Publishing with git-annex

I’ve been using git-annex and boring people with it for quite a while now. It is an excellent piece of software, and so I can’t really promise this rant will be the last.

After moving to Hakyll, I was left with a few dangling links (e.g. pictures) for various reasons detailed in the earlier post. I decided to host those files on a separate sub-domain static.cyprio.net and also used the occasion to cleanup the accumulated cruft. A few seds after, all the broken links have been squashed.

This website is normally generated with Hakyll, and then deployed by a simple rsync. However its contents are stored in Git: mainly because it makes it harder to destroy things unintentionally. Because I want the same security for the static domain, it is setup as a git-annex repository: this means git-annex manages file contents (a few Gigs), and regular git manages file’s metadata (a few Megs).

I am still “refining” the process, but right now it looks like this:

• I have clones of the full git-annex static repository on two servers (one master, and a backup).
• Those servers synchronize automatically using a cronjob (git annex sync --content): I’d rather have git hooks on the master do the sync, but I’m not sure how git-annex plugs itself in Git’s hooks. So I still have room for experimentation here. :)
• I also have a partial clone of the git-annex repository on my laptop, which allows me to push stuff online with git annex copy, or grab the data I may need when offline.

That’s all for this post. If you have other ideas, shoot me a mail, I’m all ears.