Just add Code: Fun with Terraform Modules and AWS

Note: This article originally appeared in the 2016 AWS Advent. For the original article, click here.

Note that this article was written for Terraform v0.7.x – there have been several developments since this release that makes a number of the items covered here obsolete and they will be covered in the next article. 🙂

This article is going to show you how you can use Terraform, with a little help from Packer and Chef, to deploy a fully-functional sample web application, complete with auto-scaling and load balancing, in under 50 lines of Terraform code.

You will need the sample project to follow along, so make sure you load that up before continuing with reading this article.

The Humble Configuration

Check out the code in the terraform/main.tf file.

It might be hard to think that with this mere smattering of Terraform is setting up:

  • An AWS VPC
  • 2 subnets, each in different availability zones, fully routed
  • An AWS Application Load Balancer
  • A listener for the ALB
  • An AWS Auto Scaling group
  • An ALB target group attached to the ALB
  • Configured security groups for both the ALB and backend instances

So what’s the secret?

Terraform Modules

This example is using a powerful feature of Terraform – the modules feature, providing a semantic and repeatable way to manage AWS infrastructure. The modules hide most of the complexity of setting up a full VPC behind a relatively small set of code, and an even smaller set of changes going forward (generally, to update this application, all that is needed is to update the AMI).

Note that this example is composed entirely of modules – no root module resources exist. That’s not to say that they can’t exist – and in fact one of the secondary examples demonstrates how you can use the outputs of one of the modules to add extra resources on an as-needed basis.

The example is composed of three visible modules, and one module that operates under the hood as a dependency

  • terraform_aws_vpc, which sets up the VPC and subnets
  • terraform_aws_alb, which sets up the ALB and listener
  • terraform_aws_asg, which configures the Auto Scaling group, and ALB target group for the launched instances
  • terraform_aws_security_group, which is used by the ALB and Auto Scaling modules to set up security groups to restrict traffic flow.

These modules will be explained in detail later in the article.

How Terraform Modules Work

Terraform modules work very similar to basic Terraform configuration. In fact, each Terraform module is a standalone configuration in its own right, and depending on its pre-requisites, can run completely on its own. In fact, a top-level Terraform configuration without any modules being used is still a module – the root module. You sometimes see this mentioned in various parts of the Terraform workflow, such as in things like error messages, and the state file.

Module Sources and Versioning

Terraform supports a wide variety of remote sources for modules, such as simple, generic locations like HTTP, or Git, or well-known locations like GitHub, Bitbucket, or Amazon S3.

You don’t even need to put a module in a remote location. In fact, a good habit to get into is if you need to re-use Terraform code in a local project, put that code in a module – that way you can re-use it several times to create the same kind of resources in either the same, or even better, different, environments.

Declaring a module is simple. Let’s look at the VPC module from the example:

module "vpc" {                                                       
  source                  = "github.com/paybyphone/terraform_aws_vpc?ref=v0.1.0"
  vpc_network_address     = "${var.vpc_network_address}"             
  public_subnet_addresses = ["${var.public_subnet_addresses}"]       
  project_path            = "${var.project_path}"                    

The location of the module is specified with the source parameter. The style of the parameter will dictate what kind of behaviour TF will undertake to get the module.

The rest of the options here are module parameters, which translate to variables within the module. Note that any variable that does not have a default value in the module is a required parameter, and Terraform will not start if these are not supplied.

The last item that should be mentioned is regarding versioning. Most module sources that work off of source control have a versioning parameter you can supply to get a revision or tag – with Git and GitHub sources, this is ref, which can translate to most Git references, be it a branch, or tag.

Versioning is a great way to keep things under control. You might find yourself iterating very fast on certain modules as you learn more about Terraform or your internal infrastructure design patterns change – versioning your modules ensures that you don’t need to constantly refactor otherwise stable stacks.

Module Tips and Tricks

Terraform and HCL is a work in progress, and there may be some things that seem like they may make sense that don’t necessarily work 100% – yet. There are some things that you might want to keep in mind when you are designing your modules that may reduce the complexity that ultimately gets presented to the user:

Use Data Sources

Terraform 0.7+’s data sources feature can go a long way in reducing the amount of data needs to go in to your module.

In this project, data sources are used for things such as obtaining VPC IDs from subnets (aws_subnet) and getting the security groups assigned to an ALB (using the aws_alb_listener and aws_alb data sources chained together). This allows us to create ALBs based off of subnet ID alone, and attach auto-scaling groups to ALBs with knowing only the listener ARN that we need to attach to.

