1 bup: It backs things up
2 =======================
4 bup is a program that backs things up. It's short for "backup." Can you
5 believe that nobody else has named an open source program "bup" after all
8 Despite its unassuming name, bup is pretty cool. To give you an idea of
9 just how cool it is, I wrote you this poem:
12 What rhymes with awesome?
14 But that's irrelevant.
16 Hmm. Did that help? Maybe prose is more useful after all.
19 Reasons bup is awesome
20 ----------------------
22 bup has a few advantages over other backup software:
24 - It uses a rolling checksum algorithm (similar to rsync) to split large
25 files into chunks. The most useful result of this is you can backup huge
26 virtual machine (VM) disk images, databases, and XML files incrementally,
27 even though they're typically all in one huge file, and not use tons of
28 disk space for multiple versions.
30 - It uses the packfile format from git (the open source version control
31 system), so you can access the stored data even if you don't like bup's
34 - Unlike git, it writes packfiles *directly* (instead of having a separate
35 garbage collection / repacking stage) so it's fast even with gratuitously
36 huge amounts of data. bup's improved index formats also allow you to
37 track far more filenames than git (millions) and keep track of far more
38 objects (hundreds or thousands of gigabytes).
40 - Data is "automagically" shared between incremental backups without having
41 to know which backup is based on which other one - even if the backups
42 are made from two different computers that don't even know about each
43 other. You just tell bup to back stuff up, and it saves only the minimum
44 amount of data needed.
46 - You can back up directly to a remote bup server, without needing tons of
47 temporary disk space on the computer being backed up. And if your backup
48 is interrupted halfway through, the next run will pick up where you left
49 off. And it's easy to set up a bup server: just install bup on any
50 machine where you have ssh access.
52 - Bup can use "par2" redundancy to recover corrupted backups even if your
53 disk has undetected bad sectors.
55 - Even when a backup is incremental, you don't have to worry about
56 restoring the full backup, then each of the incrementals in turn; an
57 incremental backup *acts* as if it's a full backup, it just takes less
60 - You can mount your bup repository as a FUSE filesystem and access the
61 content that way, and even export it over Samba.
63 - It's written in python (with some C parts to make it faster) so it's easy
64 for you to extend and maintain.
67 Reasons you might want to avoid bup
68 -----------------------------------
70 - This is a very early version. Therefore it will most probably not work
71 for you, but we don't know why. It is also missing some
72 probably-critical features.
74 - It requires python >= 2.5, a C compiler, and an installed git
75 version >= 1.5.3.1. It also requires par2 if you want fsck to be
76 able to generate the information needed to recover from some types
79 - It currently only works on Linux, MacOS X >= 10.4,
80 NetBSD, Solaris, or Windows (with Cygwin). Patches to support
81 other platforms are welcome.
83 - Any items in "Things that are stupid" below.
93 - Check out the bup source code using git:
95 git clone git://github.com/bup/bup
97 - Install the required python libraries (including the development
100 On very recent Debian/Ubuntu versions, this may be sufficient (run
103 apt-get build-dep bup
105 Otherwise try this (substitute python2.5-dev or python2.6-dev if
106 you have an older system):
108 apt-get install python2.7-dev python-fuse
109 apt-get install python-pyxattr python-pylibacl
110 apt-get install linux-libc-dev
111 apt-get install acl attr
112 apt-get install python-tornado # optional
114 On CentOS (for CentOS 6, at least), this should be sufficient (run
117 yum groupinstall "Development Tools"
118 yum install python python-devel
119 yum install fuse-python pyxattr pylibacl
120 yum install perl-Time-HiRes
122 In addition to the default CentOS repositories, you may need to add
123 RPMForge (for fuse-python) and EPEL (for pyxattr and pylibacl).
125 On Cygwin, install python, make, rsync, and gcc4.
127 If you would like to use the optional bup web server on systems
128 without a tornado package, you may want to try this:
132 - Build the python module and symlinks:
140 (The tests should pass. If they don't pass for you, stop here and send
141 an email to bup-list@googlegroups.com.)
143 - You can install bup via "make install", and override the default
144 destination with DESTDIR and PREFIX.
