Talking Drupal #116 - Deploying Drupal

Good show, but there was some stuff missed when you got off onto the rabbit trail about just doing a git pull from the server. Jenkins just checks out the code and gets it to the server; you need something else to actually get the code where you want it. Before switching to Acquia hosting, we used Jenkins for deployment, and we used Phing (Jenkins has a plugin) for moving the code to the correct place, creating symlinks, creating the settings.php file with the appropriate values, etc. I would use a build.xml file on the server (and not in my repo, since it had db connection info) that had all the values and steps, and the plugin would look for it and use it to get the code deployment all pretty. The step were basically:

1) Copy code from Jenkins directory to build directory
2) Copy default.settings.php to settings.php
3) Replace placeholders in settings.php with values from build.xml
4) Delete webroot symlink to previous build directory
5) Create webroot symlink to new build directory
6) Create symlink from sites/default/files to files directory (located outside the build directory)
7) Run drush cc

Before using this setup, I had tried just using git pull from the prod server, and it was a nightmare, because of things like merge conflicts that had to be resolved, and other issues. This makes it so much easier, because you're basically just creating an entirely new copy of your code instead of modifying an existing code copy. In addition, if you deploy and run into problems, you can also roll back to a previous working deployment very quickly.

Back when I did the Acquia podcast, I did an episode with Steven Merrill (now with Phase2) on using Hudson (now Jenkins), and he was an absolute guru with it (I would put in the link to it, but for some reason it's not available on the Acquia site any more). He would have it running Drupal tests and then posting updates into IRC and email if something failed, among other integrations. It was a very useful tool.