Exploit Zero Values and Defaults

Terraform follows the rules of the language it was created in regarding zero values. Hence, most of the time, supplying an empty parameter is the same as supplying none at all.

This can be advantageous when designing a module to support different kinds of scenarios. For example, the alb module supports TLS via supplying a certificate ARN. Here is the variable declaration:

// The ARN of the server certificate you want to use with the listener.
// Required for HTTPS listeners.
variable "listener_certificate_arn" {
  type    = "string"
  default = ""

And here it is referenced in the listener block:

// alb_listener creates the listener that is then attached to the ALB supplied
// by the alb resource.
resource "aws_alb_listener" "alb_listener" {
  certificate_arn   = "${var.listener_certificate_arn}"

Now, when this module parameter is not supplied, its default value becomes an empty string, which is passed in to aws_alb_listener.alb_listener. This is, most times, exactly the same as if the parameter is not passed in at all. This allows you to not have to worry about this parameter when you just want to use HTTP on this endpoint (the default for the ALB module as a whole).

Pseudo-Conditional Logic

Terraform does not support conditional logic yet, but through creative use of count and interpolation, one can create semi-conditional logic in your resources.

Consider the fact that the terraform_aws_autoscaling module supports the ability to attach the ASG to an ALB, but does not explicit require it. How can you get away with that, though?

To get the answer, check one of the ALB resources in the module:

// autoscaling_alb_target_group creates the ALB target group.
resource "aws_alb_target_group" "autoscaling_alb_target_group" {
  count    = "${lookup(map("true", "1"), var.enable_alb, "0")}"

Here, we make use of the map interpolation function, nested in a lookup function to provide essentially an if/then/else control structure. This is used to control a resource’s instance count, adding an instance if var.enable_alb is true, and completely removing the resource from the graph otherwise.

This conditional logic does not necessarily need to be limited to count either. Let’s go back to the aws_alb_listener.alb_listener resource in the ALB module, looking at a different parameter:

// alb_listener creates the listener that is then attached to the ALB supplied
// by the alb resource.
resource "aws_alb_listener" "alb_listener" {
  ssl_policy        = "${lookup(map("HTTP", ""), var.listener_protocol, "ELBSecurityPolicy-2015-05")}"

Here, we are using this trick to supply the correct SSL policy to the listener if the listener protocol is not HTTP. If it is, we supply the zero value, which as mentioned before, makes it as if the value was never supplied.

Module Limitations

Terraform does have some not-necessarily-obvious limitations that you will want to keep in mind when designing both modules and Terraform code in general. Here are a couple:

Count Cannot be Computed

This is a big one that can really get you when you are writing modules. Consider the following scenario that totally did not happen to me even though I knew of of such things beforehand 😉

  • An ALB listener is created with aws_alb_listener
  • The arn of this resource is passed as an output
  • That output is used as both the ARN to attach an auto-scaling group to, and the pseudo-conditional in the ALB-related resources’ count parameter

What happens? You get this lovely message:

value of 'count' cannot be computed

Actually, it used to be worse (a strconv error was displayed instead), but luckily that changed recently.

Unfortunately, there is no nice way to work around this right now. Extra parameters need to be supplied or you need to structure your modules in way that avoids computed values being passed into countdirectives in your workflow. (This is pretty much exactly why the terraform_aws_asg module has a enable_alb parameter).

Complex Structures and Zero Values

Complex structures are not necessarily good candidates for zero values, even though it may seem like a good idea. But by defining a complex structure in a resource, you are by nature supplying it a non-zero value, even if most of the fields you supply are empty.

Most resources don’t handle this scenario gracefully, so it’s best to avoid using complex structures in a scenario where you may be designing a module for re-use, and expect that you won’t be using the functionality defined by such a structure often.

The Application in Brief

As our focus in this article is on Terraform modules, and not on other parts of the pattern such as using Packer or Chef to build an AMI, we will only touch up briefly on the non-Terraform parts of this project, so that we can focus on the Terraform code and the AWS resources that it is setting up.

The Gem

The Ruby gem in this project is a small “hello world” application running with Sinatra. This is self-contained within this project and mainly exists to give us an artifact to put on our base AMI to send to the auto-scaling group.

The server prints out the system’s hostname when fetched. This will allow us to see each node in action as we boot things up.