146 Files are normally installed to "$DESTDIR/$PREFIX" where DESTDIR is
147 empty by default, and PREFIX is set to /usr. So if you wanted to
148 install bup to /opt/bup, you might do something like this:
150 make install DESTDIR=/opt/bup PREFIX=''
156 Binary packages of bup are known to be built for the following OSes:
159 http://packages.debian.org/search?searchon=names&keywords=bup
161 http://packages.ubuntu.com/search?searchon=names&keywords=bup
162 - pkgsrc (NetBSD, Dragonfly, and others)
163 http://pkgsrc.se/sysutils/bup
164 http://cvsweb.netbsd.org/bsdweb.cgi/pkgsrc/sysutils/bup/
170 - Get help for any bup command:
179 - Initialize the default BUP_DIR (~/.bup):
183 - Make a local backup (-v or -vv will increase the verbosity):
186 bup save -n local-etc /etc
188 - Restore a local backup to ./dest:
190 bup restore -C ./dest local-etc/latest/etc
193 - Look at how much disk space your backup took:
197 - Make another backup (which should be mostly identical to the last one;
198 notice that you don't have to *specify* that this backup is incremental,
199 it just saves space automatically):
202 bup save -n local-etc /etc
204 - Look how little extra space your second backup used (on top of the first):
208 - Get a list of your previous backups:
212 - Restore your first backup again:
214 bup restore -C ./dest-2 local-etc/2013-11-23-11195/etc
216 - Make a backup on a remote server (which must already have the 'bup' command
217 somewhere in the server's PATH (see /etc/profile, etc/environment,
218 ~/.profile, or ~/.bashrc), and be accessible via ssh.
219 Make sure to replace SERVERNAME with the actual hostname of your server):
221 ssh SERVERNAME bup init
223 bup save -r SERVERNAME: -n local-etc /etc
225 - Restore a backup from a remote server. (FAIL: unfortunately,
226 unlike "bup join", "bup restore" does not yet support remote
227 restores. See both "bup join" and "Things that are stupid" below.)
229 - Defend your backups from death rays (OK fine, more likely from the
230 occasional bad disk block). This writes parity information
231 (currently via par2) for all of the existing data so that bup may
232 be able to recover from some amount of repository corruption:
236 - Use split/join instead of index/save/restore. Try making a local
239 tar -cvf - /etc | bup split -n local-etc -vv
241 - Try restoring the tarball:
243 bup join local-etc | tar -tf -
245 - Look at how much disk space your backup took:
249 - Make another tar backup:
251 tar -cvf - /etc | bup split -n local-etc -vv
253 - Look at how little extra space your second backup used on top of
258 - Restore the first tar backup again (the ~1 is git notation for "one
259 older than the most recent"):
261 bup join local-etc~1 | tar -tf -
263 - Get a list of your previous split-based backups:
265 GIT_DIR=~/.bup git log local-etc
267 - Make a backup on a remote server:
269 tar -cvf - /etc | bup split -r SERVERNAME: -n local-etc -vv
271 - Try restoring the remote backup tarball:
273 bup join -r SERVERNAME: local-etc | tar -tf -
275 That's all there is to it!
281 - FreeBSD's default 'make' command doesn't like bup's Makefile. In order to
282 compile the code, run tests and install bup, you need to install GNU Make
283 from the port named 'gmake' and use its executable instead in the commands
284 seen above. (i.e. 'gmake test' runs bup's test suite)
286 - Python's development headers are automatically installed with the 'python'
287 port so there's no need to install them separately.
289 - To use the 'bup fuse' command, you need to install the fuse kernel module
290 from the 'fusefs-kmod' port in the 'sysutils' section and the libraries from
291 the port named 'py-fusefs' in the 'devel' section.
293 - The 'par2' command can be found in the port named 'par2cmdline'.
295 - In order to compile the documentation, you need pandoc which can be found in
296 the port named 'hs-pandoc' in the 'textproc' section.
299 Notes on NetBSD/pkgsrc
300 ----------------------
302 - See pkgsrc/sysutils/bup, which should be the most recent stable
303 release and includes man pages. It also has a reasonable set of
304 dependencies (git, par2, py-fuse-bindings).
