bup: It backs things up
bup is a program that backs things up. It's short for "backup." Can you believe that nobody else has named an open source program "bup" after all this time? Me neither.
Despite its unassuming name, bup is pretty cool. To give you an idea of just how cool it is, I wrote you this poem:
Bup is teh awesome What rhymes with awesome? I guess maybe possum But that's irrelevant.
Hmm. Did that help? Maybe prose is more useful after all.
Reasons bup is awesome
bup has a few advantages over other backup software:
It uses a rolling checksum algorithm (similar to rsync) to split large files into chunks. The most useful result of this is you can backup huge virtual machine (VM) disk images, databases, and XML files incrementally, even though they're typically all in one huge file, and not use tons of disk space for multiple versions.
It uses the packfile format from git (the open source version control system), so you can access the stored data even if you don't like bup's user interface.
Unlike git, it writes packfiles directly (instead of having a separate garbage collection / repacking stage) so it's fast even with gratuitously huge amounts of data.
Data is "automagically" shared between incremental backups without having to know which backup is based on which other one - even if the backups are made from two different computers that don't even know about each other. You just tell bup to back stuff up, and it saves only the minimum amount of data needed.
Even when a backup is incremental, you don't have to worry about restoring the full backup, then each of the incrementals in turn; an incremental backup acts as if it's a full backup, it just takes less disk space.
It's written in python (with some C parts to make it faster) so it's easy for you to extend and maintain.
Reasons you might want to avoid bup
This is a very early version. Therefore it will most probably not work for you, but we don't know why. It is also missing some probably-critical features.
It requires python >= 2.4, a C compiler, and an installed git version >= 188.8.131.52.
It currently only works on Linux, MacOS X >= 10.4, or Windows (with Cygwin). Patches to support other platforms are welcome.
It has almost no documentation. Not even a man page! This file is all you get for now.
check out the bup source code using git:
git clone git://github.com/apenwarr/bup
install the python 2.5 development libraries. On Debian or Ubuntu, this is: apt-get install python2.5-dev
build the python module and symlinks:
run the tests:
(The tests should pass. If they don't pass for you, stop here and send me an email.)
Try making a local backup as a tar file:
tar -cvf - /etc | bup split -n local-etc -vv
Try restoring your backup tarball:
bup join local-etc | tar -tf -
Look at how much disk space your backup took:
du -s ~/.bup
Make another backup (which should be mostly identical to the last one; notice that you don't have to specify that this backup is incremental, it just saves space automatically):
tar -cvf - /etc | bup split -n local-etc -vv
Look how little extra space your second backup used on top of the first:
du -s ~/.bup
Restore your old backup again (the ~1 is git notation for "one older than the most recent"):
bup join local-etc~1 | tar -tf -
get a list of your previous backups:
GIT_DIR=~/.bup git log local-etc
make a backup on a remote server (which must already have the 'bup' command somewhere in the PATH, and be accessible via ssh; make sure to replace SERVERNAME with the actual hostname of your server):
tar -cvf - /etc | bup split -r SERVERNAME: -n local-etc -vv
try restoring the remote backup tarball:
bup join -r SERVERNAME: local-etc | tar -tf -
try using the new (slightly experimental) 'bup index' and 'bup save' style backups, which bypass 'tar' but have some missing features (see "Things that are stupid" below):
bup index -uv /etc bup save -n local-etc /etc
do it again and see how fast an incremental backup can be:
bup index -uv /etc bup save -n local-etc /etc
(You can also use the "-r SERVERNAME:" option to 'bup save', just like with 'bup split' and 'bup join'. The index itself is always local, so you don't need -r there.)
That's all there is to it!
How it works
bup stores its data in a git-formatted repository. Unfortunately, git itself doesn't actually behave very well for bup's use case (huge numbers of files, files with huge sizes, retaining file permissions/ownership are important), so we mostly don't use git's code except for a few helper programs. For example, bup has its own git packfile writer written in python.
Basically, 'bup split' reads the data on stdin (or from files specified on the command line), breaks it into chunks using a rolling checksum (similar to rsync), and saves those chunks into a new git packfile. There is one git packfile per backup.