The built gem is loaded on to an AMI using Packer, for which the code is contained within packer/ami.json. We use chef-solo as a provisioner, which works off a self-contained cookbook named packer_payload in the cookbooks directory. This allows us a bit more of a higher-level workflow than we would have simply with shell scripts, including the ability to better integration test things and also possibly support multiple build targets.

Note that the Packer configuration takes advantage of a new Packer 0.12.0 feature that allows us to fetch an AMI to use as the base right from Packer. This is the source_ami_filter directive. Before Packer 0.12.0, you would have needed to resort to a helper, such as ubuntu_ami.sh, to get the AMI for you.

The Rakefile

The Rakefile is the build runner. It has tasks for Packer (ami), Terraform (infrastructure), and Test Kitchen (kitchen). It also has prerequisite tasks to stage cookbooks (berks_cookbooks), and Terraform modules (tf_modules). It’s necessary to pre-fetch modules when they are being used in Terraform – normally this is handled by terraform get, but the tf_modules task does this for you.

It also handles some parameterization of Terraform commands, which allows us to specify when we want to perform something else other than an apply in Terraform, or use a different configuration.

All of this is in addition to standard Bundler gem tasks like build, etc. Note that install and releasetasks have been explicitly disabled so that you don’t install or release the gem by mistake.

The Terraform Modules

Now that we have that out of the way, we can talk about the fun stuff!

As mentioned at the start of the article, This project has 4 different Terraform modules. Also as mentioned, one of them (the Security Group module) is hidden from the end user, as it is consumed by two of the parent modules to create security groups to work with. This exploits the fact that Terraform can, of course, nest modules within each other, allowing for any level of re-usability when designing a module layout.

The AWS VPC Module

The first module, terraform_aws_vpc, creates not only a VPC, but also public subnets as well, complete with route tables and internet gateway attachments.

We’ve already hidden a decent amount of complexity just by doing this, but as an added bonus, redundancy is baked right into the module by distributing any network addresses passed in as subnets to the module across all availability zones available in any particular region via the aws_availability_zones data source. This process does not require previous knowledge of the zones available to the account.

The module passes out pertinent information, such as the VPC ID, the ID of the default network ACL, the created subnet IDs, the availability zones for those subnets as a map, and the ID of the route table created.

The ALB Module

The second module, terraform_aws_alb allows for the creation of AWS Application Load Balancers. If all you need is the defaults, use of this module is extremely simple, creating an ALB that will answer requests on port 80. A default target group is also created that can be used if you don’t have anything else mapped, but we want to use this with our auto-scaling group.

The Auto Scaling Module

The third module, terraform_aws_asg, is arguably the most complex of the three that we see in the sample configuration, but even at that, its required options are very slim.

The beauty of this module is that, thanks to all the aforementioned logic, you can attach more than one ASG to the same ALB with different path patterns (mentioned below), or not attach it to an ALB at all! This allows this same module to be used for a number of scenarios. This is on top of the plethora of options available to you to tune, such as CPU thresholds, health check details, and session stickiness.

Another thing to note is how the AMI for the launch configuration is being fetched from within this module. We work off the tag that we used within Packer, which is supplied as a module variable. This is then searched for within the module via an aws_ami data source. This means that no code or variables need to change when the AMI is updated – the next Terraform run will pick up the most recent AMI with the tag.

Lastly, this module supports the rolling update mechanism laid out by Paul Hinze in this post oh so long ago now. When a new AMI is detected and the auto-scaling group needs to be updated, Terraform will bring up the new ASG, attach it, wait for it to have minimum capacity, and then bring down the old one.

The Security Group Module

The last module to be mentioned, terraform_aws_security_group, is not shown anywhere in our example, but is actually used by the ALB and ASG modules to create Security Groups.

Not only does it create security groups though – it also allows for the creation of 2 kinds of ICMP allow rules. One for all ICMP, if you so choose, but more importantly, allow rules for ICMP type 3 (host unreachable) are always created, as this is how path MTU discovery works. Without this, we might end up with unnecessarily degraded performance.

Give it a Shot

After all this talk about the internals of the project and the Terraform code, you might be eager to bring this up and see it working. Let’s do that now.

Assuming you have the project cloned and AWS credentials set appropriately, do the following:

  • Run bundle install --binstubs --path vendor/bundle to load the project’s Ruby dependencies.
  • Run bundle exec rake ami. This builds the AMI.
  • Run bundle exec rake infrastructure. This will deploy the project.

After this is done, Terraform should return a alb_hostname value to you. You can now load this up in your browser. Load it once, then wait about 1 second, then load it again! Or even better, just run the following in a prompt:

while true; do curl http://ALBHOST/; sleep 1; done

And watch the hostname change between the two hosts.

Tearing it Down

Once you are done, you can destroy the project simply by passing a TF_CMD environment variable in to rake with the destroy command:

TF_CMD=destroy bundle exec rake infrastructure

And that’s it! Note that this does not delete the AMI artifact, you will need to do that yourself.

More Fun

Finally, a few items for the road. These are things that are otherwise important to note or should prove to be helpful in realizing how powerful Terraform modules can be.


You may have noticed the modules have a project_path parameter that is filled out in the example with the path to the project in GitHub. This is something that I think is important for proper AWS resource management.

Several of our resources have machine-generated names or IDs which make them hard to track on their own. Having a easy-to-reference tag alleviates that. Having the tag reference the project that consumes the resource is even better – I don’t think it gets much clearer than that.

SSL/TLS for the ALB

Try this: create a certificate using Certificate Manager, and change the alb module to the following:

module "alb" {
  source                   = "github.com/paybyphone/terraform_aws_alb?ref=v0.1.0"
  listener_subnet_ids      = ["${module.vpc.public_subnet_ids}"]
  listener_port            = "443"
  listener_protocol        = "HTTPS"
  listener_certificate_arn = "arn:aws:acm:region:account-id:certificate/certificate-id"
  project_path             = "${var.project_path}"

Better yet, see the example here. This can be run with the following command:

TF_DIR=terraform/with_ssl bundle exec rake infrastructure

And destroyed with:

TF_CMD=destroy TF_DIR=terraform/with_ssl bundle exec rake infrastructure

You now have SSL for your ALB! Of course, you will need to point DNS to the ALB (either via external DNS, CNAME records, or Route 53 alias records – the example includes this), but it’s that easy to change the ALB into an SSL load balancer.

Adding a Second ASG

You can also use the ASG module to create two auto-scaling groups.

module "autoscaling_group_foo" {
  source            = "github.com/paybyphone/terraform_aws_asg?ref=v0.1.1"
  subnet_ids        = ["${module.vpc.public_subnet_ids}"]
  image_tag_value   = "vancluever_hello"
  enable_alb        = "true"
  alb_listener_arn  = "${module.alb.alb_listener_arn}"
  alb_rule_number   = "100"
  alb_path_patterns = ["/foo/*"]
  alb_service_port  = "4567"
  project_path      = "${var.project_path}"

module "autoscaling_group_bar" {
  source            = "github.com/paybyphone/terraform_aws_asg?ref=v0.1.1"
  subnet_ids        = ["${module.vpc.public_subnet_ids}"]
  image_tag_value   = "vancluever_hello"
  enable_alb        = "true"
  alb_listener_arn  = "${module.alb.alb_listener_arn}"
  alb_rule_number   = "101"
  alb_path_patterns = ["/bar/*"]
  alb_service_port  = "4567"
  project_path      = "${var.project_path}"

There is an example for the above here. Again, run it with:

TF_DIR=terraform/multi_asg bundle exec rake infrastructure

And destroy it with:

TF_CMD=destroy TF_DIR=terraform/multi_asg bundle exec rake infrastructure

You now have two auto-scaling groups, one handling requests for /foo/*, and one handling requests for /bar/*. Give it a go by reloading each URL and see the unique instances you get for each.


I would like to take a moment to thank PayByPhone for allowing me to use their existing Terraform modules as the basis for the publicly available ones at https://github.com/paybyphone. Writing this article would have been a lot more painful without them!

Also thanks to my editors, Anthony Elizondo and Andrew Langhorn for for their feedback and help with this article, and the AWS Advent Team for the chance to stand on their soapbox for my 15 minutes! 🙂


AWS World Detour – Packer and Terraform

NOTE: Click here to get to the sample project quickly.

So I know it’s been a while – quite a while – since I’ve posted anything. I’ve been far from inactive though, and I will touch up on that in a later post.

I will be getting back on track here with my AWS World Tour series soon, but first, I want to take a bit of a detour and discuss some things that have been highly relevant to what I have been working on as of the last little while – infrastructure provisioning with Terraform.

My Journey the Last 6 Months

I have had a very eventful last 6 months since I started to use AWS “in anger”. I’m not the biggest fan of that term, but I can’t really think of other way to Put it right now. There have indeed been some frustrating moments, but I can’t really say I was angry.

