Using Git to Deploy a Website

I use git at work and I think it’s a fantastic version control system. I’ve also been using it on some personal projects and I’ve written a post-receive hook that is very useful for managing simple web applications. Once configured, you can deploy to a staging server and/or a production server by simply running git-push. Skim down to Step 4 if you need an example to be convinced.

This script is great for a simple site with the following requirements:

  • Site can be updated by hosting contents in a symlinked directory. When the symlink is updated to point to a different directory, the website updates without a server restart. This is basically how Capistrano works… a PHP website is a good example where this works.

  • You want to immediately update a staging server with you latest git push and update a production server with specific tagged commits.

  • You can configure two different urls such as http://staging.yourdomain.com -> /path/to/staging and http://www.yourdomain.com -> /path/to/current. This is easy to do on a hosting service like DreamHost.

  • You have SSH keys enabled so you can easily connect to your server.

  • You’re lazy and you want a script to do all your deployment work.

Assumptions:

  • You’re comfortable using git.

  • You’ve read the man pages for git-push, git-pull, git-tag, git-status and git-diff. (Can’t hurt to read all the git man pages… there aren’t that many.)

  • You know how to use an .htaccess file to prevent anyone from seeing the contents of your git repo or the data directory we’ll create. In my apache configuration, I use RewriteRule ^.git.*$ - [F]" and "RewriteRule ^data.*$ - [F].

Step 1:

Create a directory that we’ll call /path/to for the rest of this post.

Configure your server to host a staging url from /path/to/staging and a production url from /path/to/current. We’ll create the contents of those directories in step 3.

Step 2:

Create the git repo on your server:

$ cd /path/to
$ mkdir repo
$ cd repo
$ git init --bare
$ cd hooks
$ wget https://www.tonyscelfo.com/dl/git/post-receive
$ chmod u+x post-receive

Step 3:

Clone the repo on your local machine:

$ git clone -b master <user>@<hostname>:/path/to/repo

On the initial empty checkout, you will see: “warning: You appear to have cloned an empty repository.” Once we push, you won’t see that again if/when you clone to other local machines.

$ cd
$ mkdir data
$ echo "p0" > data/live
$ git add data/live
$ git commit -a -m "initial commit to create valid repo"
$ git tag p0
$ git push origin +master:refs/heads/master --tags

You will now have the following on your server:

/path/to/repo (a bare git repo)
/path/to/staging (contains the latest code)
/path/to/current -> /path/to/p0
/path/to/p0 (the directory containing what we tagged as p0)

Step 4 (repeat for as long as you work on the project):

Work on your site and push to staging:

$ # work, work, work
$ git commit
$ git push
$ # repeat

Every time you push, the staging directory will be updated immediately. When you want to promote to production, do the following:

  1. Checkout the appropriate commit if its not the current HEAD

  2. git tag <tag_name> to tag your repo with a release name

  3. Edit the contents of data/live to contain <tag_name>

  4. git push –tags to push the tags and trigger the update to the current symlink

If you ever want to rollback, you can simply:

  1. Run git tag -n to see a list of all the previous release tags with annotations

  2. Edit the contents of data/live to contain <tag_name> of the previous release

  3. Run git push to push and trigger the update to the current symlink. (You don’t need to push the tags when doing a rollback because you had previously pushed the tag definition to the git repo.)

Enjoy. If you use this script, please comment with feedback!

comments powered by Disqus