When deciding whether to write a particular chunk into the new packfile, bup first checks all the other packfiles that exist to see if they already have that chunk. If they do, the chunk is skipped.
git packs come in two parts: the pack itself (.pack) and the index (.idx). The index is pretty small, and contains a list of all the objects in the pack. Thus, when generating a remote backup, we don't have to have a copy of the packfiles from the remote server: the local end just downloads a copy of the server's index files, and compares objects against those when generating the new pack, which it sends directly to the server.
The "-n" option to 'bup split' and 'bup save' is the name of the backup you want to create, but it's actually implemented as a git branch. So you can do cute things like checkout a particular branch using git, and receive a bunch of chunk files corresponding to the file you split.
If you use '-b' or '-t' or '-c' instead of '-n', bup split will output a list of blobs, a tree containing that list of blobs, or a commit containing that tree, respectively, to stdout. You can use this to construct your own scripts that do something with those values.
The bup index:
'bup index' walks through your filesystem and updates a file (whose name is, by default, ~/.bup/bupindex) to contain the name, attributes, and an optional git SHA1 (blob id) of each file and directory.
'bup save' basically just runs the equivalent of 'bup split' a whole bunch of times, once per file in the index, and assembles a git tree that contains all the resulting objects. Among other things, that makes 'git diff' much more useful (compared to splitting a tarball, which is essentially a big binary blob). However, since bup splits large files into smaller chunks, the resulting tree structure doesn't exactly correspond to what git itself would have stored. Also, the tree format used by 'bup save' will probably change in the future to support storing file ownership, more complex file permissions, and so on.
If a file has previously been written by 'bup save', then its git blob/tree id is stored in the index. This lets 'bup save' avoid reading that file to produce future incremental backups, which means it can go very fast unless a lot of files have changed.
Things that are stupid for now but which we'll fix later
Help with any of these problems, or others, is very, very welcome. Join the mailing list (see below) if you'd like to help.
'bup save' doesn't know about file metadata.
That means we aren't saving file attributes, mtimes, ownership, hard links, MacOS resource forks, etc. Clearly this needs to be improved.
There's no 'bup restore' yet.
'bup save' saves files in the standard git 'tree of blobs' format, so you could then "restore" the files using something like 'git checkout'. But that's a git command, not a bup command, so it's hard to explain and doesn't support retrieving objects from a remote bup server without first fetching and packing an entire (possibly huge) pack, which could be very slow. Also, like 'bup save', you would need extra features in order to properly restore file metadata. And files that bup has split into chunks would need to be recombined somehow.
'bup index' is slower than it should be.
It's still rather fast: it can iterate through all the filenames on my 600,000 file filesystem in a few seconds. But sometimes you just want to change a filename or two, so this is needlessly slow. There should be a way to binary search through the file list rather than always going through it sequentially. And if you only add a couple of filenames, there's no need to rewrite the entire index; just leave the new files in a second "extra index" file or something.
bup could use inotify for really efficient incremental backups.
You could even have your system doing "continuous" backups: whenever a file changes, we immediately send an image of it to the server. We could give the continuous-backup process a really low CPU and I/O priority so you wouldn't even know it was running.
bup currently has no features that prune away old backups.
Because of the way the packfile system works, backups become "entangled" in weird ways and it's not actually possible to delete one pack (corresponding approximately to one backup) without risking screwing up other backups.
git itself has lots of ways of optimizing this sort of thing, but its methods aren't really applicable here; bup packfiles are just too huge. We'll have to do it in a totally different way. There are lots of options. For now: make sure you've got lots of disk space :)
bup has never been tested on anything but Linux, MacOS, and Linux+Cygwin.
There's nothing that makes it inherently non-portable, though, so that's mostly a matter of someone putting in some effort. (For a "native" Windows port, the most annoying thing is the absence of ssh in a default Windows installation.)
bup has no GUI. Actually, that's not stupid, but you might consider it a limitation. There are a bunch of Linux GUI backup programs; someday I expect someone will adapt one of them to use bup.
How you can help
bup is a work in progress and there are many ways it can still be improved. If you'd like to contribute patches, ideas, or bug reports, please join the bup mailing list.
You can find the mailing list archives here:
and you can subscribe by sending a message to: