43

I'm looking for a list which paths/files are safe to exclude for a full system/home backup.

Considering that I have a list of installed packages.

  • /home/*/.thumbnails
  • /home/*/.cache
  • /home/*/.mozilla/firefox/*.default/Cache
  • /home/*/.mozilla/firefox/*.default/OfflineCache
  • /home/*/.local/share/Trash

  • /home/*/.gvfs/

  • /tmp/

  • /var/tmp/
  • not real folders but can cause severe problems when 'restoring'
    • /dev
    • /proc
    • /sys

What about...

  • /var/ in general?
  • /var/backups/ - can get quite large
  • /var/log/ - does not require much space and can help for later comparison
  • /lost+found/
Isaiah
  • 58,486
  • 28
  • 133
  • 145
seb
  • 2,331
  • 8
  • 24
  • 24
  • This is an (very) related answer worth looking at (assuming this isn't closed as a duplicate) : http://askubuntu.com/questions/5596/how-to-clean-caches-in-my-homedir/5610#5610 – Scaine Mar 03 '11 at 21:44
  • I am not sure about `/home/*/.local/share/Trash`. I already restored files from trash folder on backup. It was a mistaken deletion, noticed and undone after I made backups - and on restoring backups files were still in trash. – reducing activity Jan 29 '17 at 11:59
  • ~/.local/share/Trash/* seems to be universal, but is by user. – SDsolar Jul 30 '17 at 05:01
  • Note that this was posted in 2011. Your mileage may vary in 2017. – SDsolar Jul 30 '17 at 05:08

3 Answers3

40

When I rsync my system to a backup partition, I exclude these:

--exclude=/dev/* \
--exclude=/home/*/.gvfs \
--exclude=/home/*/.mozilla/firefox/*/Cache \
--exclude=/home/*/.cache/chromium \
--exclude=/home/*/.thumbnails \
--exclude=/media/* \
--exclude=/mnt/* \
--exclude=/proc/* \
--exclude=/sys/* \
--exclude=/tmp/* \
--exclude=/home/*/.local/share/Trash \
--exclude=/etc/fstab \
--exclude=/var/run/* \
--exclude=/var/lock/* \
--exclude=/lib/modules/*/volatile/.mounted \
--exclude=/var/cache/apt/archives/* \

This way I am able to boot into the backup partition the same way I can boot to the original one.

So to sum up, I would suggest

  • not excluding /{dev,proc,media,...} themselves, just their contents

  • excluding /var/{run,lock}, and especially the big /var/cache/apt/archives/

arrange
  • 14,727
  • 4
  • 42
  • 32
  • 2
    You want to exclude ALL of /dev. – psusi Mar 01 '11 at 03:33
  • @psusi - Yes, you are right. I'm editing my post. Thanks. – arrange Mar 03 '11 at 21:20
  • 2
    I would recommend using the -x option with rsync for backups. That way, it will not cross filesystem boundaries, which will exclude things like /dev /proc /sys /mnt and /media. You still have to exclude /home/*/.gvfs. – Azendale Jul 07 '11 at 16:04
  • 3
    @Azendale: might be. On the other hand, when you use something like `--exclude=/proc/*`, it will keep the directory /proc on the backup (which is needed if you want to boot the mirror), but not the contents of it. – arrange Jul 08 '11 at 16:35
  • or Opera users , I would add `--exclude=/home/*/.opera/cache` as well :) – Grzegorz Wierzowiecki Aug 13 '13 at 18:26
  • 2
    maybe a naive question, but why `/etc/fstab` has been excluded as well? – Paolo Dec 11 '13 at 21:33
  • @Guandalino: `/etc/fstab` should be edited manually if you want to be able to boot into the backup partition. Otherwise it's not necessary to exclude that file. – arrange Dec 18 '13 at 20:06
  • What flags would you suggest for rsync? Or better what would be the full command to do a backup like this over ssh to a remote machine? – MemphiZ Apr 16 '16 at 18:48
  • Firefox [stores it's cache](http://askubuntu.com/q/616551/14601) in `/home/*/.cache/mozilla/*` these days instead of `/home/*/.mozilla/firefox/*/Cache/` – gene_wood Apr 04 '17 at 18:07
  • 1
    The huge files are /var/log/kern* and var/log/messages* - are they safe to exclude? – SDsolar Aug 13 '17 at 23:25
6

I'll recomend to exclude all /usr (except fort /usr/local/) and have a backup of dpkg --get-selections

In my opinion a Full backup isn't really usefull. I rather prefer backup data and preferences and then restore packages from the official repositories.

But is my preference...

/usr contains mainly static binary content from distribution packages. Reinstalling a package will restore /usr contents. Also /bin and /sbin contains binaries from distribution packages.

/usr/local contains manually installed packages (out of distribuition packages). so this is worth to backup.

theist
  • 832
  • 6
  • 10
  • 2
    Very interesting approach. It will yield a significantly smaller backup, but require way more trouble (time, internet bandwidth) when restoring. A compromise that may be worth in some use cases. Questions: how about `/lib`, the same applies? And how do you *restore* such package selection? – MestreLion Apr 04 '13 at 11:44
  • 2
    Yes, I think you can also exclude `/lib`. I'd add also that instead of `dpkg --get-selections` it's maybe better to use `apt-mark`, as said here: http://askubuntu.com/questions/101931/restoring-all-data-and-dependencies-from-dpkg-set-selections – gerlos May 04 '15 at 14:53
2

I also read all of the answers and all the other related questions when I was configuring my full system backup.

My 2 cents would be not to exclude anything but only when you are doing a full system backup of an offline system.

Let me elaborate now: If you backing up the system from within itself, then you may exclude some directories as others have mentioned but this is not a good way in my opinion as it can introduce some bugs later on when restoring the system. For example, if you exclude /tmp directory, then after restoring it, /tmp won't be there and system will create one and this can have stuck login screens due to chmod & chown issues on /tmp. Also not backing up ~/.cache directory can result in breaking up database cache of Ubuntu Software Centre.

What I do is use a live CD and use LuckyBackup and backup the entire system without omitting anything. Now when restoring, I chose 'delete files not present in the source'. This way you will have an entire system snapshot restored exactly to the state when you created the backup (kinda like snapshotting a VM).

Gaurav Kumar
  • 173
  • 4
  • 2
    Many files are not supposed to be saved because they are **not** real files, but rather *pseudo-files*. You may want want to save caches, but don't save the contents of `/proc`, `/dev`, `sys`. Their contents are created on boot. – admirabilis Aug 23 '14 at 00:07
  • 1
    @TeresaeJunior When this approach is followed and the backup is done from an *offline* system, the contents of these not-real files are already empty. – Gaurav Kumar Aug 26 '14 at 12:45
  • Yes, you're right, sorry! I didn't pay attention to the LiveCD you mentioned. – admirabilis Aug 26 '14 at 16:33
  • I'm afraid these systems have to run 24x7 so don't have that luxury. I have a great exclude list but keep getting hung up on 4G /var/log/kern* and /var/log/messages* - I would like to exclude them but nobody talks to that. How about excluding /var/log/* totally? – SDsolar Aug 13 '17 at 23:28