Production Kubernetes with SQL DB on AWS

Kubernetes has been the platform of choice for Container Orchestration for years now. Amazon, despite being the industry leading Cloud Provider only released Kubernetes as a Service in 2018. In this post I will explain how a Kubernetes Cluster can be created using command line on Amazon. The source code for this post can be found here. Note that you need to know AWS and Kubernetes fairly well as I will only cover certain sections sparingly.

Prerequisites

A Key Pair in your region
A Hosted Zone in Route 53( e.g. example.com)
An ACM Certificate for your domain( e.g. for example.com and *.example.com)
AWS CLI
eksctl
helm 3

Basic EKS Setup

eksctl simplifies EKS setup. To create a simple cluster create eks-basic.yaml from the gist below. Change publicKeyNameto the name of the Key Pair in your region( us-east-1 ).

The execute the command below to create the cluster.

You will see a Kubernetes cluster made up of three t2-small in us-east-1 upon successful completion of the command. As of now EKS only supports Spot Instances in Unmanaged Node Groups.

Production EKS Setup

Shell Scripts 0 through 4( *-deploy.sh) will deploy and test the infrastructure. Do exercise caution when using the destroy scripts as the deletions will be permanent.

0-eks-deploy.sh

Deploys an EKS cluster with one On- Demand Node Group and one Spot Node Group, initializes Helm, installs a Spot Interrupt Handler and configures AWS ALB Ingress Controller and Cluster Autoscaler.

1-eks-test-deploy.sh

Deploys a Docker image, exposes it as a Service and adds an Ingress to it. DNS entries in your Hosted Zone will be updated when the Ingress gets created.

2-rds-deploy.sh

Deploys Aurora on 3 Private Subnets and creates a Service pointing to the Database Host.

3-rds-test-deploy.sh

Checks the connectivity to the Database from the Kubernetes Cluster by installing a MySQL Client Pod.

4-jenkins-deploy.sh

Deploys Jenkins in the given Namespace.

Final Thoughts

It is always a good idea to use a managed Kubernetes service if possible as maintaining the infrastructure does require substantial effort from the infrastructure team. DO NOT attempt to reinvent the wheel.