54

For a while I used Dirvish to do incremental backups of my machines, but it is slightly cumbersome to configure, and if you do not carry a copy of your configuration it can be hard to reproduce elsewhere.

I am looking for backup programs for Unix, Linux that could:

  • Incrementally update my backup
  • Create "mirror" trees like dirvish did using hardlinks (to save space)
  • Ideally with a decent UI
0

9 Answers 9

25

Try rsnapshot. It uses rsync and hardlinks and is incremental.

4
  • 3
    I should mention that I have no idea what Dirvish is or how it works. Aug 17, 2010 at 3:24
  • I think it might be GUI-less so I miss that bonus... but since you said 'Ideally' Aug 17, 2010 at 3:27
  • 3
    A GUI does not a good UI make.
    – Eli Frey
    Aug 17, 2010 at 3:32
  • 2
    i've been using rsnapshot for years
    – cmcginty
    Aug 17, 2010 at 19:35
22

This crude -but functional- script will backup everything under the sun to your external hard drive under a hard link farm. The directory name is a timestamp, and it maintains a symlink to the latest sucessful backup. Think of it as a Time Machine sans the fancy GUI.

#!/bin/sh
DATE=`/bin/date +%Y%m%d%H%M%S`
RSYNC=/usr/bin/rsync
BASE=/mnt/externalhd
TARGET=$BASE/daily
$RSYNC -av --exclude $TARGET --exclude-from=/etc/backup/rsync.exclude --link-dest=$TARGET/latest/ / $TARGET/$DATE/
touch $TARGET/$DATE/
rm $TARGET/latest
ln -s $TARGET/$DATE $TARGET/latest

Set it up creating an empty $TARGET and symlink a dummy $TARGET/latest to it. Populate /etc/backup/rsync.exclude with lost+found, tmp, var/run and everything else you need to skip during backup, or go for --include-from if it fits you better; man rsync is your friend.

Proper sanity checks, error control, remote backup and pretty GNOME GUI are left as an exercise to the reader ;-)

1
  • 1
    +1 I do something very similar to this. --link-dest for the win.
    – kbyrd
    Aug 17, 2010 at 19:37
9

The Backup-Comparison of backup tools at the Ubuntu-Stackexchange is not really Ubuntu-specific. Perhaps you get some suggestions there.

I recommend Restic. It does incremental backups by default, has a decent UI but doesn't use hardlinks.


Previously, I recommended DAR - the Disk ARchive program. It does not come with a GUI, but its config is easy to reproduce. It has great incremental backup support. It does not use hardlink mirror trees, but it has a convenient shell for navigating the filesystem view of different snapshots.

I don't recommend it anymore because Restic is faster, implements modern crypto and supports various targets (such as S3 object storage).

3
  • DAR has inconvenient restoration procedure: each incremental backup physically overrides files from previous step. So, if your file changes 7 times, it would be extracted 7 times, and 6 copies would be wasted, overridden by the 7th.
    – ayvango
    May 20, 2017 at 4:28
  • I would not recommend DAR, since DAR uses a nonstandard archive format. Also, are there a sufficient number of successful incremental restores to verify usability? Note that GNU tar advertizes to support incremental backups since 1992, but still fails to restore non-trivial deltas.
    – schily
    May 12, 2020 at 8:02
  • @schily Using a committee-standardized format isn't the only relevant criterion for judging the quality of a backup program. DAR's format is openly specified. I don't understand what you want to get at with your question. Such a number would be hard to quantify for any backup program. I don't see how GNU tar is relevant here. For example, star advertises compression support but last time I checked it failed reporting errors when compression was enabled during archive creation. Jul 10 at 18:07
8

I use backintime, which is primarily targeted towards Gnome/KDE desktops. However, it can work from the commandline as well.

I describe backintime as a backup system with "poor man's deduplication".

If you were to write your own backup script to use rsync and hardlinks, you would end up with something similar to backintime.