306 - The "fuse-python" package referred to is hard to locate, and is a
307 separate tarball for the python language binding distributed by the
308 fuse project on sourceforge. It is available as
309 pkgsrc/filesystems/py-fuse-bindings and on NetBSD 5, "bup fuse"
312 - "bup fuse" presents every directory/file as inode 0. The directory
313 traversal code ("fts") in NetBSD's libc will interpret this as a
314 cycle and error out, so "ls -R" and "find" will not work.
316 - There is no support for ACLs. If/when some entrprising person
317 fixes this, adjust t/compare-trees.
323 - There is no support for ACLs. If/when some enterprising person
324 fixes this, adjust t/compare-trees.
326 - In t/test.sh, two tests have been disabled. These tests check to
327 see that repeated saves produce identical trees and that an
328 intervening index doesn't change the SHA1. Apparently Cygwin has
329 some unusual behaviors with respect to access times (that probably
330 warrant further investigation). Possibly related:
331 http://cygwin.com/ml/cygwin/2007-06/msg00436.html
337 - There is no support for ACLs. If/when some enterprising person
338 fixes this, adjust t/compare-trees.
347 bup stores its data in a git-formatted repository. Unfortunately, git
348 itself doesn't actually behave very well for bup's use case (huge numbers of
349 files, files with huge sizes, retaining file permissions/ownership are
350 important), so we mostly don't use git's *code* except for a few helper
351 programs. For example, bup has its own git packfile writer written in
354 Basically, 'bup split' reads the data on stdin (or from files specified on
355 the command line), breaks it into chunks using a rolling checksum (similar to
356 rsync), and saves those chunks into a new git packfile. There is one git
359 When deciding whether to write a particular chunk into the new packfile, bup
360 first checks all the other packfiles that exist to see if they already have that
361 chunk. If they do, the chunk is skipped.
363 git packs come in two parts: the pack itself (*.pack) and the index (*.idx).
364 The index is pretty small, and contains a list of all the objects in the
365 pack. Thus, when generating a remote backup, we don't have to have a copy
366 of the packfiles from the remote server: the local end just downloads a copy
367 of the server's *index* files, and compares objects against those when
368 generating the new pack, which it sends directly to the server.
370 The "-n" option to 'bup split' and 'bup save' is the name of the backup you
371 want to create, but it's actually implemented as a git branch. So you can
372 do cute things like checkout a particular branch using git, and receive a
373 bunch of chunk files corresponding to the file you split.
375 If you use '-b' or '-t' or '-c' instead of '-n', bup split will output a
376 list of blobs, a tree containing that list of blobs, or a commit containing
377 that tree, respectively, to stdout. You can use this to construct your own
378 scripts that do something with those values.
383 'bup index' walks through your filesystem and updates a file (whose name is,
384 by default, ~/.bup/bupindex) to contain the name, attributes, and an
385 optional git SHA1 (blob id) of each file and directory.
387 'bup save' basically just runs the equivalent of 'bup split' a whole bunch
388 of times, once per file in the index, and assembles a git tree
389 that contains all the resulting objects. Among other things, that makes
390 'git diff' much more useful (compared to splitting a tarball, which is
391 essentially a big binary blob). However, since bup splits large files into
392 smaller chunks, the resulting tree structure doesn't *exactly* correspond to
393 what git itself would have stored. Also, the tree format used by 'bup save'
394 will probably change in the future to support storing file ownership, more
395 complex file permissions, and so on.
397 If a file has previously been written by 'bup save', then its git blob/tree
398 id is stored in the index. This lets 'bup save' avoid reading that file to
399 produce future incremental backups, which means it can go *very* fast unless
400 a lot of files have changed.
403 Things that are stupid for now but which we'll fix later
404 ========================================================
406 Help with any of these problems, or others, is very welcome. Join the
407 mailing list (see below) if you'd like to help.
409 - 'bup restore' can't pull directly from a remote server.
411 So in one sense "save -r" is a dead-end right now. Obviously you
412 can use "ssh SERVER bup restore -C ./dest..." to create a tree you
413 can transfer elsewhere via rsync/tar/whatever, but that's *lame*.
415 Until we fix it, you may be able to mount the remote BUP_DIR via
416 sshfs and then restore "normally", though that hasn't been
419 - 'bup save' and 'bup restore' have immature metadata support.
