README.md

Introduction

Terraform cluster definition leveraging the Docker provider, using Docker containers as Kubernetes nodes.

This configuration can be used for having a convenient way to start a Kubernetes cluster in your local laptop, using regular Docker containers as nodes of your cluster.

How does it work?

The Docker daemon can be run in a Docker container in what is usually called a DnD (Docker-in-Docker) configuration. This requires a special image/Dockerfile that has been tweaked for starting systemd as the the entrypoint. systemd will them start the Docker daemon as well as the kubelet. Once all these elements are running, we can run kubeadm as in any other platform for starting a Kubernetes cluster.

Pre-requisites

• Docker

  You will need a functional Docker daemon running. Make sure the ${var.daemon} is properly set, pointing to a daemon where you can launch containers.

• kubectl

  A local kubectl executable.

Contents

• Cluster definition
• Variables

Machine access

By default all the machines will have the following users:

• All the instances have a root user with linux password.

Topology

The cluster will be made by these machines:

• ${var.master_count} master nodes, with kubeadm and the kubelet pre-installed.
• ${var.worker_count} worker nodes, with kubeadm and the kubelet pre-installed.

You should be able to ssh these machines, and all of them should be able to ping each other.

Status

There is a bug in the Terraform Docker provider that prevents containers from being stopped when using rm=true, so there are some problems when re-creating resources.