Creating Terraform Templates on OpenStack¶
Overview¶
This advanced tutorial will guide you through setting up a Terraform project and deploying virtual machines (VMs) with Docker & lightweight Kubernetes (Rancher K3s).
Goals
Understand how to install software and provision multiple VMs
Understand how Terraform is used to orchestrate VMs together
Ability to create a Terraform deployment for a multi-node JupyterHub
Ability to re-provisioning an already deployed cluster on OpenStack through Terraform
Things we won't cover
Basic management of Kubernetes clusters (that is a different lesson)
All of Terraform's features
Prerequisites¶
-
Basic understanding of OpenStack and Terraform as was covered in the prior lesson
-
Access to an OpenStack cloud (we will use Jetstream2)
-
Terraform installed on your workstation or local machine
Outcomes
By the end of this tutorial, you will
have launched a lightweight Kubernetes (K3s) cluster
know how to modify the size of cluster to scale up and down with more or less resources
started and stopped a JupyterHub using a public Docker image
Updating Terraform Configurations¶
We will re-use our configurations from the Terraform basics lesson
We will update and append the following files:
variables.tf - we are going to be adding new variables to include specific features related to our K3s and JupyterHub
instances.tf - we are going to be launching multiple VM flavors to support our cluster. We want the server VM to be large enough to manage all of the JupyterHub worker nodes.
security.tf - we need to specify all of the new security and networking settings for K3s and JupyterHub
We will create new configuration files for:
k3s.tf - we will keep our Kubernetes (Rancher) configuration files here
docker.tf - we will specify the JupyterHub Docker image here
jupyterhub.tf - we will configure our JupyterHub here
instances.tf¶
Our new cluster is going to be variably sized, meaning we can re-size or re-scale the number of VMs running in our cluster to account for demand by our users.
k3s.tf¶
docker.tf¶
We are going to be using the featured Datascience Latest image from the Project Jupyter team