421 On the plus side, they actually do have support now, but it's new,
422 and not remotely as well tested as tar/rsync/whatever's. However,
423 you have to start somewhere, and as of 0.25, we think it's ready
424 for more general use. Please let us know if you have any trouble.
426 Also, if any strip or graft-style options are specified to 'bup
427 save', then no metadata will be written for the root directory.
428 That's obviously less than ideal.
430 - bup is overly optimistic about mmap. Right now bup just assumes
431 that it can mmap as large a block as it likes, and that mmap will
432 never fail. Yeah, right... If nothing else, this has failed on
433 32-bit architectures (and 31-bit is even worse -- looking at you,
436 To fix this, we might just implement a FakeMmap[1] class that uses
437 normal file IO and handles all of the mmap methods[2] that bup
438 actually calls. Then we'd swap in one of those whenever mmap
441 This would also require implementing some of the methods needed to
442 support "[]" array access, probably at a minimum __getitem__,
443 __setitem__, and __setslice__ [3].
445 [1] http://comments.gmane.org/gmane.comp.sysutils.backup.bup/613
446 [2] http://docs.python.org/2/library/mmap.html
447 [3] http://docs.python.org/2/reference/datamodel.html#emulating-container-types
449 - 'bup index' is slower than it should be.
451 It's still rather fast: it can iterate through all the filenames on my
452 600,000 file filesystem in a few seconds. But it still needs to rewrite
453 the entire index file just to add a single filename, which is pretty
454 nasty; it should just leave the new files in a second "extra index" file
457 - bup could use inotify for *really* efficient incremental backups.
459 You could even have your system doing "continuous" backups: whenever a
460 file changes, we immediately send an image of it to the server. We could
461 give the continuous-backup process a really low CPU and I/O priority so
462 you wouldn't even know it was running.
464 - bup currently has no way to prune *old* backups.
466 Because of the way the packfile system works, backups become "entangled"
467 in weird ways and it's not actually possible to delete one pack
468 (corresponding approximately to one backup) without risking screwing up
471 git itself has lots of ways of optimizing this sort of thing, but its
472 methods aren't really applicable here; bup packfiles are just too huge.
473 We'll have to do it in a totally different way. There are lots of
474 options. For now: make sure you've got lots of disk space :)
476 Until we fix this, one possible workaround is to just start a new
477 BUP_DIR occasionally, i.e. bup-2013-10, bup-2013-11...
479 - bup has never been tested on anything but Linux, MacOS, and Windows+Cygwin.
481 There's nothing that makes it *inherently* non-portable, though, so
482 that's mostly a matter of someone putting in some effort. (For a
483 "native" Windows port, the most annoying thing is the absence of ssh in
484 a default Windows installation.)
486 - bup needs better documentation.
488 According to a recent article about bup in Linux Weekly News
489 (https://lwn.net/Articles/380983/), "it's a bit short on examples and
490 a user guide would be nice." Documentation is the sort of thing that
491 will never be great unless someone from outside contributes it (since
492 the developers can never remember which parts are hard to understand).
494 - bup is "relatively speedy" and has "pretty good" compression.
496 ...according to the same LWN article. Clearly neither of those is good
497 enough. We should have awe-inspiring speed and crazy-good compression.
498 Must work on that. Writing more parts in C might help with the speed.
502 Actually, that's not stupid, but you might consider it a limitation.
503 There are a bunch of Linux GUI backup programs; someday I expect someone
504 will adapt one of them to use bup.
510 bup has an extensive set of man pages. Try using 'bup help' to get
511 started, or use 'bup help SUBCOMMAND' for any bup subcommand (like split,
512 join, index, save, etc.) to get details on that command.
514 For further technical details, please see ./DESIGN.
520 bup is a work in progress and there are many ways it can still be improved.
521 If you'd like to contribute patches, ideas, or bug reports, please join the
524 You can find the mailing list archives here:
526 http://groups.google.com/group/bup-list
528 and you can subscribe by sending a message to:
530 bup-list+subscribe@googlegroups.com
532 Please see <a href="HACKING">./HACKING</a> for
533 additional information, i.e. how to submit patches (hint - no pull
534 requests), how we handle branches, etc.