Creating signed debian apt repository

Update: 2014.03.19 Use freight

Recently had the pleasure(?) of having to resetup an apt repository. It had been managed by reprepro, but there apparently the underlying concept of a single version per repository that is central to the implementation. Seems kinda weird to me, as was it to some of the users of the repository. In an effort to resolve this and simplify our repository management I did some research and came up with a very simple solution.

Using the following references:

We already had a GPG key, so I didn’t have to create a key, i just followed the instructions in How to use GnuPG in an automated environment to generate a passwordless signing subkey using the existing gpg key. If you don’t have a key already there are lots of resources available if you just search around a bit. Here is one.

Once the private key was created I removed the original gpg keypair for that user off the system, storing it some place safe, and replaced it with the subkey’s files I just created. This allows us to revoke the subkey if it or the server is compromised. Then we can generate a new subkey using the original key and the consumers of the repository won’t need to change their installed key.

Next, we need to lay down a two things. A directory and a config file. In my example the directory will be the root of your domain on a web server. I’m not going to go into how to setup a web server here.

The config file is for apt-ftparchive to generate the Releases file. Following the instructions from Automatic Debian Package Repository HOWTO we generate /srv/repo.com/debian/Releases.conf

In my repository we only have noarch packages. No source packages, no architecture specific. Adjust that field as is necessary for your repository. We also only have one component section. I don’t even think it ends up being relevant to this simple of a repository, the same as Suite. What matters is that the codename matches your directory. I used debian for both, because ours is not specific to a Debian release such as “squeeze” or “wheezy”.

With these files in place I created the following script to build the repository, and saved it as /usr/local/bin/buildrepo.sh:

This script assumes that the packages are being uploaded to the RELEASES directory of the user bob.

I create both Packages and Packages.gz because i didn’t like seeing the Ign message when I did an apt-get update. You could consolidate it down to the one line and have just a single file.

Now for the fun part. With Incron you can configure the execution of a command/script based on filesystem events. Since packages are being pushed into /home/bob/RELEASES we want to monitor that directory. We have a large number of file event types we could trigger from, as can be seen under inotify events of this man page. For our purposes IN_CREATE was what I used. I need to verify as IN_CLOSE_WRITE might be better for larger files. To configure this I run incrontab -e and added the following entry:

That’s it! Now if you push a deb package into /home/bob/RELEASES the repository will get generated. The resulting repository can be accessed by configuring the following source on a client system.

Extra credit: add your public key to the /srv/repo.example.com/ directory so that users can add it to their local apt-key ring, which will otherwise complain.

Published by

xaeth

So I'm in my 30s. I'm a career computer geek, but of the skilled and suitably employed variety, not the variety that runs around in one of a fleet of identical vehicles to wage viral warfare. I have spent well over half my life online, and was done with most forms of social networking by the time I hit 23. For those of you that doubt it IRC, forums, and even the good old BBS's of yester-year (which I missed out on since my parents would not let me connect the modem on my commodore 64) are all social networking. We just didn't have such a fancy accepted term for it then. Through out that time I have considered starting a blog on occasion. Not because I'm all that interesting (the level varies year to year), but because I so often end up putting together pieces of technology in a way that I have a hard time finding good online resources for, and its only fair to try and give back. But alas, I tend to be a bit lazy, or busy, and never got around to it. Until now (I hope, and so far have failed). The point of this blog is to be a bit more of a collection of thoughts, helpful hints, or maybe commentary on kewl things. I'll try to leave the details of my harrowing treks down ten inch deep rapids or the details of my last family gathering out of it. For your safety and well-being as much as my own. This blog is my personal blog. The views expressed on these pages are mine alone and not those of my past, present or any future employer.

Leave a Reply