Packaging Drupal for Habitat. How hard can it be?
Early in my Habitat journey I packed a super-simple Python application that I had developed. It served no purpose other than to be packaged. Once I started to get my head around the fundamentals of packaging with Habitat, my mind turned to more useful things.
These are interesting in their own right and you'll be hearing more about them after ChefConf. For now, I want to talk about Drupal.
Drupal is "just" some PHP. How hard can it be? 1 Well, I saw enough raised eyebrows from my friends at Chef, to understand this might not be as easy as it seems.
Let's forget Drupal for now. The first step is working out how to package any piece of PHP for Habitat. Well, we really only need three ingredients:
No rocket science here. We just need to glue it all together. Most of
the hard work is already done for us, since Habitat already has
core/php packages available.
All I really wanted to do was unpack my PHP in a location where NGINX could access it and then point NGINX to my PHP-FPM socket/ip. This is all easy enough when you know how.
I didn't know how. Thankfully, the Habitat public depot has a plan available for Wordpress 2 which gave me some insight. The "trick" here is to launch PHP-FPM and then use an init hook to also launch NGINX with a config you give it.
OK, so now Drupal.
The first attempt a Drupal was easy enough:
This plan was built exactly as I described above: I download the source tarball, unpack it in a specific location and configure NGINX to use that location as a site root. When it is run, the user is presented with a shiny new Drupal install page:
So, there you have it, a successful Drupal package for Habitat!
Or is it?
The problem with my first attempt is that it is not very Habitat-y. Or
Drupal-y for that matter. The cool kids use the Drupal shell
(drush) to deploy/configure Drupal and here
comes my first problem: there's no Habitat package for
Not to worry:
drush could not be any simpler to install; it is used
as a dependency in composer. Thankfully, there is a
package I can use for this purpose. I rolled out a
package with minimal effort. It can be found here.
Drush is important for one particular reason: having installed the base Drupal ("core"?) system, drush can be used to build a preconfigured site, including default user credentials and database configuration. This means we do not need to force the user of the Drupal package to go through the post-install install script. They'd be presented with a ready-made site, all configured and ready to go.
Providing most of the details needed to configure a Drupal site using
drush is simple, we can simply dump the info in
default.toml to be
pulled out anytime. Mildly trickier is the database config.
Just as with the default account credentials, I could simply dump all
the default database config into
default.toml. This would not be a
very Habitat-y thing to do.
Much better is to use service binding, to bind my Drupal service to an existing database service. In Habitat a service bind is the mechanism that allows me to say, "this service relies on this other service in order to run properly". At runtime, I simply tell my service which other service to bind to:
hab start baggerspion/drupal --bind database:mysql.default
I will spare the full details on how binds work. There is plenty of documentation available on the Habitat website. What's important here is that, because I have bound my Drupal service to a Mysql service, I can now configure my service to use that database automatically. What does the user see now?
This is much better. I think.