My personal Jekyll workflow, involving Git, Rsync, and my own homebrewed script, the almighty Jekyllkenny…


For people who already know about the “GitHub Pages” workflow (you know, the weird two-branches workflow, one branch for the source, the other for the compiled static site), I stop you right now. This is not what I’m going to talk about.

My setup is slightly different, simpler and much more easier to picture. I do not version the compiled static site. I use Git to version ONLY the source. Then I use Rsync to update the compiled site on the server. That’s all. It’s just the usual way of doing things, nothing fancy. And it’s real easy to setup.

Source versionning with git

So, just to tell you the context: I have a VPS where I version the source of my blog with git. On the VPS, I manage my git repositories with Gitolite (I have another post talking about that).

OK, so how do you configure that ? It’s damn easy if you’re used to git. First, as soon as you create your new blog, start versionning it.

jekyll new blog
cd blog
git init
git add .gitignore *
git commit -m"Initial commit: jekyll skeleton"

Now, let’s assume that your have a vps somewhere, which is known under the name vps. I also assume that you access it through the user git (just the usual way), and that you already created an empty git repository on the server, named blog.git. The following commands come easy:

git remote add origin git@vps:blog
git push --set-upstream origin master

That’s all, your blog is now versionned and safe, you can commit and push whenever you want, and undo your worst mistakes (like the dreaded rm -rf _posts) without stress :)

Versionning with git is really the basic. Blogging is pretty much like coding, you can’t afford to loose it all once you started investing some time in it. So you must version, at least for the sake of having a backup.

Pushing to the server with rsync

This is only one of the several ways of sending your static blog to the server. I have several reasons to choose rsync for this task:

  • it’s made for syncing data, you can’t find a better tool for the task
  • it uses SSH authentication with the server, something that’s already setup
  • I never heard nobody blame rsync :)
  • it’s easy to use

It just boils down to a simple command:

rsync --archive --delete --compress \
  --chown www-data:www-data \
  ./_site/ user@vps:/var/www/blog

Notice some smart options here and there:

  • --chown to change the user and group of the synced files
  • a trailing slash at the end of the source ./_site/ to avoid creating a sub-directory inside /var/www/blog

Of course, it’s tedious to remember such command each time. To make it easier, you can alias it, or if you’re brave enough, you can start writing a little script.

Writing posts

I create a new post in the _drafts sub-directory. I work on it and preview locally, using the --draft option of Jekyll.

mkdir _drafts
vi _drafts/
jekyll server --draft

When I want to publish, I twist the usual Jekyll workflow a little bit. The common way would be just to move the draft to the post directory, and name it according to the pattern expected by Jekyll, that’s to say, prefixed by the date. So the command would be:

mv _drafts/ _posts/

This is no problem, except that I don’t like too much this naming. This is because my blog is mainly a memento for myself, where I write commands and stuff that I would forget otherwise. When I need to recall something, I don’t look for it in my web browser. I’m a command-line guy, so I prefer to go in my Jekyll post directory, and search in my posts directly.

That’s one of the reasons I like Jekyll, because it doesn’t bury my content in a database, it just let the posts live in simples markdown files, that I can browse anytime I feel like it. So when I list my posts with a simple ls command, I would like them to be sorted according to their name (which make sense to me) rather than their date.

In order to achieve that, I take advantage of the fact the Jekyll ignores the file that don’t have the proper naming. So I just name my files as I want, them add a symlink with the proper name.

mv _drafts/ _posts/
ln -s _posts/

That’s all, it’s a simple trick, but I do appreciate that Jekyll let me do that.


In the end, I started to write my own little script to automate a few things, like syncing or publishing. Basically, a kind of Octopress but much more simple, and taylored just for me.

If your Jekyll workflow meets mine, you can check it out on Gitlab and give it a try. Here is the address: