Infrastructure as Code Explained

Introduction

Cloud computing provides on-demand computing resources, which are decoupled from physical hardware and necessary underlying configuration. Autonomous software systems provision these computing resources in the cloud to achieve the automation that cloud computing offers. Because of such automation, it’s possible to control and manipulate the available resources programmatically by interfacing with the cloud providers. This way, infrastructure changes (such as resource scaling) can be implemented more quickly and reliably and operated mostly without manual interaction, but still with the ability to oversee the whole process and revert changes if something does not go according to plan.

Infrastructure as Code (IaC) is the approach of automating infrastructure deployment and changes by defining the desired resource states and their mutual relationships in code. The code is written in specialized, human-readable languages of IaC tools. The actual resources in the cloud are created (or modified) when you execute the code. This then prompts the tool to interface with the cloud provider or deployment system on your behalf to apply the necessary changes, without using a cloud provider’s web interface. The code can be modified whenever needed—upon code execution the IaC tool will take care of finding the differences between the desired infrastructure in code and the actual infrastructure in the cloud, taking steps to make the actual state equal to the desired one.

For IaC to work in practice, created resources must not be manually modified afterward (an immutable infrastructure), as this creates discord between the expected infrastructure in code and the actual state in the cloud. In addition, the manually modified resources could get recreated or deleted during future code executions, and all such customization would be lost. The solution to this is to incorporate the modifications into the infrastructure code.

In this conceptual article, we’ll explore the IaC approach, its benefits, and examples of real-world implementations. We’ll also introduce Terraform, an open source IaC provisioning tool. We’ll review Terraform’s role in this approach and how it compares to other IaC tools.

Benefits of IaC

With IaC, you can quickly create as many instances of your entire infrastructure as you need, in multiple provider regions, from a single source of truth: your declarative code. This has many advantages that ensures you’re creating resources consistently without error while reducing management and manual setup time.

The main benefits of IaC are:

• Deployment: removing the manual provisioning interaction with cloud providers means a quicker deployment speed.
• Recovery: identifying issues in the configuration can mean quick recovery from failures.
• Consistency: deploying resources are the same each time, resolving infrastructure fragility.
• Modification: modifying resources can have a quick turnaround time.
• Reusability: reusing parts of the infrastructure architecture in future projects.
• Version Control: storing the infrastructure code in version control systems.
• Visibility: writing configuration as code acts as documentation for the infrastructure.

Within an IaC workflow you can repeatedly spin up infrastructure in a standardized fashion, which means software development and testing is a quicker process because development, staging, quality-assurance testing, and production environments are separated. You can repeat the process of writing code and testing it live by deploying the infrastructure as many times as needed. Once your written infrastructure fulfills all requirements, you can deploy it in desired cloud environments. When new requirements arise, you can reiterate the process.

IaC, being based on code, should always be coupled with version control software (VCS), such as Git. Storing your infrastructure declarations in VCS makes it easily retrievable, with changes visible to everyone on your team, and provides snapshots at historical points, so you can always roll back to an earlier version if new modifications create errors. Advanced VCS can be configured to automatically trigger the IaC tool to update the infrastructure in the cloud when an approved change is added.

You now know what the IaC approach is, and what benefits it brings. You’ll now learn about states, the resource tracking mechanism employed in IaC. Then, you’ll follow up with the role of Terraform and other tools using IaC.

What is ‘state’?

In an IaC environment, the term ‘state’ refers to the state of affairs of desired infrastructure resources in a deployment. There are at least three states at a given moment: the actual one in the cloud, the ideal state presented in code, and the cached state that the IaC tool maintains. The cached state describes the state in the cloud as it was when the code was last executed. Terraform allows you to deploy the same code multiple times, forming multiple states for each deployment.

The actual state in the cloud (of the managed resources) should always be the same as the cached state of the tool. When executing the code, the tool will compare the ideal state with the cached one and apply the detected differences to the cloud. If the cached and actual states do not match, it’s highly likely that the execution will fail or that resources will be incorrectly provisioned.

Role of Terraform in IaC

Terraform is an open source IaC resource provisioning tool, written in Go and developed by Hashicorp. It supports multiple cloud providers, including DigitalOcean. The infrastructure definitions are written in the Hashicorp Configuration Language (HCL), and source code files written in it have the file extension tf.

Terraform works by reading the code describing your infrastructure and generating a graph containing all resources with their mutual relationships. It then compares it to the cached state of the resources in the cloud, and prepares an execution plan that details what will be applied to the cloud, and in what order, to reach the desired state.

The two main types of underlying components in Terraform are providers and provisioners. Providers are responsible for interacting with a given cloud provider, creating, managing, and deleting resources, while provisioners are used to execute specific actions on created remote resources or the local machine the code is executing on.

Terraform supports managing basic cloud provider components, such as compute instances, load balancers, storage, and DNS records, though more providers and provisioners can be added, owing to its extensible nature.

In IaC Terraform’s role is to ensure that the state of resources in the cloud is equal to the state expressed in code. It does not monitor the deployed resources, and its main focus is not on further bootstrapping of the provisioned compute instances with software and tasks. In the next section, you’ll learn how it compares to other tools and how they extend each other in a typical workflow.

Tools Using IaC

The IaC approach is widespread in modern deployment, configuration management, virtualization, and orchestration software. Docker and Kubernetes, the leading tools used for container creation and orchestration, both use YAML as their language for declaring the desired end result. On the other hand, Hashicorp Packer, a tool for creating snapshots of deployments, uses JSON to declare the template and variables from which a snapshot of the system will be built.

Ansible, Chef, and Puppet, the three most popular configuration management tools, all use the IaC approach to define the desired state of the servers they manage.

Ansible bootstraps provided servers according to a given playbook. A playbook is a textual file written in appropriate YAML, instructing Ansible what operations to perform on the existing target resources. Examples of such operations include running and starting services, installing packages using the system-provided package manager, or executing custom bash commands. To learn more about writing Ansible playbooks, read Configuration Managment 101: Writing Ansible Playbooks.

Chef and Puppet both require central servers with agents installed on each of the managed ones. Unlike Ansible, Chef uses Ruby, and Puppet uses its own declarative language for describing the resources.

Terraform is not mutually exclusive with other IaC tools and DevOps systems. Its strength is in provisioning hardware resources, rather than further software installation and initial server setup.

Unlike configuration management tools such as Ansible and Chef, Terraform is not suitable for installing software on the target resources and setting up tasks. Instead Terraform offers providers for interacting with their supported resources.

Terraform can work from a single machine and does not require central servers with client agents installed on the provisioned resources, unlike some other tools. It does not check their actual state and reapplies the configuration automatically, because its main focus is on provisioning them. A typical workflow is to provision the infrastructure resources using Terraform and then bootstrap them using a configuration management tool, if needed.

For Chef, Terraform has a built-in provisioner that sets up its client agent on provisioned remote resources. With it you can automatically have all your provisioned servers added to the main server, from where you can additionally configure them using cookbooks, Chef’s infrastructure declarations. You can learn more about writing them in Configuration Management 101: Writing Chef Recipes.

Conclusion

This article covered the paradigms of the IaC approach, its advantages over traditional manual system administration, the basics of Terraform as an IaC resource provisioning tool, and how it compares to other popular infrastructure automation tools.

If you’re looking to incorporate Infrastructure as Code into your workflow, check out our Terraform series to learn the fundamentals of using this tool in your development and deployment process.

One way to start with Terraform is to read How To Structure Your Terraform Project to understand how to ensure your infrastructure stays scalable and extensible.