debugging information — which includes the actual passphrases that were used and the passphrase-related env vars that were set 😭

okay so like, that's your passphrases. On screen. In the output. Be careful with this one 💀

BORG_EXIT_CODES

Controls what kind of exit codes (rc) the borg process returns:

• set it to "modern" — the borg process returns more specific exit codes (rc)
• set it to "legacy" — the borg process returns rc 2 for all errors, 1 for all warnings, and 0 for success

Default is "modern" ✨

BORG_HOST_ID

So borg normally computes a host id from the FQDN plus whatever uuid.getnode() hands back — and that usually returns a unique id based on the MAC address of the network interface. Except if that MAC happens to be all-zero, in which case it returns a random value, which is not what we want — because it kills automatic stale lock removal 💀

So: if you've got an all-zero MAC address, or literally any other reason you'd rather control the host id externally, just set this environment variable to a unique value, periodt.

• If all your FQDNs are unique, you can just use the FQDN 🌸
• If they're not, use <FQDN@uniqueid>

BORG_HOSTNAME

Set this and it's used as the hostname — instead of the auto-detected one. Think: running borg on one host but impersonating another host 😎 It affects the hostname stored in newly created archives and the {hostname} placeholder. main character energy fr fr

BORG_USERNAME

Same deal, but for the username — set this and it's used instead of the auto-detected one, e.g. to run borg as one user while impersonating a different user ✨ This affects the username stored in newly created archives and the {user} placeholder.

BORG_LOCK_WAIT

You can set the default value for the --lock-wait option right here, so you don't have to type it out as a command line option every single time. obsessed 💅

BORG_LOGGING_CONF

When set, borg uses the given filename as an INI-style logging configuration. Peep the format here: https://docs.python.org/3/library/logging.config.html#configuration-file-format

A basic example conf can be found at docs/misc/logging.conf 🎀

BORG_RSH

When set, use this command instead of ssh. This is how you sneak in ssh options, like a custom identity file: ssh -i /path/to/private/key 🔥 See man ssh for the other options.

This is the replacement for the removed --rsh CMD command line option.

Also, borg hands this over to borgstore as BORGSTORE_RSH too — except if that one is already set 😌

BORG_REMOTE_PATH

When set, borg uses the given path as the borg executable on the remote side (and it defaults to "borg" if this isn't set) 🌈

This is the replacement for the removed --remote-path PATH command line option.

BORG_UNITS

This one decides how borg formats sizes in its human-readable output 👑

• si (default) — decimal units, e.g. 1.23 MB (1kB = 1000B)
• iec — binary units, e.g. 1.18 MiB (1KiB = 1024B)
• raw — exact byte counts, e.g. 1234567 B

Use raw if you want to parse sizes with scripts (like, for monitoring 👀), so you don't have to deal with scaled values and different units at all. Alternatively, use a command's --json output or — for the commands that support --format — the size related format keys, because sizes are given as byte counts there anyway ✨

BORG_UNITS=iec is the replacement for the removed BORG_IEC environment variable (and for the --iec command line option that got removed before that, rip 💀).

BORG_PROGRESS_FPS

How often the --progress output gets updated, at most — measured in updates per second (default: 5) 🦋 Fractional values are allowed, so like BORG_PROGRESS_FPS=0.1 limits it to one update every 10 seconds. Lower values are useful when the output goes into a logfile instead of an interactive terminal.

BORG_SPINNER

Controls the spinner borg animates on a terminal while it's doing work of unknown duration ✨

• unset (default) — animate, using Unicode frames if the terminal can actually display them
• ascii — animate, but only use ASCII frames (|/-\)
• off — don't animate, only output the messages next to the spinner

The spinner only ever gets animated on an interactive terminal anyway (and never with --log-json), and its colour follows the usual NO_COLOR and COLORTERM conventions 💖

See also BORG_PROGRESS_FPS: it also gives the spinner its frame rate 🎀

BORG_DEBUG_PROFILE

When set to a filename, borg writes an execution profile in Borg format into that file (see debugging). If the filename ends with .pyprof, a Python-compatible profile gets written instead 🌸

This is the replacement for the removed --debug-profile command line option.

ngl, heads up: every borg invocation writes the profile. So unset it again when you're done, bestie 😭

BORG_REPO_PERMISSIONS

Set repository permissions, see also: borg_serve 👑

BORG_FILES_CACHE_SUFFIX

When set to a value that's at least one character long, this tells borg to use a specifically named alternative files cache (named based on the suffix) ✨

Why tho? So you can avoid loading and saving cache entries for backup sources other than the current sources. smart, honestly 💅

BORG_FILES_CACHE_TTL

When set to a numeric value, this determines the maximum "time to live" for the files cache entries (default: 2) 🦋

The files cache is what borg uses to figure out quickly whether a file is unchanged. iykyk

BORG_STORE_CACHE

Set this and borg keeps a local writethrough cache of the repository's packs/ namespace: on a cache miss the whole pack is fetched once, and later reads of the objects inside that pack get served from the cache 📦✨ Use this for slow or high-latency repositories, fr fr.