In any case, my journey to try and find a good AWS provisioning platform took me to a few places, namely:

  • Trying to use Chef – Which prompted me to write this pull request. However, I quickly realized that having to write AWS resources for everything that we needed to do at PayByPhone might not be the best use of my time.
  • I also was going to write my own tooling around managing CloudFormation, but stopped short when, one weekend, I asked myself if I was reinventing the wheel, and I kind of was.

Enter Terraform.


Terraform is a configuration management tool created by HashiCorp with an emphasis on virtual infrastructure, versus say something like Chef or Puppet, that has an emphasis on OS-level configuration management. In fact, the two can work in tandem – Terraform has a Chef provisioner.

What makes Terraform special is its ability to support many kinds of infrastructure platforms – and in fact, this is its stated goal. Things written in Terraform should be ultimately portable from one platform to another, say one wanted to move from AWS to Azure, or some kind of OpenStack-hosted solution somewhere. Admittedly, we had looked at Terraform earlier on in the year last year, and its AWS support had not been fully fleshed out yet, but that has seemed to have changed drastically, and by the end of last year its feature set was on par, and possibly even better, than CloudFormation itself.

So what if one just wanted to stick with AWS? What’s the point in using it over, say, simply CloudFormation? The answer to that is really dependent on the use case, and like many things in life, boil down to the little things. Some examples:

  • Support for ZIP file uploads to AWS Lambda (versus being restricted to S3 on CloudFormation).
  • Support for the AWS NAT Gateway. As of this writing, about a month since the NAT Gateway’s release, it still does not appear to be supported by
    CloudFormation, according to my very light research (judging by the traffic on this thread).
  • A non-JSON DSL that has support for several kinds of programmatic interpolation operations, including some basic forms of loop operations allowing one to create a certain amount of resources with a single chunk of code. The DSL also has support for modules, allowing re-use and distribution of common templates.

This is all possible by the fact that by using Terraform, one is not beholden to the CloudFormation way of doing things. Rather, Terraform uses the AWS API (thru the Go SDK), which in some instances is much more versatile than CloudFormation is, or CloudFormation could be, for that matter (as a hosted configuration management platform, they have to restrict some of their data sources to some reliable services – this is more than likely why S3 is the only way to get a ZIP file read in for Lambda).

Terraform is a fast moving target. I have submitted several pull requests myself, for bug fixes and adding functionality alike, and don’t see an end in sight for that, not at least for a few months. For example, one of my more recent feature PRs allows one to get details on an AMI for use in a template later. In order to do this in CloudFormation, one would have to undertake the tedious process of writing a custom resource which ultimately leads to more out-of-band resources, and unnecessary technical debt, in my opinion.


I’ve been using Packer for quite some time now to build base images, namely for our VMWare infrastructure at PayByPhone.

Packer is basically an image builder. Think of it this way – gem might build a Ruby gem to deploy on other systems, and one might build a static binary with go build. Packer is like this, but for system images. In fact, one might not have a custom application, and simply may need to build an image with some typical software on it and configured a certain way – in this instance, Packer and its provisioning code could live in the same repository and serve as a complete “application”.

Incidentally, the process for building AMIs is actually much simpler than using VMWare. There is a lot less code, and it’s basically AMI-in, AMI-out.

Packer Provisioners

Just a note on provisioners – Packer supports several of these. I use the Chef provisioner. There are also Ansible and shell provisioners. I would recommend using one of the configuration management options – it allows for the better re-use code, and also allows for testing before moving on to the Packer build. Namely, with Chef, I am able to use Berkshelf to manage dependencies, and Test Kitchen to sort out most, if not all, errors that would happen with a cookbook before moving on to creating the packer JSON file.

This also makes it more portable for use with something like Vagrant – generally all that’s needed is to copy the provisioner configuration from Vagrant to Packer, or vice versa, with possibly some minor modifications.

Putting it Together

With the above tools and a competent build runner, I can actually write an entire pipeline that takes an application and deploys it to AWS with relative ease. Better than that, this all can live together! This enables someone to go to a single repository to:

  • Get hands on the code, make changes, and run unit tests on it
  • Deploy using the provisioner code to an EC2 instance, or local to their machine with Vagrant, to run integration tests and experiment with the code
  • Fully test the infrastructure by building the AMI and deploying the infrastructure in an uniform fashion to multiple environments (ie: sandbox, staging, or production).

This allows for a pipeline that should ensure a near perfect deploy once the application is ready for production.

