debmirror - Online in the Cloud

PROGRAM:

NAME

debmirror - Debian partial mirror script, with ftp, http or rsync and package pool support

SYNOPSIS

debmirror [options] mirrordir

DESCRIPTION

This program downloads and maintains a partial local Ubuntu mirror. It can mirror any
combination of architectures, distributions, and sections. Files are transferred by ftp,
and package pools are fully supported. It also does locking and updates trace files.

The partial mirror created by this program is not suitable to be used as a public Debian
mirror. If that is your aim, you should instead follow the instructions at
<http://www.debian.org/mirrors/ftpmirror>.

This program mirrors in three steps.

1. download Packages and Sources files
First it downloads all Packages and Sources files for the subset of Ubuntu it was
instructed to get.

2. download everything else
The Packages and Sources files are scanned, to build up a list of all the files they
refer to. A few other miscellaneous files are added to the list. Then the program
makes sure that each file in the list is present on the local mirror and is up-to-
date, using file size (and optionally checksum) checks. Any necessary files are
downloaded.

3. clean up unknown files
Any files and directories on the local mirror that are not in the list are removed.

OPTIONS

mirrordir
This required (unless defined in a configuration file) parameter specifies where the
local mirror directory is. If the directory does not exist, it will be created. Be
careful; telling this program that your home directory is the mirrordir is guaranteed
to replace your home directory with an Ubuntu mirror!

-p, --progress
Displays progress bars as files are downloaded.

-v, --verbose
Displays progress between file downloads.

--debug
Enables verbose debug output, including ftp protocol dump.

--dry-run
Simulate a mirror run. This will still download the meta files to the ./.temp working
directory, but won't replace the old meta files, won't download debs and source files
and only simulates cleanup.

--skip-installer=foo[,bar,..]
Don't download debian-installer files for the specified distribution.

--help
Display a usage summary.

-h, --host=remotehost
Specify the remote host to mirror from. Defaults to archive.ubuntu.com, you are
strongly encouraged to find a closer mirror.

-r, --root=directory
Specifies the directory on the remote host that is the root of the Ubuntu archive.
Defaults to ubuntu, which will work for most mirrors. The root directory has a dists
subdirectory.

--method=method
Specify the method to download files. Currently, supported methods are ftp, http,
https, and rsync. The file method is experimentally supported.

--passive
Download in passive mode when using ftp.

-u, --user=remoteusername
Specify the remote user name to use to log into the remote host. Defaults to
"anonymous".

--passwd=remoteuserpassword
Specify the remote user password to use to log into the remote ftp host. It is used
with --user and defaults to "anonymous@".

--proxy=http://user:pass@url:port/
Specifies the http proxy (like Squid) to use for http or ftp methods.

-d, --dist=foo[,bar,..]
Specify the distribution (lucid, oneiric, precise) of Ubuntu to mirror. This switch
may be used multiple times, and multiple distributions may be specified at once,
separated by commas.

You may also use the stable, testing, unstable, names.

--omit-suite-symlinks
With this option set, debmirror will not create the symlink from suite to codename.
This is needed for example when mirroring archived Debian releases as they will all
have either "stable" or "oldstable" as suite in their Release files.

-s, --section=foo[,bar,..]
Specify the section of Ubuntu to mirror. Defaults to
"main,contrib,non-free,main/debian-installer".

-a, --arch=foo[,bar,..]
Specify the architectures to mirror. The default is --arch=i386. Specifying
--arch=none will mirror no archs.

--rsync-extra=foo[,bar,..]
Allows you to also mirror files from a number of directories that are not part of the
package archive itself.

Debmirror will always use rsync for the transfer of these files, irrespective of what
transfer method is specified in the --method option. This will therefore not work if
your remote mirror does not support rsync, or if the mirror needs a different --root
option for rsync than for the main transfer method specified with --method.

Note that excluding individual files in the directories is not supported.

The following values are supported.

doc
Download all files and subdirectories in doc directory, and all README files in the
root directory of the archive.

indices
Download all files and subdirectories in indices directory. Note that this directory
can contain some rather large files; don't include this type unless you know you
need these files.

tools
Download all files and subdirectories in tools directory.

trace
Download the remote mirror's trace files for the archive (project/trace/*). This is
enabled by default.

none
This can be used to disable getting extra files with rsync.

If specified, the update of trace files will be done at the beginning of the mirror
run; the other types are done near the end.

This switch may be used multiple times, and multiple values may be specified at once,
separated by commas; unknown values are ignored.

--di-dist=dists | foo[,bar,..]
Mirror current Debian Installer images for the specified dists. See further the
section "Mirroring Debian Installer images" below.

--di-arch=arches | foo[,bar,..]
Mirror current Debian Installer images for the specified architectures. See further
the section "Mirroring Debian Installer images" below.

--source
Include source in the mirror (default).

--nosource
Do not include source.