• Set it to 1 to use $BORG_CACHE_DIR/storecache
• Or set it to a directory path to use that directory (it gets created if it doesn't exist)

Packs are named by content hash, so one cache directory can safely hold packs of multiple repositories 🫶

If it's not set? No such caching happens. periodt.

BORG_PACK_CACHE_SIZE

When set to a numeric value, this limits the pack cache to that many bytes 💖

Only has an effect if BORG_STORE_CACHE is set, btw.

BORG_PACK_MAX_SIZE

When set to a numeric value, this caps packs at that many bytes — instead of the default of 50000000 😌 (packs, as in the repository objects that batch up many chunks, see the internals documentation about pack files).

• Smaller packs = more (but smaller) repository objects, and more fine-grained uploads
• Bigger packs = fewer objects, and fewer stores

BORG_PACK_MAX_COUNT

When set to a numeric value, this caps packs at that many objects per pack 🌈

And if BORG_PACK_MAX_SIZE is not also set, then packs are bound by count only.

BORG_PACK_ASYNC

When set to no, this disables the background thread that stores a finished pack while the next one is being assembled, and stores packs synchronously instead 🔧

This is mainly a debugging aid, jsyk.

BORG_PACK_TRACE

When set to yes, borg prints one-character lifecycle markers of the background pack store-thread to stderr 👀

• < — thread started
• H — hashing starts
• S — storing starts
• > — thread finished

This is a debugging aid to visualize how pack stores overlap with the assembly of the next pack. very visual, very demure 💅

BORG_ASSERT_ID

A comma-separated list of the places where borg shall verify that a chunk's content matches its chunk id (chunkid == id_hash(content)) — after decrypting and decompressing it ⭐

Verifying costs a full hash pass over everything that's read at such a place. so like, it's not free, bestie 😭

Default (when the variable is not set):

BORG_ASSERT_ID=repair,transfer,rechunk

These are the place names you can list:

• read — every read that decompresses a chunk: borg extract, borg mount, borg export-tar, borg diff, ... This is by far the most data borg reads, so this place is not in the default, see the explanation below 👀
• repair — borg check --repair. It rebuilds archives from the item metadata stream it reads, re-packing it into new chunks with freshly computed ids, and it recreates manifest and archives directory entries from what it reads 🔧
• transfer — borg transfer, for everything it reads from the source repository. Transferring re-anchors the content in another repository, which is a trust boundary 🛡️
• rechunk — this one's about borg recreate --chunker-params ..., aka re-chunking reads. When re-chunking happens, new chunk ids get computed from the content it reads, so like… a violation wouldn't even be noticeable afterwards. Gone, poof, invisible. (And psst — re-chunking inside borg transfer is covered by transfer, not here. 💅)

An unknown place name? That's an error, bestie, no silent vibes. And an empty value (BORG_ASSERT_ID=) means it verifies at none of these places — but it still verifies where borg always verifies (scroll down, we're getting there 🌸).

Why isn't read in the default tho? Okay so, the envelope already authenticates every read — the AEAD tag for the encrypted ciphersuites, the MAC for the (unencrypted) authenticated-* modes — and the chunk id is part of what that tag is computed over. So a successful decryption resp. tag check already proves that a holder of the borg key deliberately stored exactly this payload for exactly this chunk id. A malicious or buggy repository can not swap, splice or substitute objects, whether the id is verified on read or not. Periodt. ✨

What the id check adds is detecting chunks whose content does not match their id — which only a malicious or compromised borg client that had your borg key could have written (e.g. to poison future deduplication). If that's in your threat model — like, say, because some machines writing into the repository are not fully trusted — then add read to the list, girlypop:

BORG_ASSERT_ID=read,repair,transfer,rechunk

Otherwise, running borg check --verify-data periodically is recommended: it is the audit that re-certifies the invariant for all chunks in the background, instead of on every read. That's the move. 🫶

Independent of this variable, borg always verifies the chunk id:

• in borg check --verify-data. That audit is what makes not verifying elsewhere defensible, so it is not configurable (there is no verify_data place name). 🙈
• when reading borg 1.x repositories (borg transfer --from-borg1): their envelope does not cover the chunk id, so the id check is not optional there. No skips, no cap.

