Using Rancher, K3S, and Longhorn for single node Kubernetes deployment

Prerequisites: This process assumes the use of a clean Ubuntu 20.04 server with a public IP address and an authorized key recognised by the server in the current users ~/.ssh.

Rancher Deployment Process

Get the docker-ansible project to run an ansible controller in a docker container

user@dev$ git clone https://github.com/kartoza/docker-ansible.git

Build the container image

user@dev$ cd docker-ansible

user@dev$ docker build . -t ansible

Replace the remote IP address in the inventory file

user@dev$ echo "<server-ip> ansible_user=root ansible_private_key_file=/root/.ssh/id_ed25519" > ${PWD}/ansible/inventory/hosts.ini

Remove the default playbooks directory

user@dev$ rm -r ./ansible/playbooks

Fetch the playbook repository which includes the singlenode rancher deployment

user@dev$ cd ./ansible && git clone https://github.com/kartoza/playbooks.git

Go back to the repository root

user@dev$ cd /..

Run the docker container (Powershell/ Windows users should read the docker-ansible readme)

user@dev$ docker run -dt -v $PWD/ansible:/ansible \
               -v ~/.ssh:/root/win-ssh:ro \
               -e ANSIBLE_CONFIG=/ansible/ansible.cfg \
               --restart=unless-stopped \
               --name ansible ansible

Once the container is running run the playbook

user@dev$ docker exec -it ansible ansible-playbook /ansible/playbooks/ubuntu20.04/app/rancher-singlenode.yaml

Save the logs and kill off the ansible container

user@dev$ docker logs ansible >> $PWD/ansible_log.txt && docker rm -f ansible

At this point, the server should be configured to use https://rancher.<server-ip>.sslip.io. By default the admin user is admin and the password is chickensoup but that can be changed in the rancher-singlenode.yaml file. Note self-signed TLS is enabled so there'll be browser notices, but TLS configuration is environment specific and not considered in the scope of this process definition.

Application Deployment Process

Now the project workload will need to be deployed to the server. This will use kubectl and can be done in various ways. If you wish to deploy from the local machine, the kubectl configuration for each node should be available in the ansible/data directory.

For the purpose of demonstrating this process, it is assumed that this is done by using ssh to connect to the server as the root user, which should have the kubectl configuration already setup by the ansible deployment.

user@dev$ ssh root@<server-ip>

Clone the project repository

root@app-srv$ mkdir /manifests && cd /manifests && git clone https://gitlab.com/zacharlie/test-manifest.git

Replace the hostname field in the ingress with the fully qualified domain name of the server.

root@app-srv$ cp /manifests/test-manifest/ingress.yaml /manifests/test-manifest/ingress.yaml.bak && awk '/nginx.test.localhost/ {sub("nginx.test.localhost", "123.456.7.89.sslip.io", $0)} {print}' /manifests/test-manifest/ingress.yaml.bak > /manifests/test-manifest/ingress.yaml

Apply the project resources using kustomize

root@app-srv$ cd /manifests/test-manifest && kubectl apply -k .

Be patient and monitor the resources from the nifty rancher dashboard (Workload >> Pods is a good place to start). Once everything is running you may view the landing page at the hostname address that was configured within the ingress configuration.