  • I use cron to kick off the backintime job once per night.
  • As the documentation says: The real magic is done by rsync (take snapshots and restore), diff (check if somethind changed) and cp (make hardlinks).
  • backintime can be configured with different schedules. I keep monthly backups for 1 year, weeklies for 1 month, and dailies for 1 week.
  • backintime uses hardlinks. I have 130GB worth of data, and I back this up nightly. It only uses 160GB worth of space on the second drive because of the magic of hardlinks.
  • Restoring data from the backup location is as simple as running cp /u1/backintime/20100818-000002/backup/etc/rsyslog.conf /etc/rsyslog.conf. You don't need to use the GUI.
  • On the second drive, the initial copy was expensive (since you can't do hardlinks between two different filesystems), but subsequent copies are fast.
  • I copy data from my primary filesystems to a second filesystem onto a second hot-swappable drive, and periodically rotate the secondary drive.
2
  • Surely you want the initial copy to be expensive, otherwise you don't have a backup, just another link to a single file? Of course, it's also possible that I'm missing some crucial point which makes this comment pointless :-)
    – dr-jan
    Aug 25, 2010 at 13:05
  • @Dr-jan : I agree with you. However, I think some users expect the initial copy to be fast. Aug 26, 2010 at 17:11
5

Rdiff Backup is really good http://rdiff-backup.nongnu.org/

Note that it is abandoned, with latest stable and unstable releases from 2009.

2
  • But currently unmaintained. Nov 28, 2015 at 17:31
  • It is now, as of 2021-12-07
    – toraritte
    Dec 7, 2021 at 10:31
3

I've had some success with RIBS (Rsync Incremental Backup System)

It uses rsync so hardlinks are supported and can do incremental backups hourly, daily, weekly and monthly.

However, it is a PHP script only. To set up you need to edit the settings and then set up related cronjobs. It works, but it's not the most user friendly and requires PHP.

1

I've been using epitome for about a year now for deduplicated backups of my personal data . It has a tar like interface so it's quite comfortable for a unix user and setup is a breeze, at least, on OpenBSD. You can easily cron it to backup your directories on a daily basis, and it takes care of the deduplication of your data. You basically are left with a meta-file that you can use to restore your snapshot at a later date. As I said the interface is tar-like so doing a backup is as easy as:

# epitomize -cvRf 2010-08-16-home.md /home

Note that epitome is abandoned, only partial copy of website at https://web.archive.org/web/20140908075740/https://www.peereboom.us/epitome/ remains.

1
  • It's currently experimental but, works quite well. I've been able to do full restores from arbitrary meta files and recover information that I needed, and have had 0 problems with it in ~1 year of use.
    – gabe.
    Aug 17, 2010 at 5:01
1

BackupPC sounds like it fits the bill. It manages a tree of hard links for dedupe and can backup many machines, or just the local machine.

1
  • +1 for BackupPC I use it to backup a group of servers regularly. It also has a good web-based UI.
    – dr-jan
    Aug 25, 2010 at 13:00
1

NB: Obnam's author retired it in 2017.


Lars Wirzenius's obnam:

  • Does deduplication when it backs up things, which means that backups are likely to take little space, potentially a lot more than simply hardlinking files.
  • As the backups are with deduplication, every backup is "full", with no need of having incremental backups. It simply detects that not many things have changed and only does what is needed.
  • Each backup is, effectively, a snapshot of your system, without the need to recover the last full backup and each incremental backup in turn to get the system to be restored.
  • Contrary to bup (which is another strong contender with deduplication), obnam is able to delete previous backups to save space of unnecessary backups.
  • Besides using the regular recovery methods of a backup program, there is a fuse filesystem that provides a view of obnam's backups as a plain filesystem and that can choose which snapshot/backup/generation to mount, which is super handy, as far as "user" interfaces go (given that we are in a Unix-related site, a flexible command line interface is highly valued).
  • It supports encryption as an integral part of the backups (and not as an afterthought).
  • It was written with support for remote backups in mind.

In my opinion, one serious contender for the Backup World Day (and not only that day).

2
  • "As the backups are with deduplication, every backup is "full", with no need of having incremental backups. It simply detects that not many things have changed and only does what is needed" -as it relies on previous backup versions to provide data it means that it IS an incremental backup. Apr 16, 2016 at 15:37
  • The author seems to have some plans for obnam2. FWIW, back in the days, obnam1 was unusably slow. I tested it on a $HOME where it reached only 2.4 MiB/s backup speed whereas other backup programs had no issues saturating USB 2.0 speeds (i.e. 25 MiB/s or so). Jul 10 at 17:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.