The Pattern in Action

As an example, I have taken the code from my previous article, (see here, here, here, and the code here), and applied the same idea, but with some changes. Again, I am deploying a VPC with ELB and 3 instances, but this time, I have skipped some of the details irrelevant to a VPC of this kind, especially since I will never be logging into it – namely the private network and NAT part of things.

The code can be found here. Let’s go over it together:

The application

The application is a simple Ruby “hello world” style application running with Sinatra. The application is bundled up as a Ruby gem – this is actually a pretty easy way to create this kind of application as it produces a single artifact that can be deployed where ever, especially on a “bare metal”, single- purpose system. This application could be even deployed even using the system’s base Ruby, if it is current enough (I don’t do that though, as I will show in later sections).

The layout of the application part of the sample repo is such:

  vancluever_hello       <-- Executable binary part of gem package
  vancluever_hello.rb    <-- Application &amp;amp;quot;entry point&amp;amp;quot; and Sinatra code
    version.rb           <-- Gem version file
pkg/                     <-- Output directory (gem gets built to here)
Gemfile                  <-- Bundler dependency file (chained to gemspec)
Rakefile                 <-- Rake build runner configuration file
vancluever_hello.gemspec <-- RubyGems package spec file

The bulk of the test code is in the lib/vancluever_hello.rb file, a simple file whose contents is shown below:

require 'sinatra/base'
require 'socket'

module VanclueverHello
  # Run the test server.
  class Server < Sinatra::Base
    def self.run_server
      set :bind, ''

      get '/' do
        "Hello from #{Socket.gethostname}!!!"


This is the Sinatra self-hosted version of what we were doing with the index.html files, Apache, and user data in the previous version of this stack. Rather than use a static file this time, we are using Sinatra and Ruby to demonstrate how this small app can be bundled onto an image and deployed from there, without any post-creation package installation and content writing.

The Rakefile contains the bundler/gem_tasks helper that allows us to easily build this gem from the details in the vancluever_hello.gemspec file. By running rake build, the .gem is dropped into the pkg/ directory, ready for the next step.

The packer_payload Chef cookbook

The next piece of the puzzle is the packer_payload Chef cookbook, self-contained in the cookbooks directory. This cookbook is not like other Chef cookbooks one might see – the metadata is stripped down (no version info, description, or even version locking). This is because this cookbook is not intended to be used anywhere else other than the Packer build that I will be discussing shortly.

Why Chef then? Why not shell scripts if this is all the cookbook is going to be used for? A couple of quick reasons that come to mind:

  • I’m not the biggest fan of shell scripts – I will use them when necessary, but I’m more a fan these days of writing things in a way that they can be easily re-used, and in a way that makes it easy to pull in things that make my job easier. Using Chef allows me to do that. For example, rather than having to write code to manage a non-distro Ruby, I use poise-ruby and poise-ruby-build to manage the Ruby version and gem package. Taking it further, rather than having to write scripts to template out upstart or systemd, I use poise-service, which supports both.
  • Even if this cookbook is not suitable for Supermarket, or to sit on a Chef server, it’s re-usability is not completely diminished. Test Kitchen can still be used with this, and in fact there is a .kitchen.cloud.yml file in the directory. Kitchen was used to test this cookbook before putting it into packer, ensuring that most, if not all, of the code worked before starting the process to build the AMI. This cookbook can also be used in Vagrant with minimal effort as well, should the need arise.

Getting the data to Chef

One thing that deserves mention is how I actually get the data to the Chef cookbook. The artifact does need to be delivered in some fashion to the cookbook itself. This is okay, mainly because I have Packer and Test Kitchen to help out with that.

In attributes/default.rb I control the location of the artifact:

default['packer_payload']['source_path'] = '/tmp/gem_pkg'

This is where the artifact is copied to with Packer (more on that soon). However, with Kitchen, things are a little different, because of how the data directory stuff works:

  name: chef_zero
  data_path: ../../pkg/

data_path controls the directory that contains any non-cookbook data that I want to send to the server. After this is done, I need to change the source_path node attribute:

  - name: default
      - recipe[packer_payload::default]
        source_path: /tmp/kitchen/data
        app_version: <%= ENV['KITCHEN_APP_VERSION'] %>

Also note the ERB in app_version – this is an environment variable passed in from Rake, which gets the data from the VanclueverHello::VERSION module in the gem code. More on this below.


