Blog-Technical-Content VS atlantis

We will set up a GitHub repository with a simple Terraform file that deploys an EC2 instance on AWS. You can find the demo contents here.

Next, we will go through an example of creating a custom module that takes as input a string that represents an epoch timestamp and converts it to its human-readable equivalent of type datetime in Python. You can find the code for this tutorial on this repository.

This is an example play to try out our new webserver role. Let’s go ahead and execute this play. To follow along, you should first run the vagrant up command from the top directory of this repository to create our target remote host.

Let’s examine how we can use all this in a real-world example. In this GitHub repository, we define the Terraform configuration for this example’s infrastructure. To follow along, you will need to install Terraform, have an AWS account ready, and authenticate with your AWS keys via the command line. Note that you might be charged a few dollars in your AWS account if you follow along.

If you are still learning how to use Ansible, you might also find helpful the introductory Ansible Tutorial or Working with Ansible Playbooks blog posts. You can find this article’s code on this repository if you wish to follow along.

Our primary entry point is main.tf, and in simple use cases, we can add all our resources there. We define our variables in variables.tf and assign values to them in terraform.tfvars. We use the file outputs.tf to declare output values. You can find a similar example project structure here.

None of these are a replacement of Terraform Cloud (recently rebranded to HCP Terraform). For example, when you create a PR, it could affect multiple workspaces. The new experimental version of TFC/TFE (I refuse to call it HCP!) implements Stacks, which is something like a workflow, and links one workspace output to other workspace inputs. None of the open-source solutions, including the paid Digger [0], support this - only the paid one, such as Spacelift [1] (which is the closest to TFC if you ask me). Having a monorepo of Terraform is a common design pattern, so, if I change an embedded module, it could trigger changes it many workspaces. As far as I know, Atlantis [2] can't really help in this case.

By the way, the reason I singled-out Spacelift is due to its quality, and the great Terraform provider it has. Scalr [3], for example, has a really low-quality Terraform provider. I extensively use the hashicorp/tfe provider to manage TFC itself.

[0]: https://digger.dev/

[1]: https://spacelift.io/

[2]: https://www.runatlantis.io/

[3]: https://www.scalr.com/

Atlantis is a pull request automation tool that works well with plain Terraform right away. But what if we're already using Terramate to generate Terraform code?

Atlantis automates reviewing and deploying Terraform via pull requests, streamlining collaboration and ensuring consistency across Terraform deployments.

For example, Atlantisgo for Terraform, Zapier’s Kubechecks for Argo CD, Quizlet’s GitHub action all do something similar to this. But a generic, extensible tool for IaC providers doesn’t seem to exist. Additionally, many of them require exposing your Kubernetes cluster or other infrastructure to third-party access, webhooks, etc.

Our first attempt was to introduce other engineering teams to Terraform - the Platform team was already using it extensively with Terragrunt, and using Atlantis to automate plan and apply operations in a Git flow to ensure infrastructure was consistent. We'd written modules, with documentation, and an engineer would simply need to raise a PR to use the module and provide the right values, and Atlantis (once the PR was approved by Platform) would go ahead and set it up for them.

Alternatively, you can look at solutions like Atlantis or spacelift.

Atlantis: https://www.runatlantis.io/

Looking at the commits at https://github.com/runatlantis/atlantis, it looks like 1.6.5. Am I right?

We use Atlantis [0] for CI/CD automation of Terraform pull requests to a centralized repository. It's pretty good too, especially for a self-hosted solution. I can't see how Terraform Cloud's costs would be justifiable for us without a custom contract.

[0] https://www.runatlantis.io/