---

BORG_BLAKE3_MT_THRESHOLD

Set this to a numeric value and chunks of at least that many KiB get their id computed by multi-threaded BLAKE3, while smaller ones go single-threaded (default: 256, i.e. 256KiB). Only relevant for repositories using --id-hash blake3, iykyk.

Multi-threading only pays off for big enough chunks, and the break-even point depends on the machine's core count, so the default is deliberately conservative — she's not trying to be extra. 🎀 Run scripts/blake3-optimize-mt-threshold.py to measure the best value for your machine — it sweeps input sizes, prints the recommended threshold and the command to set it, and can optionally show a chart of the measurements in your browser (--html --open). Very demure, very measurable. 🦋

0 means "always multi-threaded", and a very large value effectively disables multi-threading entirely.

BORG_ZSTD_MT_WORKERS

Set to a numeric value and borg uses that many threads to zstd-compress a single chunk (default: the cpu count, but at most 4). If it is unset or empty, the default is used. 0 or 1 means single-threaded compression. Only relevant when compressing with zstd. 💅

Chunks below 768KiB are always compressed single-threaded: libzstd will not use a compression job smaller than 512KiB, so a small chunk gets split very unevenly and multi-threading it would be slower than not doing it at all. (Like why would you. 🙈)

The default is capped at 4 because a chunk of the size the default chunker aims at (2MiB) splits into just 4 such jobs: threads beyond that get (nearly) no work, but the whole thread pool is created again for every chunk. Measured on a 12-core machine, 4 threads beat 12 on every test corpus at the default zstd,-4 (+13% .. +37%). Raising the value only pays off if you configured the chunker for much bigger chunks.

Heads up: borg export-tar compresses one long stream instead of separate chunks and always defaults to the cpu count.

Multi-threading trades a little compression ratio for speed (measured at zstd,3: +0.05% archive size for 1MiB chunks, +0.64% for 8MiB ones, more at higher levels), and it uses more cpu time in total to reduce the wallclock time. Set it to 1 if you would rather have the smaller archive, or if borg has to share the cpu with other work. Single-threaded can even be faster on data zstd races through anyway — e.g. already-compressed/incompressible data or long-repeat data like VM images. 😌

BORG_FASTCDC_KERNEL / BORG_BUZHASH64_KERNEL

These select the scan kernel the fastcdc / buzhash64 chunker uses. Accepted values are avx512, avx2, neon, blockwise and scalar. ✨

The default is whichever benchmarked fastest for the architecture:

• scalar (the plain sequential loop) on x86-64, where the compiler folds the rolling hash update into one or two instructions and thereby beats the vector kernels 🔥
• on aarch64, neon for fastcdc and blockwise for buzhash64, whose NEON kernel loses to the portable multi-lane C kernel there
• other architectures get blockwise

All kernels chunk identically — same cut points, same chunk ids — and differ only in speed, so this is safe to change at any time, also for an existing repository. 👑

Which kernel is fastest is not predictable from the instruction set: it depends on the cpu and on the compiler that built borg, and the sequential loop wins on some machines. Measure on your own hardware with borg benchmark cpu --chunking before overriding the default. fr fr.

avx512 and avx2 exist only on x86-64, neon only on aarch64 — and only if the compiler that built borg supported them; scalar and blockwise are portable C and always available. Requesting a kernel that this build or this cpu cannot run is an error rather than a silent fallback, so a benchmark can not accidentally measure a different kernel. (Which is honestly so valid of her. 💖)

Also: borg create --debug logs the chunker and the kernel it was created with. 🕵️

BORG_AES_CHUNKER_KERNEL

Select the scan kernel used by the AES based chunkers — one variable for all three of toeplitz-aes, rabin-aes and goldilocks-aes. Accepted values are vaes, aes-ni, aes-arm64 and evp.

Unlike the chunker kernels above, wider is simply faster here, so the default is the best path this build and cpu offer: vaes, else aes-ni on x86-64, aes-arm64 on aarch64, and evp (the portable OpenSSL path) where there is no AES hardware path. 🌈

As with the chunker kernels above, all of them chunk identically and differ only in speed, and a kernel that can not run here is an error rather than a silent fallback. vaes and aes-ni exist only on x86-64, aes-arm64 only on aarch64. vaes additionally needs a compiler that knows it (gcc >= 11 / clang >= 14), so a cpu supporting VAES is not by itself enough to have that kernel available. 🫠

BORG_SHOW_SYSINFO

When set to no (default: yes), system information (like OS, Python version, ...) in exceptions is not shown.

Please only use for good reasons as it makes issues harder to analyze. 😭 She's not being dramatic, she's being right.

