1 % bup-index(1) Bup %BUP_VERSION%
2 % Avery Pennarun <apenwarr@gmail.com>
7 bup-index - print and/or update the bup filesystem index
11 bup index \<-p|-m|-s|-u|\--clear|\--check\> [-H] [-l] [-x] [\--fake-valid]
12 [\--no-check-device] [\--fake-invalid] [-f *indexfile*] [\--exclude *path*]
13 [\--exclude-from *filename*] [\--exclude-rx *pattern*]
14 [\--exclude-rx-from *filename*] [-v] \<paths...\>
18 `bup index` prints and/or updates the bup filesystem index,
19 which is a cache of the absolute paths, attributes, and sha-1
20 hashes of each file and directory in the filesystem. The
21 bup index is similar in function to the `git`(1) index, and
22 can be found in `$BUP_DIR/bupindex`.
24 Creating a backup in bup consists of two steps: updating
25 the index with `bup index`, then actually backing up the
26 files (or a subset of the files) with `bup save`. The
27 separation exists for these reasons:
29 1. There is more than one way to generate a list of files
30 that need to be backed up. For example, you might want to
31 use `inotify`(7) or `dnotify`(7).
33 2. Even if you back up files to multiple destinations (for
34 added redundancy), the file names, attributes, and hashes
35 will be the same each time. Thus, you can save the trouble
36 of repeatedly re-generating the list of files for each
39 3. You may want to use the data tracked by bup index for
40 other purposes (such as speeding up other programs that
41 need the same information).
45 At the moment, bup will ignore Linux attributes (cf. chattr(1) and
46 lsattr(1)) on some systems (any big-endian systems where sizeof(long)
47 < sizeof(int)). This is because the Linux kernel and FUSE currently
48 disagree over the type of the attr system call arguments, and so on
49 big-endian systems there's no way to get the results without the risk
50 of stack corruption (http://lwn.net/Articles/575846/). In these
51 situations, bup will print a warning the first time Linux attrs are
52 relevant during any index/save/restore operation.
54 bup makes accommodations for the expected "worst-case" filesystem
55 timestamp resolution -- currently one second; examples include VFAT,
56 ext2, ext3, small ext4, etc. Since bup cannot know the filesystem
57 timestamp resolution, and could be traversing multiple filesystems
58 during any given run, it always assumes that the resolution may be no
59 better than one second.
61 As a practical matter, this means that index updates are a bit
62 imprecise, and so `bup save` may occasionally record filesystem
63 changes that you didn't expect. That's because, during an index
64 update, if bup encounters a path whose actual timestamps are more
65 recent than one second before the update started, bup will set the
66 index timestamps for that path (mtime and ctime) to exactly one second
67 before the run, -- effectively capping those values.
69 This ensures that no subsequent changes to those paths can result in
70 timestamps that are identical to those in the index. If that were
71 possible, bup could overlook the modifications.
73 You can see the effect of this behavior in this example (assume that
74 less than one second elapses between the initial file creation and
78 # A "sleep 1" here would avoid the unexpected save.
80 $ bup save -n src src # Saves 1 and 2.
83 $ date > src/2 # Not indexed.
84 $ bup save -n src src # But src/2 is saved anyway.
86 Strictly speaking, bup should not notice the change to src/2, but it
87 does, due to the accommodations described above.
92 : recursively update the index for the given paths and their
93 descendants. One or more paths must be specified, and if a path
94 ends with a symbolic link, the link itself will be indexed, not
95 the target. If no mode option is given, `--update` is the
96 default, and paths may be excluded by the `--exclude`,
97 `--exclude-rx`, and `--one-file-system` options.
100 : print the contents of the index. If paths are
101 given, shows the given entries and their descendants.
102 If no paths are given, shows the entries starting
103 at the current working directory (.).
106 : prints only files which are marked as modified (ie.
107 changed since the most recent backup) in the index.
111 : prepend a status code (A, M, D, or space) before each
112 filename. Implies `-p`. The codes mean, respectively,
113 that a file is marked in the index as added, modified,
114 deleted, or unchanged since the last backup.
117 : carefully check index file integrity before and after
118 updating. Mostly useful for automated tests.
121 : clear the default index.
127 : for each file printed, prepend the most recently
128 recorded hash code. The hash code is normally
129 generated by `bup save`. For objects which have not yet
130 been backed up, the hash code will be
131 0000000000000000000000000000000000000000. Note that
132 the hash code is printed even if the file is known to
133 be modified or deleted in the index (ie. the file on
134 the filesystem no longer matches the recorded hash).
135 If this is a problem for you, use `--status`.
138 : print more information about each file, in a similar
139 format to the `-l` option to `ls`(1).
141 -x, \--xdev, \--one-file-system
142 : don't cross filesystem boundaries when traversing the
143 filesystem -- though as with tar and rsync, the mount points
144 themselves will still be indexed. Only applicable if you're using
148 : mark specified paths as up-to-date even if they
149 aren't. This can be useful for testing, or to avoid
150 unnecessarily backing up files that you know are
154 : mark specified paths as not up-to-date, forcing the
155 next "bup save" run to re-check their contents.
157 -f, \--indexfile=*indexfile*
158 : use a different index filename instead of
162 : exclude *path* from the backup (may be repeated).
164 \--exclude-from=*filename*
165 : read --exclude paths from *filename*, one path per-line (may be
166 repeated). Ignore completely empty lines.
168 \--exclude-rx=*pattern*
169 : exclude any path matching *pattern*, which must be a Python regular
170 expression (http://docs.python.org/library/re.html). The pattern
171 will be compared against the full path, without anchoring, so
172 "x/y" will match "ox/yard" or "box/yards". To exclude the
173 contents of /tmp, but not the directory itself, use
174 "^/tmp/.". (may be repeated)
178 * '/foo$' - exclude any file named foo
179 * '/foo/$' - exclude any directory named foo
180 * '/foo/.' - exclude the content of any directory named foo
181 * '^/tmp/.' - exclude root-level /tmp's content, but not /tmp itself
183 \--exclude-rx-from=*filename*
184 : read --exclude-rx patterns from *filename*, one pattern per-line
185 (may be repeated). Ignore completely empty lines.
188 : don't mark an entry invalid if the device number (stat(2) st_dev)
189 changes. This can be useful when indexing remote, automounted, or
190 snapshot filesystems (LVM, Btrfs, etc.), where the device number
194 : increase log output during update (can be used more
195 than once). With one `-v`, print each directory as it
196 is updated; with two `-v`, print each file too.
200 bup index -vux /etc /var /usr
205 `bup-save`(1), `bup-drecurse`(1), `bup-on`(1)
209 Part of the `bup`(1) suite.