Horcrux: A Wrapper for Duplicity

14 July 2011

Get the code

Duplicity is a backup program that allows you to store encrypted backups offsite while also allowing for incremental updates.

I take the Voldemort approach to backups, backing up mostly the same stuff to several different places, hence using similar options with Duplicity. To make this use case easier, I wrote Horcrux, a wrapper for Duplicity’s most commonly-used options. It also includes functionality to check a restored backup to periodically ensure your backups are working.

After a simple setup, you’ll be able to run

horcrux auto offsite1

to backup selected files and directories as part of your offsite1 backup. If you want to add a new backup, the setup is minimal, and you can run

horcrux auto offsite2

Both backups keep separate caches and use separate log files. They can backup different files and directories, and to different locations.


Download Horcrux from GitHub.

Place it in your PATH somewhere, usually ~/bin.

First run

Run horcrux with no arguments. It’ll tell you that it’s created a file, ~/.horcrux/horcrux.conf. This is the global configuration file, which’ll be consulted for each backup. Open this file in your favourite text editor and edit the options (perhaps only the source directory and the encryption key ID).

Note that if you add a sign key (a good idea) and you use a subkey, you should give the subkey’s ID, not the ID of your main key. If you give the main key, the backup will still work (since gpg will choose your signing key automatically), but upon restore Duplicity will complain that the backup files have been signed with a different key.

The source directory is the directory containing the files and directories to be backed up. Each backup will then consult an exclude file, which will tell Duplicity which files should be ignored, and which should be part of the backup.

Basic usage

Horcrux is run from the command line with the following syntax:

horcrux [options] operation backup_name [restore path]

options and restore path are optional, but you must always tell Horcrux what you want it to do (the operation), and the name of the backup. The backup name can use most characters (letters, numbers, etc.) but should not use hyphens.

Configuration setup

Each backup run by Horcrux has a name and an exclude file. For example, let’s create a backup called offsite1. Create the following two files in the ~/.horcrux directory:

The config file should contain at least one variable: the destination path. If you’re backing up to a directory on a remote server somewhere, offsite1-config might look like this:


The URL follows the definitions allowed as described in Duplicity’s man page. You can also override variables set in the global configuration file; this is useful to be able to change the source directory or encryption key for one specific backup, for example.

The offsite1-exclude file also follows the syntax explains in Duplicity’s man page. Assuming the source directory is $HOME/, it looks something like this:

+ /Users/username/.emacs
+ /Users/username/emacs
+ /Users/username/bin
+ /Users/username/.git*
+ /Users/username/.ssh
+ /Users/username/.zsh*
+ /Users/username/.horcrux
+ /Users/username/Documents/**/*.md
+ /Users/username/Documents/**/*.css

Duplicity likes the exclude syntax to use absolute path names, hence the redundancy here. Items starting with a + are to be included. Those without are to be excluded. The above lines will cause Duplicity to ignore any .DS_Store files (used by Mac OS X to store icon information in the Finder), then include several files and directories. Any .css files in any directory within my Documents directory will get backed up, as will the .horcrux directory, my emacs directory, and several files and directories starting with .zsh.

Duplicity stores a local cache of changes in ~/.cache/duplicity by default. If your source directory is the root of your home directory (the Horcrux default), then make sure you exclude this directory. I add the following line to some of my exclude files:


This will ensure that Duplicity excludes files and directories with the word cache in it, which also includes ~/.cache.

Your first backup

Now you should be ready to make your first backup. Duplicity first produces a full backup, and from then on can make incremental backups (i.e., only uploading the changed bits, to save time and space). To do this automatically, run

horcrux auto offsite1

All output sent to the screen will also be put in a log file in ~/.horcrux with the appropriate name. It’ll be compressed with bzip2, or pbzip2 if available.

If at any point you want to check the names of options to use, you can run horcrux help or horcrux -h.

Zero configuration for local drives

You might have a few external USB drives to use for backups, each of which have the directory backup inside them, so for each one you’d add




to its configuration file1 (assuming you’re using the same name for both the drive and the backup). In this case, there’s no need to have a configuration file at all. Horcrux will automatically look for such a directory, and if it exists, will use this as the destination path2.

For example, if you connect a drive called offsite2, and add a directory backup, you can then add an exclude file offsite2-exclude to your Horcrux directory, and the backup files will automatically be sent to /Volumes/offsite2/backup.


Horcrux supports the following operations:

Full backup if no filesets or last full backup is old (default 360 days), else incremental.
Check md5 hashes of files in a restored fileset against local fileset.
Requires md5deep.
Clean local caches, removing files no longer needed.
Maps to duplicity cleanup --extra-clean --force.
Perform full backup.
Perform incremental backup.
List files backed up in fileset.
Maps to duplicity list-current-files.
Remove filesets, leaving latest n (default 3) full+inc filesets.
Maps to duplicity remove-all-but-n-full n --force.
Restore certain files/directories or complete filesets.
Maps to duplicity restore.
Check collection status on fileset.
Maps to duplicity collection-status.
Verify a backup. Not as thorough as check.
Maps to duplicity verify.

