I’m currently teaching myself how to do Puppet. Why? Well, at Percona we support a bunch of platforms for our software. This means we have to maintain a bunch of Jenkins slaves to build the software on. We want to add new machines and have (up until now) maintained a magic “apt-get install” command line in the Jenkins EC2 configuration. This isn’t an ideal situation and there’s been talk of getting Puppet to do the heavy lifting for a while.

So I sat down to do it.

Step 1: take the “apt-get install” line and convert it into puppet speak.

This was pretty easy. I started off with Vagrant starting a Ubuntu Lucid 32 VM (just like in the Vagrant getting started guide) and enabled the provision using puppet bit.

Step 2: find out you need to run “apt-get update”

Since the base VM I’m using was made there had been updates, so I needed to make any package installation depend on running “apt-get update” to ensure I was both installing the latest version and that the repositories would have the files I was looking for.

This was pretty easy (once I knew how):

exec {"apt-update":
       command => "/usr/bin/apt-get update",
}

Exec["apt-update"] -> Package <| |>

This simply does two things: specify to run “apt-get update” and then specify that any package install depends on having run “apt-update” first.

I’ve also needed things such as:

case $operatingsystem {
     debian, ubuntu: { $libaiodev = "libaio-dev" }
     centos, redhat: { $libaiodev = "aio-devel" }
     default: { fail("Unrecognised OS for libaio-dev") }
}
package { "libaio-dev":
          name => $libaiodev,
          ensure => latest,
}

The idea being that when I go and test all this stuff running on CentOS, it should mostly “just work” there too.

The next step? Setting up and running the Jenkins slave.