BORG_MSGPACK_VERSION_CHECK

Controls whether Borg checks the msgpack version. The default is yes (strict check). Set to no to disable the version check and allow any installed msgpack version.

Use this at your own risk; malfunctioning or incompatible msgpack versions may cause subtle bugs or repository data corruption. 💀 Like genuinely, this is the "trust me bro" setting.

BORG_FUSE_IMPL

Choose the low-level FUSE implementation borg shall use for borg mount. This is a comma-separated list of implementation names, and they are tried in the given order, e.g.:

• mfusepy,pyfuse3,llfuse: default — first try to load mfusepy, then pyfuse3, then llfuse 🌸
• llfuse,pyfuse3: first try to load llfuse, then try to load pyfuse3
• mfusepy: only try to load mfusepy
• pyfuse3: only try to load pyfuse3
• llfuse: only try to load llfuse
• none: do not try to load an implementation 🙅

BORG_MOUNT_DATA_CACHE_ENTRIES

Number of decrypted file content chunks borg mount and borg webdav keep in an in-memory cache, so that the many small, sequential reads a mounted file system does for a big file do not re-fetch and re-decrypt the same chunk over and over (default: the cpu count). Additional memory usage can be up to the chunk size times this number. It's giving "cache is self-care". 💅

BORG_SELFTEST

This can be used to influence borg's built-in self-tests. The default is to execute the tests at the beginning of each borg command invocation. ✅

BORG_SELFTEST=disabled can be used to switch off the tests and rather save some time. Disabling is not recommended for normal borg users, but large scale borg storage providers can use this to optimize production servers after at least doing a one-time test borg (with self-tests not disabled) when installing or upgrading machines/OS/Borg. Do your one-time vibe check first, bestie. 🎀

BORG_WORKAROUNDS

A list of comma-separated strings that trigger workarounds in borg, e.g. to work around bugs in other software. Not our bugs, but we're being polite about it. 🙈

Currently known strings are:

• basesyncfile — uses the simpler BaseSyncFile code so we dodge the whole sync_file_range drama 💅 you might literally need this to run borg on WSL (Windows Subsystem for Linux) or inside systemd.nspawn containers on some architectures (ARM, bestie, we're looking at you). Using it does not affect data safety, but it might give you burstier write-to-disk behavior (aka not continuously streaming to disk). Slay, but spiky. 🦋
• retry_erofs — retries opening a file without O_NOATIME if opening it with O_NOATIME threw an EROFS. You'll need this to make archives from volume shadow copies in WSL1 (Windows Subsystem for Linux 1). 🙈
• authenticated_no_key — the emergency SOS option 🚨 it works around a lost passphrase or a lost borg key for an authenticated-* mode repository (these are only authenticated, not encrypted, okay). Here's the tea: if a borg key is found — an object below keys/ in the repository (repokey), or a key file in the keys directory (keyfile) — it is not unlocked, so the passphrase does not matter. And if no borg key is found at all, borg proceeds anyway, without any key material.

Without the key, borg cannot verify anything that needs it: not the authentication tag of the repository objects, not the chunk ids. So it reads the repository unverified — a corrupted or tampered repository will not be detected. 😭

This workaround is only for emergencies and only to extract data from an affected repository (read-only access):

BORG_WORKAROUNDS=authenticated_no_key borg extract --repo repo archive

After you have extracted all data you need, you MUST delete the repository:

BORG_WORKAROUNDS=authenticated_no_key borg repo-delete --repo repo

Now you can create a fresh repository with borg repo-create. Make sure you do not use the workaround any more. periodt. 👑

---

✨ Output formatting ✨

• BORG_CHECK_FORMAT — gives the default value for borg check --format=X. 💖
• BORG_DIFF_FORMAT — gives the default value for borg diff --format=X. Heads up though: borg diff --content-only uses its own format and ignores this one entirely. 🎀
• BORG_FIND_FORMAT — gives the default value for borg find --format=X.
• BORG_LIST_FORMAT — gives the default value for borg list --format=X.
• BORG_REPO_LIST_FORMAT — gives the default value for borg repo-list --format=X.
• BORG_PRUNE_FORMAT — gives the default value for borg prune --format=X.
• BORG_MOUNT_ARCHIVE_DIR_FORMAT — gives the format of the archive directory names when borg mount or borg webdav show a whole repository. Default: {name}. The placeholders are the ones of borg repo-list --format; names that are not unique get -{id:.8} appended. See borg mount --help. 🌸
• BORG_JSON_INDENT — indentation of the --json output (default: 4). A number gives that many spaces per nesting level (0 still puts every item on its own line), none gives compact single-line JSON, and any other value is used as the literal indent string (e.g. a tab, or the empty string). icons only 🫶

