Setting up a Debian archive mirror
- Whether to mirror
- What to mirror
- Where to mirror from
- How to mirror
- Partial mirroring
- When to mirror
- Recommended additional settings
- How to add a mirror to the mirror list
While we appreciate all new mirrors, every prospective mirror maintainer should make sure that they can answer these questions before trying to start their own mirror:
- Is a mirror necessary at my location? Maybe there are already mirrors nearby.
- Do I have the resources to host a mirror? Mirrors take up considerable disk space and bandwidth, one has to be able to commit to the cost.
- Is a mirror the right choice? Sometimes people mistakenly start a mirror, when they actually want to run a caching proxy, such as apt-cacher-ng.
The main mirror page lists the archives available for mirroring.
- Users will look for the debian/ archive to install Debian over the network, to build CDs from (with jigdo), and to upgrade already installed systems.
debian-cd/ is an archive that isn't identical across all the different
mirror servers. On some sites it contains jigdo templates to build CD images
from (used in combination with files from debian/), on some it contains
already built CD images, and on some sites both.
Please see the page for mirroring the CD images for further information on this.
- debian-archive/ contains the true archive, the old and obsolete versions of Debian. It will generally be interesting only to a small segment of users.
Please see the mirror size page for more precise information about mirror sizes.
The debian-security/ archives contain the security updates released by the Debian security team. It sounds interesting to everyone, but since the security updates are sporadic, one would have to mirror it very often to be up to date (or use push mirroring) so we do not recommend it. Debian makes every effort to maintain the high availability of security.debian.org instead.
NB: the debian-non-US/ archive is deprecated since the 3.1 (sarge) release in June 2005.
Many people seem to think that
ftp.debian.org is the
canonical location of Debian packages and that it will be best for them
to mirror from that site. This is not true.
ftp.debian.org is merely one of several servers that get
updated from an internal Debian server. That address is presently located
on a single server, and it still exists mainly for
The official mirrors are meant to be a good place
to mirror from. Any of the servers in the full list of
mirrors that have the tags
Type: Push-Primary or
Type: Push-Secondary should be good to mirror from.
Please use one that is close to you both network-wise and geographically.
There is no real difference between different Push-Primary mirror servers as far as mirroring is concerned. On the other hand, if many people use ftp.debian.org (and unfortunately, they do), this needlessly wastes the donated bandwidth.
Mirror administrators in the United States should mirror from
ftp.us.debian.org (the official mirror address from the US, which is a
Push-Primary mirror). But, since that address is also a round-robin alias
for several machines, it usually makes sense to determine which of the
components is the best and then mirror from that particular one.
The upside of doing that is that it avoids the risk of failures when performing a two-stage rsync (different stages might end up on different machines during time windows where they have different data, causing a race condition). The downside is that it concentrates on a single site the performance of which may vary, but that is currently inherent to our mirror network in general.
The recommended method of mirroring is with the so called ftpsync scriptset, available in two forms:
- as a tarball from https://ftp-master.debian.org/ftpsync.tar.gz
- as a git repository: git clone https://ftp-master.debian.org/git/archvsync.git
The mirroring protocol which we strongly recommend is rsync.
Please do not mirror Debian using wget and other tools based on FTP. They may seem to work well, but will have numerous issues (they can't detect hard links, it's harder to make partial mirrors, etc).
We recommend to use the ftpsync scriptset to mirror the archive. Follow the README file included to set up mirroring.
If you do use your own scripts, the minimal functionality that must be supported is as follows:
MUST perform a 2-stage sync
The archive mirroring must be done in 2 stages. The first rsync run must ignore the index files. The correct exclude options for the first rsync run are:--exclude Packages* --exclude Sources* --exclude Release* --exclude InRelease --exclude i18n/* --exclude ls-lR*
The first stage must not delete any files.
The second stage should then transfer the above excluded files and delete files that no longer belong on the mirror.
Rationale: if archive mirroring is done in a single stage, there will be periods of time during which the index files will reference files not yet mirrored.
MUST not ignore pushes while running (for push mirroring).
If a push is received during a run of the mirror sync, it MUST NOT be ignored. The whole synchronization process must be rerun.
Rationale: most implementations of Debian mirror scripts will leave the mirror in an inconsistent state in the event of a second push being received while the first sync is still running. It is likely that in the near future, the frequency of pushes will increase.
- Run rsync with at least these options: --recursive --times --links --hard-links --delete.
- If you have extra disk space, also use the --delete-after option to avoid some of the temporary update problems.
- After rsync is done mirroring, add a time stamp file to the
project/trace/subdirectory of the Debian mirror named after your server. This means running date -u > .../debian/project/trace/your.server after your daily rsync is finished.
- We strongly advise against excluding the project/, doc/ and other subdirectories. Usually these are minor in size and yet useful to users. Especially project/trace helps very much if there are any mirror issues.
Considering the already large size of Debian archive, some people prefer to mirror only parts of it they need. If you want to exclude something, you should exclude architectures.
With the recommended ftpsync, this can be done by editing the ARCH_EXCLUDE variable.
We strongly advise against excluding the project/, doc/ and other subdirectories. Usually these are minor in size and yet useful to users. Especially project/trace helps very much if there are any mirror issues.
It is possible to use other specially written scripts, but they are usually not necessary, and not recommended for official mirrors.
The main archive gets updated four times a day. The mirrors commonly start updating around 3:00, 9:00, 15:00 and 21:00 (all times UTC), but these are never fixed times and you should not fixate your mirror on them.
Your mirror should update a few hours after one of the main archive mirror pulses. You should check if the site you're mirroring from leaves a time stamp file in their project/trace/ subdirectory. The time stamp file will be named like that site, and it will contain the completion time of their last mirror update. Add a couple of hours to that time (to be safe) and mirror then.
The easiest way to automatically have the mirror run every day is to use cron. See man crontab for details.
Note that if your site is being triggered with a push mechanism, then you don't need to worry about any of this.
Push mirroring is a form of mirroring that we have developed
to minimize the time it takes for changes to the archive to reach mirrors.
An upstream mirror uses an SSH trigger to tell the downstream mirror to
update itself. The
pushing is usually limited to a secure trigger
that takes no variable data, so the rest of the mirroring process is
pull like with a cron job.
Push mirroring is necessary to keep multiple servers in synchronization
(such as servers in DNS round-robin aliases like ftp.us.debian.org),
and we use it for our first and second tier mirrors. But as this method
requires a bit more effort to set it up, it isn't commonly done for
For a more detailed description of how this works, why it is secure, and how to set a push mirror, please see the complete explanation.
If you are going to make the Debian mirror available through HTTP, please
add the following settings to your Apache configuration (presuming, of
course, you will use Apache), within the
block, where /path/to/your/debian/mirror should be the actual
name of the directory where you keep the mirror:
Options +Indexes +SymlinksIfOwnerMatch IndexOptions NameWidth=* +SuppressDescription
This enables the directory indices and makes sure that following symlinks will work. The file names in the directory indices won't be truncated, and (mostly nonexistent) descriptions won't be shown.
In Apache 1.3.x, the following option may also be added:
Do not add that option in Apache 2.x.
Any problems or enquiries can be sent to firstname.lastname@example.org.