Running full will force a full backup; if auto is passed, Duplicity will choose this method if it doesn’t find a “recent” fileset at the backup destination3, otherwise it’ll run an incremental backup.

Restoring a backup

You can restore certain files from your backups, or certain directories, or the entire fileset. To do the latter, create a directory where you want to restore to, and then run

horcrux restore offsite1 /Some/restore/directory

The backup offsite1 will then be restored to /Some/restore/directory. You’ll be prompted for your GPG passphrase by Duplicity.

If you want to restore a certain file or directory, use the -f option:

horcrux -f somefile.txt restore offsite1 ~/restore_directory/somefile.txt

The file ~/restore_directory/somefile.txt will be written with the contents of somefile.txt from your backup. This file named with -f is relative to the source directory of the backup.

Restoring from a certain time

You can use the -t option to restore a file or directory from a certain time (assuming this time is between the first backup date and now). The following syntax is used:

YYYY-MM-DD, or interval using characters s, m, h, D, W, M, or Y. 1h78m, etc.

For example, a file could be restored from 2 weeks ago using

horcrux -t 2W -f somefile.txt restore offsite1 ~/restore_directory/somefile.txt

Again, this time format is shown in the output from horcrux help.

Checking a backup

This is a feature that isn’t provided by duplicity. It ensures that a restored backup is exactly the same as the local files that the backup is based on. Alternatively, it’ll show you exactly what files have been changed. It’s useful if you want to periodically check your backups.

First, you need to restore your backup to a directory. Then, run the command

horcrux check offsite1 /Some/restore/directory

[md5deep] will run through all files in the restore directory, and then compare the md5 hashes of each file with its equivalent in the backup’s source directory. If just one byte of a file is different its hash will also be different, and this will be recorded. The list of changed files will be stored in ~/.horcrux/offsite1-changed-hashes.log.bz2.

Removing old backup filesets

Performing a full backup followed by a few incremental backups produces the following structure on the backup side:

- full 1
    - incremental 1
    - incremental 2

If we then force a full backup then perform a few incremental backups, we’ll end up with

- full 1
    - incremental 1
    - incremental 2
- full 2
    - incremental 3
    - incremental 4

Future incremental backups (included those started by the auto operation) will be attached to full 2. If space is at a premium, the old backup chains can be removed using the remove operation. By default it’ll leave 3 chains, but this can be changed either globally in ~/.horcrux/horcrux.conf or for a particular backup using the -c option. If we run

horcrux -c1 remove offsite1

for the above backup, the result will be

- full 2
    - incremental 3
    - incremental 4


Horcrux supports a number of options that can be used to change an option from the default. For example, perhaps you want to increase the verbosity of output from the default of 5 to the maximum of 9. You would run

horcrux -v9 auto offsite1

There are a number of possible options:

If you run horcrux help then you’ll be reminded of these options, together with their current values (as set in ~/.horcrux/horcrux.conf).

Frequently asked questions

I want to check what will happen before I run a backup

Use the dry run option, giving a command like horcrux -n auto offsite1.

You could also use the paramtest operation, to very quickly check what parameters will be passed to Duplicity.
I want to backup to the same place, but with different names

I backup to a home server. Locally it has an address, but outside of my home network it has another address provided by DynDNS.

Use multiple configuration files, one for each location, with names like offsite3-local-config and offsite3-remote-config. In each one specify the correct destination_path.

Since it’s actually the same backup, it’ll just use the same exclude file as usual, offsite3-exclude.

Run the backup as either horcrux auto offsite3-local or horcrux auto offsite3-remote. The correct configuration file will be loaded.

All output will go into the same log file as usual, and the same local cache will be used.

I want to backup locally first, then upload somewhere

Create your backup-config file as usual, but also create a file called backup-local-config (for example). In this second file, make sure the backup destination is set to a local drive; the former file can still have a remote path.

Backup using horcrux auto backup-local, and then use scp, rsync, or whatever, to copy these files to the remote destination. You can run horcrux auto backup from then on. I often run something like this to maximise my upload bandwidth:

cd local_backup_dir
find . -name '*.gpg' -print0 | xargs -0I{} -n1 -P8 rsync -avhP {} \

Reporting bugs

Please send me email to the usual address.


Change log

Add support for passing extra parameters to Duplicity.
Thanks to Roland Wirth for the suggestion.
Add support for gpg-agent with -a and use_agent in horcrux.conf.
Check for backup directory in /media/drive_name as well as /Volumes/drive_name.
Produce more output with paramtest operation (for debugging).
  1. This feature assumes Mac OS X or another unix system where files are mounted in /media, such as Ubuntu. 

  2. See variable backup_basename in horcrux.conf

  3. See variable full_if_old in horcrux.conf. By default it’s 360 days.