As mentioned, the cookbook’s .kitchen.cloud.yml is fully functional, and the cookbook can be tested using the following command:

 AWS_KITCHEN_AMI_ID=ami-123456abcd \
 KITCHEN_YAML=.kitchen.cloud.yml \
 ../../bin/kitchen verify

Note the environment variables. Depending on the target being tested for, kitchen-ec2 may have trouble finding an AMI for the target. The one I seem to have the most luck with right now is Ubuntu Trusty (14.04) – but I wanted to try this against some of the more recent versions like Wily. This necessitated the need to supply the login user and the AMI ID, which I do through environment so that it can be parameterized. Also, I pass the app version, which helps show how I can control the version of the gem that gets installed. Also, it is a popular pattern to name EC2 or other cloud Kitchen config files as .kitchen.cloud.yml and call them by passing the KITCHEN_YAML environment variable.

All of this is also in the Rakefile, under the kitchen task. By doing this, I don’t have to worry about running this from the command line all the time. Further to that it takes the work out of determining the AMI to use (more on that later).

The Packer template

The Packer template lives in the packer/ami.json file, sitting at a nice 43 lines, fully beautified. It has variables that are used to tell Packer what AMI to get, what region to set it up in, and also some things to add to the description, such as the distribution and application version.

The Packer template is probably the most simplest part of this setup. All that is needed to kick off the AMI creation is a packer build packer/ami.json. Of course, the parameters need to be passed via environment variables though, but the Rakefile handles that.

Provisioners – artifact delivery and Chef

One thing that I will note about the Packer template is how it does its configuration work on the AMI – this is done via what are known in Packer as provisioners:

  "provisioners": [
      "type": "shell",
      "inline": ["mkdir /tmp/gem_pkg"]
      "type": "file",
      "source": "pkg/vancluever_hello-{{user `app_version`}}.gem",
      "destination": "/tmp/gem_pkg/vancluever_hello-{{user `app_version`}}.gem"
      "type": "chef-solo",
      "cookbook_paths": ["berks-cookbooks"],
      "run_list": ["packer_payload"],
      "json": {
        "packer_payload": {
          "app_version": "{{user `app_version`}}"

Note the first two – the shell and file provisioners, which deliver the artifact. Creation of the directory is necessary here (and Packer won’t fail if the directory does not exist – something that created about an extra 2 hours of troubleshooting work for me as I was making this example). The next one, the chef-solo provisioner, runs the packer_payload cookbook to configure things.

Note the cookbooks directory. It’s not cookbooks, but instead berks-cookbooks. This is because I’m staging the full, dependency evaluated cookbook collection in the berks-cookbooks directory, via Berkshelf. This is handled by the Rakefile ahead of the execution of Packer. I haven’t dived into the Rakefile yet, but I won’t just yet – first off, I want to introduce the star of the show.


One last thing before I move on. Tagging the AMI is important! This allows a search on this AMI afterwards. Also, it keeps me from having to parse the Packer logs for the AMI ID, which even though Packer makes easier with a machine readable output flag, I still find tagging to be less work to do (no capturing output or having to save a log file). It also gets one into the habit of tagging resources, which should be done anyway.

Note that it doesn’t have to be the application ID that is the tag or the “artifact”. In addition to this, one could also tag the build ID – which can provide even further granularity when searching on AMIs.


Finally I get to the headliner. Terraform.

The Terraform file sits in the terraform/main.tf file. This is only a single file, but it can be broken up in this directory into as much as it makes sense to break up such things. For example, a lot of my projects now have a variables.tf outputs.tf, vpc.tf, instance.tf, and more, depending on how things need to look. This allows for code chunks that are easy to read. As long as they are all in the same directory, Terraform will treat them all as the same plan.

Looking at the main.tf file, one may notice several analogs to the CloudFormation template from my last article. The key differences, save the difference in infrastructure to remove things that were not necessary for this article, are that it’s not JSON, and also that there is only one aws_instance resource, with a count of 3. count is a special Terraform DSL attribute that tells Terraform to make more than one of the specified resource. This is referenced in the aws_elb block too, with a splat operator:

resource "aws_elb" "elb" {

  instances = ["${aws_instance.web_servers.*.id}"]

Basically allowing me to reference all the instance IDs at once.

And of course, the Terraform file is parameterized. There are 3 parameters – region, ami_id, and vpc_subnet. region and ami_id are both required as they don’t have a default, but vpc_subnet does not need to be supplied if the default network is okay.

Manually, if the variables were supplied as TF_VAR_ environment variables (ie: TF_VAR_region or TF_VAR_ami_id), one could just run terraform apply and watch this thing go. In reality though, I want this going through Rake, and that’s what I do.

