2 bup: It backs things up
3 =======================
5 bup is a program that backs things up. It's short for "backup." Can you
6 believe that nobody else has named an open source program "bup" after all
9 Despite its unassuming name, bup is pretty cool. To give you an idea of
10 just how cool it is, I wrote you this poem:
13 What rhymes with awesome?
15 But that's irrelevant.
17 Hmm. Did that help? Maybe prose is more useful after all.
20 Reasons bup is awesome
21 ----------------------
23 bup has a few advantages over other backup software:
25 - It uses a rolling checksum algorithm (similar to rsync) to split large
26 files into chunks. The most useful result of this is you can backup huge
27 virtual machine (VM) disk images, databases, and XML files incrementally,
28 even though they're typically all in one huge file, and not use tons of
29 disk space for multiple versions.
31 - It uses the packfile format from git (the open source version control
32 system), so you can access the stored data even if you don't like bup's
35 - Unlike git, it writes packfiles *directly* (instead of having a separate
36 garbage collection / repacking stage) so it's fast even with gratuitously
37 huge amounts of data. bup's improved index formats also allow you to
38 track far more filenames than git (millions) and keep track of far more
39 objects (hundreds or thousands of gigabytes).
41 - Data is "automagically" shared between incremental backups without having
42 to know which backup is based on which other one - even if the backups
43 are made from two different computers that don't even know about each
44 other. You just tell bup to back stuff up, and it saves only the minimum
45 amount of data needed.
47 - You can back up directly to a remote bup server, without needing tons of
48 temporary disk space on the computer being backed up. And if your backup
49 is interrupted halfway through, the next run will pick up where you left
50 off. And it's easy to set up a bup server: just install bup on any
51 machine where you have ssh access.
53 - Bup can use "par2" redundancy to recover corrupted backups even if your
54 disk has undetected bad sectors.
56 - Even when a backup is incremental, you don't have to worry about
57 restoring the full backup, then each of the incrementals in turn; an
58 incremental backup *acts* as if it's a full backup, it just takes less
61 - You can mount your bup repository as a FUSE filesystem and access the
62 content that way, and even export it over Samba.
64 - It's written in python (with some C parts to make it faster) so it's easy
65 for you to extend and maintain.
68 Reasons you might want to avoid bup
69 -----------------------------------
71 - This is a very early version. Therefore it will most probably not work
72 for you, but we don't know why. It is also missing some
73 probably-critical features.
75 - It requires python >= 2.4, a C compiler, and an installed git version >=
78 - It currently only works on Linux, MacOS X >= 10.4,
79 NetBSD, Solaris, or Windows (with Cygwin). Patches to support
80 other platforms are welcome.
90 - Check out the bup source code using git:
92 git clone git://github.com/apenwarr/bup
94 - Install the needed python libraries (including the development
95 libraries). On Debian or Ubuntu, this is usually:
96 apt-get install python2.6-dev python-fuse
97 apt-get install python-pyxattr python-pylibacl
99 Substitute python2.5-dev or python2.4-dev if you have an older system.
101 Or on newer Debian/Ubuntu versions, you can try this:
103 apt-get build-dep bup
105 - Build the python module and symlinks:
113 (The tests should pass. If they don't pass for you, stop here and send
120 Binary packages of bup are known to be built for the following OSes:
123 http://packages.debian.org/search?searchon=names&keywords=bup
125 http://packages.ubuntu.com/search?searchon=names&keywords=bup
126 - pkgsrc (NetBSD, Dragonfly, and others)
127 http://pkgsrc.se/sysutils/bup
128 http://cvsweb.netbsd.org/bsdweb.cgi/pkgsrc/sysutils/bup/
134 - Try making a local backup as a tar file:
136 tar -cvf - /etc | bup split -n local-etc -vv
138 - Try restoring your backup tarball:
140 bup join local-etc | tar -tf -
142 - Look at how much disk space your backup took:
146 - Make another backup (which should be mostly identical to the last one;
147 notice that you don't have to *specify* that this backup is incremental,
148 it just saves space automatically):
150 tar -cvf - /etc | bup split -n local-etc -vv
152 - Look how little extra space your second backup used on top of the first:
156 - Restore your old backup again (the ~1 is git notation for "one older than
159 bup join local-etc~1 | tar -tf -
161 - Get a list of your previous backups:
163 GIT_DIR=~/.bup git log local-etc
165 - Make a backup on a remote server (which must already have the 'bup' command
166 somewhere in the server's PATH (see /etc/profile, etc/environment,
167 ~/.profile, or ~/.bashrc), and be accessible via ssh.
168 Make sure to replace SERVERNAME with the actual hostname of your server):
170 tar -cvf - /etc | bup split -r SERVERNAME: -n local-etc -vv
172 - Try restoring the remote backup tarball:
174 bup join -r SERVERNAME: local-etc | tar -tf -
176 - Try using the new (slightly experimental) 'bup index' and 'bup save'
177 style backups, which bypass 'tar' but have some missing features (see
178 "Things that are stupid" below):
181 bup save -n local-etc /etc
183 - Do it again and see how fast an incremental backup can be:
186 bup save -n local-etc /etc
188 (You can also use the "-r SERVERNAME:" option to 'bup save', just like
189 with 'bup split' and 'bup join'. The index itself is always local,
190 so you don't need -r there.)
192 That's all there is to it!
198 - FreeBSD's default 'make' command doesn't like bup's Makefile. In order to
199 compile the code, run tests and install bup, you need to install GNU Make
200 from the port named 'gmake' and use its executable instead in the commands
201 seen above. (i.e. 'gmake test' runs bup's test suite)
203 - Python's development headers are automatically installed with the 'python'
204 port so there's no need to install them separately.
206 - To use the 'bup fuse' command, you need to install the fuse kernel module
207 from the 'fusefs-kmod' port in the 'sysutils' section and the libraries from
208 the port named 'py-fusefs' in the 'devel' section.
210 - The 'par2' command can be found in the port named 'par2cmdline'.
212 - In order to compile the documentation, you need pandoc which can be found in
213 the port named 'hs-pandoc' in the 'textproc' section.
216 Notes on NetBSD/pkgsrc
217 ----------------------
219 - See pkgsrc/sysutils/bup, which should be the most recent stable
220 release and includes man pages. It also has a reasonable set of
221 dependencies (git, par2, py-fuse-bindings).
223 - The "fuse-python" package referred to is hard to locate, and is a
224 separate tarball for the python language binding distributed by the
225 fuse project on sourceforge. It is available as
226 pkgsrc/filesystems/py-fuse-bindings and on NetBSD 5, "bup fuse"
229 - "bup fuse" presents every directory/file as inode 0. The directory
230 traversal code ("fts") in NetBSD's libc will interpret this as a
231 cycle and error out, so "ls -R" and "find" will not work.
233 - It is not clear if extended attribute and POSIX acl support does
242 bup stores its data in a git-formatted repository. Unfortunately, git
243 itself doesn't actually behave very well for bup's use case (huge numbers of
244 files, files with huge sizes, retaining file permissions/ownership are
245 important), so we mostly don't use git's *code* except for a few helper
246 programs. For example, bup has its own git packfile writer written in
249 Basically, 'bup split' reads the data on stdin (or from files specified on
250 the command line), breaks it into chunks using a rolling checksum (similar to
251 rsync), and saves those chunks into a new git packfile. There is one git
254 When deciding whether to write a particular chunk into the new packfile, bup
255 first checks all the other packfiles that exist to see if they already have that
256 chunk. If they do, the chunk is skipped.
258 git packs come in two parts: the pack itself (*.pack) and the index (*.idx).
259 The index is pretty small, and contains a list of all the objects in the
260 pack. Thus, when generating a remote backup, we don't have to have a copy
261 of the packfiles from the remote server: the local end just downloads a copy
262 of the server's *index* files, and compares objects against those when
263 generating the new pack, which it sends directly to the server.
265 The "-n" option to 'bup split' and 'bup save' is the name of the backup you
266 want to create, but it's actually implemented as a git branch. So you can
267 do cute things like checkout a particular branch using git, and receive a
268 bunch of chunk files corresponding to the file you split.
270 If you use '-b' or '-t' or '-c' instead of '-n', bup split will output a
271 list of blobs, a tree containing that list of blobs, or a commit containing
272 that tree, respectively, to stdout. You can use this to construct your own
273 scripts that do something with those values.
