AWS EKS-based Services VPC

This project aims to create a blueprint of a complete Services VPC, fully containerised and based on AWS EKS, Terraform and Helm

Requirements

The automated deployment process through Makefile requires Mac OS X. You need to have fulfill the following requirements:

• AWS CLI tools installed (the aws command)
• AWS credentials (Access and Secret key of your IAM user) set up in your profile (the command aws configure will help)
• "GNU GCC make" installed (it should come by default with Mac OS and Linux)

How to deploy automatically (Mac OS X)

I got a little bit too excited with the Makefile approach and I created sections to perform every action needed for the deployment. You need to execute the highlighted commands on the root directory of the repository.

1. make mac-install: (optional) installs all dependencies in your Mac:
   • Terraform
   • Helm
   • aws-iam-authenticator
2. make init: uses Terraform to initialise your local copy of the deployment
3. make plan: (optional) uses Terraform to calculate changes to be done on an already-deployed environment once the terraform code has been updated. It's not necessary during the first deployment.
4. make apply: uses Terraform to deploy or update:
   • A VPC with some generic IP addresses.
   • Private and public subnets along with their respective NAT Gateways for Internet Access.
   • An optional bastion host to enable external access to that VPC (not needed for accessing the tools)
   • An AWS EKS cluster with a small instance (edit the ./2-k8sbase/cluster.tf to make changes on instance type and number of instances)
   • An EFS volume to store the Persistent Data Layer of used by the applications.
   • K8s EFS provider to the Kubernetes pods can create volumes on the EFS volume.
   • A k8s namespace that is not being used (everything is deployed on the default one... oooops, sorry!)
5. make app-deploy: uses Helm to deploy a Jenkins master on the EKS cluster using the EFS volume as storage backend.

How to destroy

This process will remove the whole deployment, it's very handy on testing scenarios with small budget.

make destroy-all

should destroy everything, but it's best to keep an eye on Terraform's logs in case something cannot be deleted and it's being left behind.