This repository shows examples and guides for using Terraform to provision a GKE (Google Kubernetes Engine) Cluster on Google Cloud Platform

Fork this repository into your own GitHub account and then clone (your forked version) down to your local machine.

We'll use the Google Command Line tool to get information from the cluster.

If you haven't already installed it then you can follow below to install.

To install this you can install manually by following these instructions or if you have a package manager you can use the instructions below:

Homebrew

brew install --cask google-cloud-sdk

Chocolatey

choco install gcloudsdk

Once it is installed we need to configure it to point at your Google Cloud Project - to do this run:

gcloud init

When prompted ensure you pick the correct Google Cloud Project.

The kubectl tool will be utilised to interact with your Kubernetes (K8S) cluster.

You can install this manually by following this guide below:

https://kubernetes.io/docs/tasks/tools/install-kubectl/

Or alternatively if you use a package manager it can be installed using the package manager:

Homebrew

brew install kubernetes-cli

Chocolatey

choco install kubernetes-cli

You can verify that it has installed correctly by running

kubectl version

It should print something like the below:

Client Version: version.Info{Major:"1", Minor:"20", GitVersion:"v1.20.4", GitCommit:"e87da0bd6e03ec3fea7933c4b5263d151aafd07c", GitTreeState:"clean", BuildDate:"2021-02-21T20:21:49Z", GoVersion:"go1.15.8", Compiler:"gc", Platform:"darwin/amd64"}

Dont worry if it says "Unable to connect to server" at this stage. We'll be sorting that later.

Before we go ahead and create your cluster its worth exploring the files.

Oh and before we do explore, the files in this directory could have been named whatever we like. For example the outputs.tf file can have been called foo.tf - we just chose to call it that because it contained outputs. So the naming was more of a standard than a requirement.

terraform.tfvars

Think of this as the place where you define the actual values of variables to be used by your terraform configuration.

vpc.tf

This file contains the setup for the Virtual Private Cloud (VPC - remember those from session two) that your GKE will be placed into.

gke.tf

Does the actual job of provisioning a GKE cluster and a separately managed node pool (recommended).

You'll also see that this file contains a variable called gke_num_nodes that is defaulted to 1. This means that one node will be created in each subnet. Totaling 3 nodes for your cluster.

outputs.tf

This file defines the outputs that will be produced by terraform when things have been provisioned.

versions.tf

Configures the terraform providers (in our case the GCP provider) and sets the Terraform version to at least 0.14.

Now you know the files the next step is to update the tfvars file according to your project.

Update the project ID to be that of your Google Project ID and change the region if you would like it to be anything other than the UK.

In the terraform installation video you setup a service account. You'll need to re-use that service account again in order for Terraform to authenticate with your Google cloud account.

First copy the service account to this directory.

Then rename that service account file to be called service-account.json

The reason for the rename is because you'll see it mentioned in the versions.tf file and also in the .gitignore file. Choosing the name service-account.json will ensure you don't accidentally commit it due to us adding it to the gitignore file.

We need to get terraform to pull down the google provider.

In order to do this run:

terraform init

You should see something similar to the below:

Initializing the backend...

Initializing provider plugins...
- Finding hashicorp/google versions matching "3.58.0"...
- Installing hashicorp/google v3.58.0...
- Installed hashicorp/google v3.58.0 (signed by HashiCorp)

Firstly run a plan to see if what Terraform decides will happen.

terraform plan

We can then create your cluster by applying the configuration.

NOTE If you find it complain that an API isn't enabled. Compute Engine API and Kubernetes Engine API are required for terraform apply to work on this configuration. Enable both APIs for your Google Cloud and then continue.

terraform apply

Sit back and relax - it might take 10 mins or so to create your cluster.

Once its finished it'll output something like the info below. Those outputs are defined in the outputs.tf file.

Outputs:

kubernetes_cluster_host = "35.246.22.132"
kubernetes_cluster_name = "devops-upskill-305410-gke"
project_id = "devops-upskill-305410"
region = "europe-west2"

Once its done you'll have your Kubernetes cluster all ready to go!!!

kubectl is used to issue actions on our cluster.

We need to configure kubectl to be able to authenticate with your cluster.

To do this we use the Google Cloud Command Line to get the credentials. Notice how we reference the outputs in the command below:

gcloud container clusters get-credentials $(terraform output -raw kubernetes_cluster_name) --region $(terraform output -raw region)

It should say something like:

Fetching cluster endpoint and auth data.
kubeconfig entry generated for devops-upskill-305410-gke.

You can now verify if kubectl can access your cluster.

Go ahead and see how many nodes are running:

kubectl get nodes

It should show something like:

NAME                                                  STATUS   ROLES    AGE   VERSION
gke-devops-upskill-3-devops-upskill-3-6ab12de4-6p9p   Ready    <none>   18m   v1.18.12-gke.1210
gke-devops-upskill-3-devops-upskill-3-eada08f1-793v   Ready    <none>   18m   v1.18.12-gke.1210
gke-devops-upskill-3-devops-upskill-3-fa0c8ee0-j7qt   Ready    <none>   18m   v1.18.12-gke.1210

Exciting eh!!!

Finally lets get a container running on your cluster.

Firstly check if there are any running pods

kubectl get pods

It should probably say something like this:

No resources found in default namespace.

Lets deploy the nginx deployment.

kubectl apply -f kubernetes/nginx-deployment.yaml

Now lets see the pods

kubectl get pods

And it will show something like this:

NAME                                READY   STATUS    RESTARTS   AGE
nginx-deployment-5cd5cdbcc4-b46m7   1/1     Running   0          14s
nginx-deployment-5cd5cdbcc4-knxss   1/1     Running   0          14s
nginx-deployment-5cd5cdbcc4-sqm2p   1/1     Running   0          14s

Actually you know what, I think 3 replicas is a bit overkill. Lets bring it down to 2 replicas.

Update the nginx-deployment.yaml file and change it down to 2 and save the file.

Then re-run your deployment.

kubectl apply -f kubernetes/nginx-deployment.yaml

Now lets see the pods

kubectl get pods

And you should see something like this:

NAME                                READY   STATUS        RESTARTS   AGE
nginx-deployment-5cd5cdbcc4-b46m7   1/1     Running       0          3m47s
nginx-deployment-5cd5cdbcc4-knxss   1/1     Running       0          3m47s
nginx-deployment-5cd5cdbcc4-sqm2p   0/1     Terminating   0          3m47s

Finally lets get that web server exposed to the internet with a service

We can do this by creating the service (which will sit in front of our pods) and the ingress point

Firstly create the service

kubectl apply -f kubernetes/nginx-service.yaml

And then create the ingress

kubectl apply -f kubernetes/nginx-ingress.yaml

Kubernetes will now create the required load balancer and create an external IP address for your ingress point.

Keep running the command below until you see an external IP address

kubectl get ing

It should output something like this:

NAME                      CLASS    HOSTS   ADDRESS         PORTS   AGE
nginx-webserver-ingress   <none>   *       34.117.49.187   80      95s

After around 5 to 10 mins you should be able to hit the endpoint with your browser. Using the example above I would go to: http://34.117.49.187

NOTE It does take around 5 to 10 mins, for some time you might see a Google 404 page.

Finally we want to destroy our cluster.

Firstly lets remove the service and ingress

kubectl delete -f kubernetes/nginx-ingress.yaml
kubectl delete -f kubernetes/nginx-service.yaml

Then we can destroy the cluster in full.

terraform destroy