Cloud-init with Terraform in vSphere environment… a step forward

I already write here about the ability to use Cloud-init in conjunction with Terraform to build an efficient and infrastructure-as-code system to cover infrastructure and configuration setup using the same deploy engine. Now. it’s time to take a step forward, showing how to use a for-each statement with Terraform to deploy different virtual machines using a common baseline, but specializing them according to their cloud-init specifications.

Terraform map and for-each usage

Coding the infrastructure assumes you write a “code” that describes your infrastructure introducing also some “tricky” element that increases the “agility” and like happens when you write code, you should ensure that all identical statements must be optimized using cycles, function calls, etc.

With Terraform it’s possible to put variables and declarations in two separate files: variables.tf and terraform.tfvars. The first is the “description” of the variables involved in the elements declared in the other files.

Let’s focus on “vms”: this is a map of objects that creates a dynamic structure that could be the source of an iteration. With the for_each statement is possible to iterate “vms” variable grabbing and using each key inside the resource brackets:

Variables declaration in terraform.tfvars should be:

You’ll find this example in my GitHub repo: https://github.com/linoproject/terraform/tree/master/rancher-lab

The configurations directories

Using the object-maps with the fore-each pattern is possible to dynamically pass variables but also pass entire configurations with cloud-init. We already saw in the previous post, how it’s easy to place a baseline configuration and a user-defined configuration which describes what specialization must be performed for every virtual machine.

In the using object element as a string pèattern for a multiple directory definition it’s possible to organize multiple cloud-init files that could be applied for every virtual machine:

So user-data files should contain only what is “special” for this virtual machine like packages, IP hostname,…

Passing variables to cloud-init declarations

Let’s do a step more… In order to pass variables to the cloud-init file, it’s possible to use the template_file terraform data source. Then, it’s possible to implement the for-each statement to pass the specific values to the template, cycling to “vms” object list:

In this way, using this template mechanism opens the ability to reduce the number of cloud-init files separating the common element from the VM specific elements. In this example, kickstart.yaml and metadata.yaml could be transformed into two templates. Here the templates:

Finally, in the main file the injection of the rendered template should be:

Then it’s possible to transform kickoff file in a template, in order to pass user/password credentials.

Time to show

You’ll find the fully functional example here: https://github.com/linoproject/terraform/tree/master/rancher-lab

Don’t forget to change/create terraform.tvars for your environment.

   Send article as PDF   

Un pensiero riguardo “Cloud-init with Terraform in vSphere environment… a step forward

I commenti sono chiusi