--i18n
Additionally download Translation-<lang>.bz2 files, which contain translations of
package descriptions. Selection of specific translations is possible using the
--include and --exclude options. The default is to download only the English file.

--getcontents
Additionally download Contents.<arch>.gz files. Note that these files can be
relatively big and can change frequently, especially for the testing and unstable
suites. Use of the available diff files is strongly recommended (see the --diff
option).

--checksums
Use checksums to determine if files on the local mirror that are the correct size
actually have the correct content. Not enabled by default, because it is too paranoid,
and too slow.

When the state cache is used, debmirror will only check checksums during runs where
the cache has expired or been invalidated, so it is worth considering to use these two
options together.

--ignore-missing-release
Don't fail if the Release file is missing.

--check-gpg, --no-check-gpg
Controls whether gpg signatures from the Release.gpg file should be checked. The
default is to check signatures.

--keyring=file
Use file as an additional gpg-format keyring. May be given multiple times.

Note that these will be used in addition to $GNUPGHOME/trustedkeys.gpg. The latter
can be removed from the set of keyrings by setting $GNUPGHOME to something non-
existent when using this option.

On a typical Debian system, the Debian archive keyring can be used directly with this
option:

debmirror --keyring /usr/share/keyrings/debian-archive-keyring.gpg ...

--ignore-release-gpg
Don't fail if the Release.gpg file is missing. If the file does exist, it is mirrored
and verified, but any errors are ignored.

--ignore=regex
Never delete any files whose filenames match the regex. May be used multiple times.

--exclude=regex
Never download any files whose filenames match the regex. May be used multiple times.

--include=regex
Don't exclude any files whose filenames match the regex. May be used multiple times.

--exclude-deb-section=regex
Never download any files whose Debian Section (games, doc, oldlibs, science, ...)
match the regex. May be used multiple times.

--limit-priority=regex
Limit download to files whose Debian Priority (required, extra, optional, ...) match
the regex. May be used multiple times.

--exclude-field=fieldname=regex
Never download any binary packages where the contents of fieldname match the regex.
May be used multiple times. If this option is used and the mirror includes source
packages, only those source packages corresponding to included binary packages will be
downloaded.

--include-field=fieldname=regex
Don't exclude any binary packages where the contents of fieldname match the regex. May
be used multiple times. If this option is used and the mirror includes source
packages, only those source packages corresponding to included binary packages will be
downloaded.

-t, --timeout=seconds
Specifies the timeout to use for network operations (either FTP or rsync). Set this
to a higher value if you experience failed downloads. Defaults to 300 seconds.

--max-batch=number
Download at most max-batch number of files (and ignore rest).

--rsync-batch=number
Download at most number of files with each rsync call and then loop.

--rsync-options=options
Specify alternative rsync options to be used. Default options are "-aL --partial".
Care must be taken when specifying alternative options not to disrupt operations, it's
best to only add to those options.

The most likely option to add is "--bwlimit=x" to avoid saturating the bandwidth of
your link.

--postcleanup
Clean up the local mirror but only after mirroring is complete and only if there was
no error.

This is the default, because it ensures that the mirror is consistent at all times.

--precleanup
Clean up the local mirror before starting mirroring.

This option may be useful if you have limited disk space, but it will result in an
inconsistent mirror when debmirror is running.

The deprecated --cleanup option also enables this mode.

--nocleanup
Do not clean up the local mirror.

--skippackages
Don't re-download Packages and Sources files. Useful if you know they are up-to-date.

--diff=use|mirror|none
If --diff=use is specified and the Release file contains entries for diff files, then
debmirror will attempt to use them to update Packages, Sources, and Contents files
(which can significantly reduce the download size for meta files), but will not
include them in the mirror. This is the default behavior and avoids having time
consuming diff files for a fast local mirror.

Specifying --diff=mirror does the same as use, but will also include the downloaded
diff files in the local mirror. Specify --diff=none to completely ignore diff files.

Note that if rsync is used as method to download files and the archive being mirrored
has "rsyncable" gzipped meta files, then using --diff=none may be the most efficient
way to download them. See the gzip(1) man page for information about its rsyncable
option.

--gzip-options=options
Specify alternative options to be used when calling gzip(1) to compress meta files
after applying diffs. The default options are "-9 -n --rsyncable" which corresponds
with the options used to gzip meta files for the main Debian archive.

These options may need to be modified if the checksum of the file as gzipped by
debmirror does not match the checksum listed in the Release file (which will result in
the gzipped file being downloaded unnecessarily after diffs were successfully
applied).

--slow-cpu
By default debmirror saves some bandwidth by performing cpu-intensive tasks, such as
compressing files to generate .gz and .xz files. Use this mode if the computer's CPU
is slow, and it makes more sense to use more bandwidth and less CPU.

This option implies --diff=none.

