terraform

Terraform

This folder contains the files to deploy the T2-Project with Terraform for different target environments: local (kind), AWS (EKS) and Azure (AKS).

kind is a tool for running local Kubernetes clusters using Docker container "nodes". Amazon Elastic Kubernetes Service (EKS) is the managed Kubernetes offering by AWS. Azure Kubernetes Service (AKS) is the managed Kubernetes offering by Azure.

Quick Start

For illustration purposes, AWS is used here as the target environment, but can easily be replaced with Azure or kind.

1. Take a look into the file environments/aws/main.tf to check if your desired modules are enabled/disabled.
2. Login to AWS in your terminal.
3. Run the start script: ./aws-start.sh
4. Get yourself a cup of tea, make yourself comfortable and wait for around 20 minutes
5. Check if everything works:
   • kubectl cluster-info
   • kubectl get pods

If you don't need the cluster anymore, stop everything and delete the cluster: ./aws-stop.sh

Folder Structure

Isolation via file layout is used to separate the different environments and state files from each other:

• environments/aws
• environments/azure
• environments/kind

To execute a terraform command you have to specify which environment you want to use.

Example using aws environment:

terraform -chdir=./environments/aws/ init  -upgrade

If you are using the scripts provided in this folder (see section Scripts below), the correct environment directory is used automatically.

Terraform Providers

Terraform requires that providers know their configuration during plan time. However, the providers kubernetes, helm and kubectl know how to connect to the cluster after it was created by Terraform.

Therefore, we have to ensure that the cluster is created as a separate apply-step to prevent any errors. For that we are using the Terraform apply parameter -target in the start scripts.

Example:

terraform -chdir=./environments/aws/ apply -target="module.eks" -auto-approve

For more information see:

• Stacking with managed Kubernetes cluster resources – Terraform Docs
• How to make Terraform provider dependent on a resource being created – Stack Overflow

Modules

To see which modules are used when executing terraform apply take a look into the file main.tf for your target environment (e.g. environments/aws/main.tf). To disable a module you don't need, e.g. Kepler, comment out the respective module in the file main.tf.

Scripts

For each environment there are a start, a start-with-microservices and a stop script, e.g. aws-start.sh, aws-start-with-microservices.sh and aws-stop.sh. The basic start script only creates a cluster in the respective environment. The start-with-services script creates the cluster and installs also the T2-Project services. The stop script deletes all resources and the cluster.

Optionally you can provide an argument to the scripts that is used as the namespace for the T2-Project services. For example, ./aws-start-with-microservices.sh t2-project will deploy the services of the T2-Project in the namespace t2-project.

The scripts are using the terraform apply parameter -auto-approve. Please be aware of that!

There is another script called run.sh that can be used to execute arbitrary terraform commands for a specific environment in a simplified way.

Example:

./run.sh aws init -upgrade

Rigorous Clean-up

The stop scripts executes terraform destroy that deletes the most resources. However, typically, it leaves some resources. If you want to delete all AWS resources associated with your account in a specific region, you can use cloud-nuke. This is a highly destructive operation! Please be aware of that!

Run the script:

./aws-nuke.sh