Deploying highly available K8s cluster using RKE

RKE is a useful tool that help with configuring K8s clusters. In this use case we will deploy a highly available K8s cluster using RKE. We will also use Terraform in order to deploy the required compute resources into AWS.

The following infra diagram describes the resources that will be deployed

Services that will be installed on the ndoes

• Docker
• RKE
• kubectl
• Helm

Use Terraform to deploy the infrastructure

1. Install Terraform: How to install Terraform

2. Edit the Terraform provider information in order to deploy to AWS. You can use access keys, or you could configure the AWS CLI. Another option is to use an IAM role.

3. Run the following to initiate the terraform deployment

terraform init

terraform apply

Post install

1. SSH into the master node using its public IP

2. Edit the rancher-cluster.yml and update the IP addresses of the nodes. Also, update the path to your SSH private key (How to create SSH key pairs) for authentication between the nodes.

nodes:
  - address: <Node01 IP Address>
    user: ubuntu
    role: [controlplane,etcd,worker]
    ssh_key_path: /path/to/privatekey.pem
  - address: <Node02 IP Address>
    user: ubuntu
    role: [controlplane,etcd,worker]
    ssh_key_path: /path/to/privatekey.pemm
  - address: <Node03 IP Address>
    user: ubuntu
    role: [controlplane,etcd,worker]
    ssh_key_path: /path/to/privatekey.pem
addon_job_timeout: 120

3. Run the up command to initiate the creation of the cluster rke up config rancher-cluster.yml

You will then see the following output. This output indecates that the setup has started and that the tunnels to the hosts have been created. Using the SSH key pair fro the previous step.

After a few moments you will get a success output:

4. Create a symbolic link and export path to the kubeconfig

mkdir -p /home/ubuntu/.kube
ln -s /home/ubuntu/kube_config_rancher-cluster.yml /home/ubuntu/.kube/config

export KUBECONFIG=/home/ubuntu/kube_config_rancher-cluster.yml

5. Check that the cluster is running properly

kubectl get nodes

This should give you the running nodes in the cluster