--state-cache-days=number
Save the state of the mirror in a cache file between runs. The cache will expire after
the specified number of days, at which time a full check and cleanup of the mirror
will be done. While the cache is valid, debmirror will trust that the mirror is
consistent with this cache.

The cache is only used for files that have a unique name, i.e. binary packages and
source files. If a mirror update fails for any reason, the cache will be invalidated
and the next run will include a full check.

Main advantage of using the state cache is that it avoids a large amount of disk
access while checking which files need to be fetched. It may also reduce the time
required for mirror updates.

--ignore-small-errors
Normally debmirror will report an error if any deb files or sources fail to download
and refuse to update the meta data to an inconsistent mirror. Normally this is a good
things as it indicates something went wrong during download and should be retried. But
sometimes the upstream mirror actually is broken. Specifying --ignore-small-errors
causes debmirror to ignore missing or broken deb and source files but still be
pedantic about checking meta files.

--allow-dist-rename
The directory name for a dist should be equal to its Codename and not to a Suite. If
the local mirror currently has directories named after Suites, debmirror can rename
them automatically. An existing symlink from codename to suite will be removed, but
debmirror will automatically create a new symlink suite -> codename (immediately after
moving meta files in place). This conversion should only be needed once.

--disable-ssl-verification
When https is used, debmirror checks that the SSL certificate is value.

If the server has a self-signed certificate, the check can be disabled with this
option.

--debmarshal
On each pull, keep the repository meta data from dists/* in a numbered subdirectory,
and maintain a symlink latest to the most recent pull. This is similar to Debmarshal
in tracking mode, see debmarshal.debian.net for examples and use. debmirror cleanup
is disabled when this flag is specified. Separate pool and snapshot cleanup utilities
are available at
http://code.google.com/p/debmarshal/source/browse/#svn/trunk/repository2

--config-file=file
Specify a configuration file. This option may be repeated to read multiple
configuration files. By default debmirror reads /etc/debmirror.conf and
~/.debmirror.conf (see section FILES).

Experimental options
--retry-rsync-packages=number
While downloading Packages and related files via rsync, try up to this many times if
rsync fails to connect. Defaults to 1, to try only once. (A typical nondefault value
is 10. To try an unlimited number of times, use -1 or 0.)

USING DEBMIRROR

Using regular expressions in options
Various options accept regular expressions that can be used to tune what is included in
the mirror. They can be any regular expression valid in perl, which also means that
extended syntax is standard. Make sure to anchor regular expressions appropriately: this
is not done by debmirror.

The --include and --exclude options can be combined. This combination for example will, if
the --i18n option is used, exclude all Translation files, except for the ones for
Portuguese (pt) and Brazillian (pt_BR):

--exclude='/Translation-.*\.bz2$' --include='/Translation-pt.*\.bz2$'

Mirroring Debian Installer images
Debmirror will only mirror the "current" images that are on the remote mirror. At least
one of the options --di-dist or --di-arch must be passed to enable mirroring of the
images.

The special values "dists" and "arches" can be used to tell debmirror to use the same
dists and architectures for D-I images as for the archive, but it is also possible to
specify different values. If either option is not set, it will default to the same values
as for the archive.

If you wish to create custom CD images using for example debian-cd, you will probably also
want add the option "--rsync-extra=doc,tools".

Limitations

There are no progress updates displayed for D-I images.

Archive size
The tables in the file /usr/share/doc/debmirror/mirror_size give an indication of the
space needed to mirror the Debian archive. They are particularly useful if you wish to set
up a partial mirror. Only the size of source and binary packages is included. You should
allow for around 1-4 GB of meta data (in ./dists/<dist>) per suite (depending in your
settings). Plus whatever space is needed for extra directories (e.g. tools, doc) you wish
to mirror.

The tables also show how much additional space is required if you add a release on top of
its predecessor. Note that the additional space needed for testing and (to a lesser
extent) unstable varies during the development cycle of a release. The additional space
needed for testing is zero immediately after a stable release and grows from that time
onwards.

Note Debmirror keeps an extra copy of all meta data. This is necessary to guarantee that
the local mirror stays consistent while debmirror is running.

EXAMPLES

Simply make a mirror in /srv/mirror/debian, using all defaults (or the settings defined in
debmirror.conf):

debmirror /srv/mirror/debian

Make a mirror of i386 and amd64 binaries, main and universe only, and include both LTS and
latest versions of Ubuntu; download from 'archive.ubuntu.com':

debmirror -a i386,amd64 -d lucid -d precise -s main,universe --nosource \
-h archive.ubuntu.com --progress $HOME/mirror/debian

Make a mirror using rsync (rsync server is 'ftp.debian.org::debian'), excluding the
section 'debug' and the package 'foo-doc':

debmirror -e rsync $HOME/mirror/debian --exclude='/foo-doc_' \
--exclude-deb-section='^debug$'

Use debmirror online using onworks.net services