Reprovisioning machines with null_resource

null_resource doesn't create anything. It's a container for provisioners. Because it is not directly connected to any piece of the infrastructure, it's not a big deal to destroy it in order to retrigger provisioners it has defined on.

There are two types of provisioning we are doing right now with Puppet: the one-time Puppet installation and Puppet run, which should be retriggered in case the manifest changes--imagine that repository IP changed and somehow we still don't have a proper DNS server in place.

Note

If you have Puppet Master, all of it makes zero sense: modules and manifests are stored on the master, and the Puppet agent is running as a system service and applies manifests automatically every N minutes. On the contrary, this approach can be very handy if you decide to go for masterless setup, because in this case, you have a whole new set of problems of how to distribute your Puppet code to all the servers you have.

Slim down the provisioners...