Skip to content

sourcefrog/conserve

Repository files navigation

Conserve: a robust backup program

https://github.com/sourcefrog/conserve/

GitHub build status crates.io Maturity: Beta

Conserve's guiding principles:

  • Safe: Conserve is written in Rust, a fast systems programming language with compile-time guarantees about types, memory safety, and concurrency. Conserve uses a conservative log-structured format.

  • Robust: If one file is corrupted in storage or due to a bug in Conserve, or if the backup is interrupted, you can still restore what was written. (Conserve doesn't need a large transaction to complete for data to be accessible.)

  • Careful: Backup data files are never touched or altered after they're written, unless you choose to purge them.

  • When you need help now: Restoring a subset of a large backup is fast, because it doesn't require reading the whole backup.

  • Always making progress: Even if the backup process or its network connection is repeatedly killed, Conserve can quickly pick up where it left off and make forward progress.

  • Ready for today: The storage format is fast and reliable on on high-latency, limited-capability, unlimited-capacity, eventually-consistent cloud object storage.

  • Fast: Conserve exploits Rust's fearless concurrency to make full use of multiple cores and IO bandwidth. (In the current release there's still room to add more concurrency.)

  • Portable: Conserve is tested on Windows, Linux (x86 and ARM), and OS X.

Quick start guide

Conserve storage is within an archive directory created by conserve init:

conserve init /backup/home.cons

conserve backup copies a source directory into a new version within the archive. Conserve copies files, directories, and (on Unix) symlinks. If the conserve backup command completes successfully (copying the whole source tree), the backup is considered complete.

conserve backup /backup/home.cons ~ --exclude /.cache

conserve diff shows what's different between an archive and a source directory. It should typically be given the same --exclude options as were used to make the backup.

conserve diff /backup/home.cons ~ --exclude /.cache

conserve versions lists the versions in an archive, whether or not the backup is complete, the time at which the backup started, and the time taken to complete it. Each version is identified by a name starting with b.

$ conserve versions /backup/home.cons
b0000                      complete   2016-11-19T07:30:09+11:00     71s
b0001                      incomplete 2016-11-20T06:26:46+11:00
b0002                      incomplete 2016-11-20T06:30:45+11:00
b0003                      complete   2016-11-20T06:42:13+11:00    286s
b0004                      complete   2016-12-01T07:08:48+11:00     84s
b0005                      complete   2016-12-18T02:43:59+11:00      4s

conserve ls shows all the files in a particular version. Like all commands that read a band from an archive, it operates on the most recent by default, and you can specify a different version using -b. (You can also omit leading zeros from the backup version.)

conserve ls -b b0 /backup/home.cons | less

conserve restore copies a version back out of an archive:

conserve restore /backup/home.cons /tmp/trial-restore

conserve validate checks the integrity of an archive:

conserve validate /backup/home.cons

conserve delete deletes specific named backups from an archive:

conserve delete /backup/home.cons -b b1

Exclusions

The --exclude GLOB option can be given to commands that operate on files, including backup, restore, ls and list-source.

A / at the start of the exclusion pattern anchors it to the top of the backup tree (not the root of the filesystem.) ** recursively matches any number of directories. *.o matches anywhere in the tree.

--exclude-from reads exclusion patterns from a file, one per line, ignoring leading and trailing whitespace, and skipping comment lines that start with a #.

The syntax is comes from the Rust globset crate.

Directories marked with CACHEDIR.TAG are automatically excluded from backups.

S3 support

From 23.9 Conserve supports storing backups in Amazon S3. AWS IAM credentials are read from the standard sources: the environment, config file, or, on EC2, the instance metadata service.

S3 support is not built by default, but can be turned on by passing cargo install --features s3.

To use this, just specify an S3 URL for the archive location. The bucket must already exist.

conserve init s3://my-bucket/
conserve backup s3://my-bucket/ ~

Files are written in the INTELLIGENT_TIERING storage class.

(This should work on API-compatible services but has not been tested; experience reports are welcome.)

Install

To build Conserve you need Rust and a C compiler that can be used by Rust.

To install the most recent release from crates.io, run

cargo install conserve

To install from a git checkout, run

cargo install -f --path .

Optional features

The following options can be enabled with --features:

  • s3: support for storing backups in Amazon S3 (or compatible services)
  • sftp: support for storing backups on SFTP servers, addressed with sftp:// URLs

Arch Linux

To install from from available AUR packages, use an AUR helper:

yay -S conserve

More documentation

Performance on Windows

Windows Defender and Windows Search Indexing can severely slow down any program that does intensive file IO, including Conserve. I recommend you exclude the backup directory from both systems.

Project status

Conserve is at a reasonable level of maturity; the format is stable and the basic features are complete. I have used it as a primary backup system for over a year. There is still room for several performance improvements and features.

The current data format (called "0.6") will be readable by future releases for at least two years.

Be aware Conserve is developed as a part-time non-commercial project and there's no guarantee of support or reliability. Bug reports are welcome but I cannot promise they will receive a resolution within any particular time frame.

Licence and non-warranty

Copyright 2012-2023 Martin Pool.

This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.