The Rakefile

Maybe not the star of the show, but definitely the thing keeping the lights on, is the Rakefile.

In addition to having a full DSL of its own to run builds with, Rake can be extended with standard Ruby. This can come in the form of simple methods within the Rakefile, or full-on helper libraries that can provide a suite of common tasks for a project. As mentioned, I used the bundler/gem_tasks helper to provide the basic RubyGems building tasks (and if in the Rakefile I have even disabled a few to ensure the gem doesn’t get accidentally pushed).

Incidentally, the Rake tasks make up only a small portion of this Rakefile. There are 4 user-defined tasks, berks-cookbooks, ami, infrastructure, and kitchen.

ami calls two prerequisites, build and berks-cookbooks. The former is a gem task, which builds the gem and puts it into the pkg/ directory. The second runs Berkshelf on the cookbooks/packer_payload/Berksfile file, Vendoring the cookbooks into the berks-cookbooks directory so that Packer has all cookbooks available to it during its chef-solo run. After these are done, packer can run, after getting some variables, of course. The same goes for the infrastructure task, which has no prerequisites, but sends some variables to the terraform command. Finally, the kitchen task allows for me to easily run tests on the packer_payload cookbook.

The Rakefile helper methods

This is where things really come together. There are a couple of kinds of helpers that I have here. The first kind are very simple and handle a few variables from the environment. Honestly, if Rake had something better to handle this, I would use that instead, but from what I’ve seen, it doesn’t (a bit more RTFM may be necessary). Also, I would rather use environment instead of the parameter system that Rake uses by default to provide parameters – it’s more in line with what the rest of our toolchain uses.

The second kind is where Rake really shines though. These are the functions ubuntu_ami_id, app_ami_id, and rfc3339_to_unix:

  • ubuntu_ami_id uses the ubuntu_ami gem to find the latest Ubuntu AMI for the distribution (default trusty) and root store type (ebs-ssd). This is fed to Packer.
  • After Packer is done and has tagged the AMI with our application tag, app_ami_id can go and get the latest built image for our system. Sorting is helped by the rfc3339_to_unix, which helps convert the timestamp. This AMI ID is then fed to Terraform.

The tasks to deploy

Finally, after all of this write up, what is needed to make this run? Three very easy commands:

  • After mirroring the repo, bundle install --binstubs --path vendor/bundle ensures all the dependencies are gathered up within the working directory tree.
  • Then, bundle exec rake ami will build the AMI.
  • Finally, bundle exec rake infrastructure will deploy the infrastructure with Terraform.

And presto! A 3-node ELB cluster in AWS. The Terraform output will have the DNS name of the ELB – after a few minutes when everything is available, connect over HTTP to port 4567 to see what has been created – refresh to see the page cycle through the IP addresses.

Destroying the infrastructure

After completing the exercise, I want to shut down these resources to ensure that they are not going to rack up a nice big bill for me. This is easily done with the setup I have:

TF_CMD=destroy bundle exec rake infrastructure

This will destroy all created resources. Afterwards, I delete the AMI and the snapshot that Packer made using the AWS CLI or console. And it’s like it never existed!

Final Word – the terraform.tfstate File

One important thing about this file. The terraform.tfstate contains a working state of the infrastructure and must be treated with respect. This is the part that’s actually hidden from when CloudFormation is used as AWS handles this part.

The terraform.tfstate file can be managed in one of two ways:

  • By checking it into source control (if using the example repo as a starting point, note that it is, by default, in the .gitignore file).
  • By using remote state to store the config. There are several options, such as S3, Consul, or Atlas, and a few others not mentioned here. Remote state also has the advantage of easy retrieval for use with other projects (other Terraform files, for example, can access its outputs).

Final Final Word – Modules

One thing that didn’t get mentioned here at all has been Terraform’s ability to use modules.

I suggest checking this out if you plan to really dive into Terraform. So much of your repeatable infrastructure code can be put in a module. In fact, the template in this example can serve as a module, if it lived in its own repository – then, all the template that referenced it would need is the 3 variables defined at the top, which would ultimately turn the main.tf file in its referencing project into an approximately half-dozen lines of code.

Well, that’s all for this article! As usual, I hope you found the material informative! I hope to get back on track with my initial intention of evaluating AWS services soon, but don’t hold it against me if things are slow. I have been far from inactive though – I will mention some of the things I have been up to in my next post. Until then, take care!