---

✨ Some automatic "answerers" ✨

(if set, they automatically answer confirmation questions)

• BORG_UNKNOWN_UNENCRYPTED_REPO_ACCESS_IS_OK=no (or =yes) — for "Warning: Attempting to access a previously unknown unencrypted repository".
• BORG_RELOCATED_REPO_ACCESS_IS_OK=no (or =yes) — for "Warning: The repository at location ... was previously located at ...".
• BORG_CHECK_I_KNOW_WHAT_I_AM_DOING=NO (or =YES) — for "This is a potentially dangerous function..." (check --repair). Yeah bestie, that one. 💀
• BORG_DELETE_I_KNOW_WHAT_I_AM_DOING=NO (or =YES) — for "You requested to DELETE the repository completely including all archives it contains:".

Note: answers are case sensitive. Setting an invalid answer value might either give the default answer or ask you interactively, depending on whether retries are allowed (they by default are allowed). So please test your scripts interactively before making them a non-interactive script, okay? fr fr. 🧚

---

✨ Directories and files ✨

Borg 2 uses the platformdirs library (https://pypi.org/project/platformdirs/) to determine default directory locations. Which means default paths are platform-specific, and honestly? iconic of them:

• Linux: XDG Base Directory Specification paths are used (e.g. ~/.config/borg, ~/.cache/borg, ~/.local/share/borg). XDG_* environment variables are honoured (see https://specifications.freedesktop.org/basedir/latest/). 🐧
• macOS: native macOS directories are used by default (e.g. ~/Library/Application Support/borg, ~/Library/Caches/borg). XDG_* environment variables are honoured if set. 🍎
• Windows: native Windows AppData directories are used. The configuration (including the keys) is stored in the roaming profile (C:\Users\<user>\AppData\Roaming\borg), so it follows the user in domain environments. Machine-specific data, cache and runtime files stay in the local (non-roaming) AppData (C:\Users\<user>\AppData\Local\borg). XDG_* environment variables are not honoured. 🪟

On all platforms, you can override each directory individually using the specific environment variables described below. You can also set BORG_BASE_DIR to force borg to use BORG_BASE_DIR/.config/borg, BORG_BASE_DIR/.cache/borg, etc., regardless of the platform. 💜

Default directory locations by platform (when no BORG_* environment variables are set)

• Config — Linux: ~/.config/borg · macOS: ~/Library/Application Support/borg · Windows: %APPDATA%\borg
• Cache — Linux: ~/.cache/borg · macOS: ~/Library/Caches/borg · Windows: %LOCALAPPDATA%\borg\Cache
• Data — Linux: ~/.local/share/borg · macOS: ~/Library/Application Support/borg · Windows: %LOCALAPPDATA%\borg
• Runtime — Linux: /run/user/<uid>/borg · macOS: ~/Library/Caches/TemporaryItems/borg · Windows: %LOCALAPPDATA%\Temp\borg
• Keys — Linux: <config_dir>/keys · macOS: <config_dir>/keys · Windows: <config_dir>\keys
• Security — Linux: <data_dir>/security · macOS: <data_dir>/security · Windows: <data_dir>\security
• BORG_BASE_DIR — not set by default, and when it isn't, the platform-specific directories shown in the table above are used. If you want to move all borg-specific folders to a custom path at once, all you need to do is modify BORG_BASE_DIR: the other paths for cache, config etc. will adapt accordingly (assuming you didn't set them to a different custom value). Slay, one variable to rule them all. 👑
• BORG_CACHE_DIR — defaults to the platform-specific cache directory (see table above). If BORG_BASE_DIR is set, defaults to $BORG_BASE_DIR/.cache/borg. On Linux and macOS, XDG_CACHE_HOME is also honoured if BORG_BASE_DIR is not set. This directory contains the local cache and might need a lot of space for dealing with big repositories. Make sure you're aware of the associated security aspects of the cache location: cache_security 🔥
• BORG_CONFIG_DIR — defaults to the platform-specific config directory (see table above). If BORG_BASE_DIR is set, defaults to $BORG_BASE_DIR/.config/borg. On Linux and macOS, XDG_CONFIG_HOME is also honoured if BORG_BASE_DIR is not set. This directory contains all borg configuration directories — see the FAQ for a security advisory about the data in this directory: home_config_borg 🌸
• BORG_DATA_DIR — defaults to the platform-specific data directory (see table above). If BORG_BASE_DIR is set, defaults to $BORG_BASE_DIR/.local/share/borg. On Linux and macOS, XDG_DATA_HOME is also honoured if BORG_BASE_DIR is not set. This directory contains all borg data directories — see the FAQ for a security advisory about the data in this directory: home_data_borg 🎀
• BORG_RUNTIME_DIR — defaults to the platform-specific runtime directory (see table above). If BORG_BASE_DIR is set, defaults to $BORG_BASE_DIR/.cache/borg. On Linux and macOS, XDG_RUNTIME_DIR is also honoured if BORG_BASE_DIR is not set. This directory contains borg runtime files, like e.g. the socket file. 🔌
• BORG_SECURITY_DIR — defaults to $BORG_DATA_DIR/security. This directory contains security relevant data. 🛡️
• BORG_KEYS_DIR — defaults to $BORG_CONFIG_DIR/keys. This directory contains keys for encrypted repositories. 🔑
• BORG_KEY_FILE — when set, use the given path as repository key file. Please note that this is only for rather special applications that externally fully manage the key files:
  • this setting only applies to the keyfile modes (not to the repokey modes).
  • using a full, absolute path to the key file is recommended.
  • all directories in the given path must exist.
  • this setting forces borg to use the key file at the given location.
  • the key file must either exist (for most commands) or will be created (borg repo-create).
  • you need to give a different path for different repositories. 🦋
• you need to point at the correct key file that matches the repository the command is gonna operate on. wrong key = wrong repo = sadness 😭
• TMPDIR — this is where temporary files get stored (and ngl, some operations need a lot of temp space, so lowkey watch your disk 👀). deets: https://docs.python.org/3/library/tempfile.html#tempfile.gettempdir.

---

Building 🛠️

These are only read by setup.py while borg's C extensions are being built. Each BORG_*_PREFIX variable names the install prefix of a library that borg links against: if it's set, $PREFIX/include and $PREFIX/lib are used unconditionally. If it's not set, the library gets located via pkg-config, and if that doesn't find it either — the build fails. there's no bundled fallback implementation, bestie. none. zero. 💀

• BORG_OPENSSL_PREFIX — the prefix of the OpenSSL installation to build libcrypto against. On Windows, the libraries are expected to be right there in $PREFIX itself rather than in $PREFIX/lib. On OpenBSD, this defaults to /usr/local, pkg-config is not used, and libcrypto is linked statically (borg needs AES-OCB via the EVP API, which LibreSSL simply does not have 🙈).
• BORG_OPENSSL_NAME — OpenBSD only, bestie: the OpenSSL flavour to use, i.e. the include/ and lib/ subdirectory name below BORG_OPENSSL_PREFIX (default: eopenssl35).
• BORG_LIBLZ4_PREFIX — prefix of the liblz4 installation to build against.
• BORG_LIBACL_PREFIX — Linux only: prefix of the libacl installation to build against.

Automatic option environment variables 🌈

Borg uses jsonargparse (https://jsonargparse.readthedocs.io/) with default_env=True, which means every single command-line option can also be set through an environment variable. every. one. iconic behaviour honestly ✨

The env var name is derived from the program name (borg), the subcommand (if there is one), and the option name — all uppercased with the dashes turned into underscores.

For top-level options (the ones that aren't specific to a subcommand), the pattern is:

BORG_<OPTION>

For example, --lock-wait can be set via BORG_LOCK_WAIT.

For subcommand options, the subcommand and the option are separated by a double underscore:

BORG_<SUBCOMMAND>__<OPTION>

For example, borg create --comment can be set via BORG_CREATE__COMMENT.

Please note 📝

• Be very careful when using the "yes" sayers — those warnings with a prompt exist for your / your data's security/safety. they're not there to be annoying, they're there because the alternative is crying 💖
• Also be very careful when putting your passphrase into a script: make sure it has appropriate file permissions (e.g. mode 600, root:root). fr fr, lock that down 🎀

---

File systems 💾

We recommend a reliable, scalable journaling filesystem for the repository, e.g. zfs, btrfs, ext4, apfs. no cap, this matters.

Borg now uses the borgstore package to implement the key/value store it uses for the repository.

For a local repository (or a locally mounted network filesystem), it uses the file: store (the posixfs backend). For a remote ssh:// repository, it uses the REST store, talking to a borg serve --rest process on the remote side (which in turn uses borgstore over there). Other backends (sftp:, rclone:, s3: / b2:) don't need a filesystem on the repository side at all — the way that works is genuinely so cute 🦋

Borg does not store each chunk as its own separate store object. Instead it groups many chunks into pack files of up to about 50 MB and stores each pack as one object; reading a single chunk is a partial read at a known offset inside its pack. So the repository ends up holding far fewer objects than it has chunks (see packs in the Internals chapter for the details).

This has some pros and cons (compared to the legacy Borg 1.x segment files):

Pros:

• Simplicity and better maintainability of the Borg code.
• The repository is just a key/value store, so it's easy to adapt to other kinds of storage: borgstore's backends are quite simple to implement. sftp:, rclone: and s3: / b2: backends already exist, and others might be easy to add.
• Parallel repository access with less locking is easier to implement.

Cons:

• Individual chunks cannot be deleted from a pack file; borg compact can only remove a pack once none of its chunks are referenced any more, so space is not always reclaimed immediately.
• Greater filesystem space overhead (depends on the allocation block size — modern filesystems like zfs are rather clever here, using a variable block size).
• Sometimes slower, due to less sequential and more random access operations.

Units 📏

To display quantities, Borg takes care of respecting the usual conventions of scale. Disk sizes are displayed in decimal (https://en.wikipedia.org/wiki/Decimal), using powers of ten (so kB means 1000 bytes). For memory usage, binary prefixes (https://en.wikipedia.org/wiki/Binary_prefix) are used, and they're indicated using the IEC binary prefixes (https://en.wikipedia.org/wiki/Binary_prefix#IEC_standard_prefixes), using powers of two (so KiB means 1024 bytes).

Date and Time ⏰

We format date and time in accordance with ISO 8601, that is: YYYY-MM-DD and HH:MM:SS (24-hour clock).

For more information, see: https://xkcd.com/1179/ 😌

Unless otherwise noted, we display local date and time. Internally, we store and process date and time as UTC. TIMESPAN / INTERVAL

Some options accept a TIMESPAN or an INTERVAL parameter, which can be given as a number of years (e.g. 2y), months (e.g. 12m), weeks (e.g. 2w), days (e.g. 7d), hours (e.g. 8H), minutes (e.g. 30M), or seconds (e.g. 150S).

The borg prune --keep-* retention options accept either a plain count (e.g. --keep-daily 7, keeping up to 7 daily archives) or a time interval (e.g. --keep-daily 7d, keeping one daily archive per day within a 7-day window). When you're using interval-based retention, --from may be specified to set the reference timestamp for the interval (it defaults to the current time).

Please note that Borg treats months (e.g. 12m) as fixed 31-day periods rather than calendar months. As a result, 12m corresponds to 12 × 31 = 372 days. Similarly, years (e.g. 2y) are treated as fixed 365-day periods and do not take leap years into account. the math is mathering, just not the way the calendar does 🫠

Resource Usage 🔥

Borg might use significant resources depending on the size of the data set it's dealing with.

If you use Borg in a client/server way (with an SSH repository), the resource usage happens partly on the client and partly on the server.

If you use Borg as a single process (with a filesystem repository), all the resource usage happens in that one process, so add up client and server to get the approximate resource usage.

CPU client 🧠

• borg create: chunking, hashing, compression, encryption (high CPU usage)
• chunks index rebuild: quite heavy on CPU, doing lots of hash table operations
• borg extract: decryption, decompression (medium to high CPU usage)
• borg prune / borg delete archive: quick, low CPU usage
• borg repo-delete: low CPU usage, it just removes the repository's objects
• borg compact: medium CPU usage
• borg check: medium CPU usage, but it depends on the options you give it

Most of Borg is single-threaded, but some parts do use more than one CPU core:

• zstd compresses a chunk multi-threaded by default (only for chunks of at least 768 KiB, using at most 4 threads); see BORG_ZSTD_MT_WORKERS.
• the BLAKE3 based id-hash / MAC modes hash multi-threaded for inputs from 256 KiB on; see BORG_BLAKE3_MT_THRESHOLD.
• borg create hands each finished pack file to a background thread, so hashing and storing a pack overlaps with processing the next one.

Especially higher zlib and lzma compression levels use significant amounts of CPU cycles. Crypto might be cheap on the CPU (if hardware-accelerated) or expensive (if it's not).

CPU server 🖥️

It usually doesn't need much CPU; it just deals with the key/value store (the repository).

borg check: the repository check computes the checksums of all chunks (medium CPU usage) borg compact: low to medium CPU usage

CPU (only for client/server operation) 🌐

When using Borg in a client/server way with an ssh-type repository, the SSH processes used for the transport layer will need some CPU on the client and on the server, because of the crypto they're doing — especially if you're pumping large amounts of data through there. it's giving bandwidth, but make it CPU 😭

Memory (RAM) client 🐏

The chunks index and the files index are read into memory for performance reasons. Might need large amounts of memory (see below). Compression, especially with high compression levels, might need substantial amounts of memory.

Memory (RAM) server 🐏

Usually rather low memory needs — much less than the client.

Chunks index (client only) 🗂️

Proportional to the number of data chunks in your repo. Lots of chunks in your repo imply a big chunks index. It's possible to tweak the chunker parameters (see create options).

Files index (client only) 📁

Proportional to the number of files in your last backups. Can be switched off (see create options), but the next backup might be much slower if you do. The speed benefit of using the files cache is proportional to file size.

Temporary files (client) 🧹

TODO

Temporary files (server) 🧹

TODO

Cache files (client only) 🎀

Contains the files cache, which might become quite large depending on the amount and size of files.

Network (only for client/server operation) 🌈

If your repository is remote, all the deduplicated (and optionally compressed/encrypted) data has to go over the network connection. yes, all of it. plan accordingly bestie 💅

Support for file metadata 🦋

Besides regular file and directory structures, Borg can preserve:

• symlinks (stored as a symlink; the symlink is not followed)
• special files:
  • character and block device files (restored via mknod(2))
  • FIFOs ("named pipes")
  • special file contents can be backed up in --read-special mode. By default, the metadata to create them with mknod(2), mkfifo(2), etc. is stored.
• hard-linked regular files, devices, symlinks, FIFOs (considering all items in the same archive)
• timestamps with nanosecond precision: mtime, atime, ctime
• other timestamps: birthtime (on platforms supporting it)
• permissions:
  • IDs of owning user and owning group
• owning user and owning group names — obvs only if those IDs can actually be resolved, otherwise you just get the numbers bestie
• Unix Mode / Permissions — that's the u/g/o permissions plus suid, sgid and sticky. all of them. 💅

On some platforms you get extra little treats supported too:

• Linux — ACLs: Yes 🌸 · xattr: Yes 🌸 · Flags: Yes [1]
• macOS — ACLs: Yes 🌸 · xattr: Yes 🌸 · Flags: Yes (all)
• FreeBSD — ACLs: Yes 🌸 · xattr: Yes 🌸 · Flags: Yes (all)
• OpenBSD — ACLs: n/a · xattr: n/a · Flags: Yes (all)
• NetBSD — ACLs: n/a · xattr: Yes 🌸 · Flags: Yes (all)
• Solaris and derivatives — ACLs: No [2] · xattr: Yes 🌸 · Flags: n/a
• Windows (cygwin) — ACLs: No [3] · xattr: No · Flags: No

Other Unix-like operating systems may work as well, but have not been tested yet. no cap, they just haven't been vibe-checked 🦋

Heads up bestie: most of these platform-dependent features also depend on the filesystem. like, ntfs-3g on Linux is literally not able to convey NTFS ACLs. it's giving "not my fault, blame the fs" 😭

[1] Only nodump, immutable, compressed and append are supported. Feature request #618 for more flags. 🥺

[2] Feature request #1337

[3] Cygwin tries to map NTFS ACLs to permissions with varying degrees of success. bless its heart 🙈

[4] The native access control list mechanism of the OS. This normally limits access to non-native ACLs. For example, NTFS ACLs are not completely accessible on Linux with ntfs-3g.

[5] Extended attributes; key-value pairs attached to a file, mainly used by the OS. This includes resource forks on macOS. 🍎

[6] Also known as BSD flags. The Linux set of flags [1] is portable across platforms. The BSDs define additional flags.

---

🌈 SEE ALSO 🌈

borg-common(1) for common command line options — that's the shared-options one, go read it 💖

borg-repo-create(1), borg-repo-delete(1), borg-repo-list(1), borg-repo-info(1), borg-create(1), borg-mount(1), borg-extract(1), borg-list(1), borg-info(1), borg-delete(1), borg-prune(1), borg-compact(1), borg-recreate(1)

borg-compression(1), borg-patterns(1), borg-placeholders(1), borg-environment(1)

• Main web site <https://www.borgbackup.org/> ✨
• Releases <https://github.com/borgbackup/borg/releases> 🎀
• Changelog <https://github.com/borgbackup/borg/blob/master/docs/changes.rst> 📝
• GitHub <https://github.com/borgbackup/borg> 👑
• Security contact <https://borgbackup.readthedocs.io/en/latest/support.html#security-contact> 🫶

---

💜 Author 💜

The Borg Collective. yes, the Borg Collective. iconic, no notes, they ate 💅

---

and that's the whole page bestie!! this was borg(1) — the one and only — coming at you from the BorgBackup source, dated 2026-10-07, authored by someone going by orphan (mysterious, main character energy, we love that for them 👑). stay backed up, hydrate, and don't forget your passphrase 💖✨🦋