277 'bup index' walks through your filesystem and updates a file (whose name is,
278 by default, ~/.bup/bupindex) to contain the name, attributes, and an
279 optional git SHA1 (blob id) of each file and directory.
281 'bup save' basically just runs the equivalent of 'bup split' a whole bunch
282 of times, once per file in the index, and assembles a git tree
283 that contains all the resulting objects. Among other things, that makes
284 'git diff' much more useful (compared to splitting a tarball, which is
285 essentially a big binary blob). However, since bup splits large files into
286 smaller chunks, the resulting tree structure doesn't *exactly* correspond to
287 what git itself would have stored. Also, the tree format used by 'bup save'
288 will probably change in the future to support storing file ownership, more
289 complex file permissions, and so on.
291 If a file has previously been written by 'bup save', then its git blob/tree
292 id is stored in the index. This lets 'bup save' avoid reading that file to
293 produce future incremental backups, which means it can go *very* fast unless
294 a lot of files have changed.
297 Things that are stupid for now but which we'll fix later
298 --------------------------------------------------------
300 Help with any of these problems, or others, is very welcome. Join the
301 mailing list (see below) if you'd like to help.
303 - 'bup save' and 'bup restore' have immature metadata support.
305 On the plus side, they actually do have support now, but it's new,
306 and not remotely as well tested as tar/rsync/whatever's. If you'd
307 like to help test, please do; something like 'rsync -niaHAX src/
308 restore/' may be useful on that front.
310 In addition, at the moment, if any strip or graft-style options
311 are specified to 'bup save', then no metadata will be written for
312 the root directory. That's obviously less than ideal.
314 - 'bup index' is slower than it should be.
316 It's still rather fast: it can iterate through all the filenames on my
317 600,000 file filesystem in a few seconds. But it still needs to rewrite
318 the entire index file just to add a single filename, which is pretty
319 nasty; it should just leave the new files in a second "extra index" file
322 - bup could use inotify for *really* efficient incremental backups.
324 You could even have your system doing "continuous" backups: whenever a
325 file changes, we immediately send an image of it to the server. We could
326 give the continuous-backup process a really low CPU and I/O priority so
327 you wouldn't even know it was running.
329 - bup currently has no features that prune away *old* backups.
331 Because of the way the packfile system works, backups become "entangled"
332 in weird ways and it's not actually possible to delete one pack
333 (corresponding approximately to one backup) without risking screwing up
336 git itself has lots of ways of optimizing this sort of thing, but its
337 methods aren't really applicable here; bup packfiles are just too huge.
338 We'll have to do it in a totally different way. There are lots of
339 options. For now: make sure you've got lots of disk space :)
341 - bup has never been tested on anything but Linux, MacOS, and Windows+Cygwin.
343 There's nothing that makes it *inherently* non-portable, though, so
344 that's mostly a matter of someone putting in some effort. (For a
345 "native" Windows port, the most annoying thing is the absence of ssh in
346 a default Windows installation.)
348 - bup needs better documentation.
350 According to a recent article about git in Linux Weekly News
351 (https://lwn.net/Articles/380983/), "it's a bit short on examples and
352 a user guide would be nice." Documentation is the sort of thing that
353 will never be great unless someone from outside contributes it (since
354 the developers can never remember which parts are hard to understand).
356 - bup is "relatively speedy" and has "pretty good" compression.
358 ...according to the same LWN article. Clearly neither of those is good
359 enough. We should have awe-inspiring speed and crazy-good compression.
360 Must work on that. Writing more parts in C might help with the speed.
364 Actually, that's not stupid, but you might consider it a limitation.
365 There are a bunch of Linux GUI backup programs; someday I expect someone
366 will adapt one of them to use bup.
372 bup has an extensive set of man pages. Try using 'bup help' to get
373 started, or use 'bup help SUBCOMMAND' for any bup subcommand (like split,
374 join, index, save, etc.) to get details on that command.
376 For further technical details, please see ./DESIGN.
382 bup is a work in progress and there are many ways it can still be improved.
383 If you'd like to contribute patches, ideas, or bug reports, please join the
386 You can find the mailing list archives here:
388 http://groups.google.com/group/bup-list
390 and you can subscribe by sending a message to:
392 bup-list+subscribe@googlegroups.com