This cluster is used daily to carry out my personnal and professional projects (poc, deployment strategies, ci/cd, ...).
- HA Kubernetes cluster on Raspberry Pi
- 6 Raspberry PI 4B for the Kubernetes (K8S) cluster
- 3 RPI 4B 4Go for master nodes
- 3 RPI 4B 8Go for worker nodes
- 1 Raspberry PI 3B+ for the load balancer (LBA) node
- 7 micro SD cards
- 1 => 32G (for LBA node)
- 6 => 128G (for K8S cluster nodes)
- 1 Ethernet Network Switch with 8 ports
- 1 Wireless Portable/Travel router
- 1 USB charger 60W with 8 usb ports
- 1 Raspberry PI cluster case
- 6 USB-C cables for RPI 4B
- 2 USB-A cables (LBA and Wifi router)
- 8 RJ45 cables Cat. 6
There are several methods and tools to install Linux OS on a SD card for RPI. We will use the Raspberry Imager tool to install the Ubuntu Server 20.04.4 LTS 64 on each micro SD card of our RPI cluster.
The default password of the ubuntu user is ubuntu.
ssh ubuntu@10.11.13.21
...
...
...
New password:
Retype new password:
passwd: password updated successfully
Connection to 10.11.13.23 closed.
Set the new password and reconnect with ssh.
In this step, we configure the Ansible inventory file (cluster.yml) to match with our RPI cluster.
all:
vars:
ansible_connection: ssh
ansible_user: ubuntu
ansible_ssh_pass: raspberry
ansible_become: true
ansible_become_user: root
ansible_python_interpreter: /usr/bin/python3
hosts:
rpi-k8s-lba-01:
ansible_host: 10.11.13.20
rpi-k8s-master-01:
ansible_host: 10.11.13.21
rpi-k8s-master-02:
ansible_host: 10.11.13.22
rpi-k8s-master-03:
ansible_host: 10.11.13.23
rpi-k8s-worker-01:
ansible_host: 10.11.13.31
rpi-k8s-worker-02:
ansible_host: 10.11.13.32
rpi-k8s-worker-03:
ansible_host: 10.11.13.33
children:
load_balancer:
hosts:
rpi-k8s-lba-01:
vars:
node: "lba"
# List of backend servers.
backend_servers:
- name: rpi-k8s-master-01
address: 10.11.13.21
- name: rpi-k8s-master-02
address: 10.11.13.22
- name: rpi-k8s-master-03
address: 10.11.13.23
master:
hosts:
rpi-k8s-master-01:
primary_master: true
lba_ip: 10.11.13.20
rpi-k8s-master-02:
rpi-k8s-master-03:
vars:
node: "master"
worker:
hosts:
rpi-k8s-worker-01:
rpi-k8s-worker-02:
rpi-k8s-worker-03:
vars:
node: "worker"
...Let's use the ansible playbook check.yml to check all the cluster hosts.
ansible-playbook -i cluster.yml ansible/playbooks/check.yml
...
rpi-k8s-lba-01 : ok=15 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
rpi-k8s-master-01 : ok=15 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
rpi-k8s-master-02 : ok=15 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
rpi-k8s-master-03 : ok=15 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
rpi-k8s-worker-01 : ok=15 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
rpi-k8s-worker-02 : ok=15 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
rpi-k8s-worker-03 : ok=15 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
All the cluster nodes are OK ;)
We can use these following Ansible playbooks to upgrade OS.
ansible-playbook -i cluster.yml ansible/playbooks/update-host.yml
ansible-playbook -i cluster.yml ansible/playbooks/upgrade.yml
It takes a few minutes to upgrade all the cluster nodes.
Now, let's overclock all the RPI nodes with the ansible playbook overclock.yml.
ansible-playbook -i cluster.yml ansible/playbooks/overclock.yml
ansible-playbook -i cluster.yml ansible/site.yml --tags common
ansible-playbook -i cluster.yml ansible/site.yml --tags lba
After installation, we can take a look at our HAProxy dashboard at http://10.11.13.20:8080/stats
All our backends servers are red because we haven't installed Kubernetes and Traefik yet.
ansible-playbook -i cluster.yml ansible/site.yml --tags kubernetes
Now we can go to take a coffee and come back later (5-10 minutes).
ansible-playbook -i cluster.yml ansible/playbooks/copy-kubernetes-config.yml
This playbook will copy the .kube/config file in our local host under the ~/.kube directory. The file is named cluster-k8s-rpi-config.
kubectl get nodes
NAME STATUS ROLES AGE VERSION
rpi-k8s-master-01 Ready control-plane,master 2d10h v1.23.5
rpi-k8s-master-02 Ready control-plane,master 2d10h v1.23.5
rpi-k8s-master-03 Ready control-plane,master 2d10h v1.23.5
rpi-k8s-worker-01 Ready <none> 2d10h v1.23.5
rpi-k8s-worker-02 Ready <none> 2d10h v1.23.5
rpi-k8s-worker-03 Ready <none> 2d10h v1.23.5
Our Kubenetes cluster is running, let's deploy Traefik included in the addons directory.
kubectl create namespace proxy
kubectl apply -f addons/traefik
Now all our backend servers should be in green state.
Let's put the domain traefik.rpi.lan in the /etc/hosts to view the Traefik dashboard.
...
...
10.11.13.20 traefik.rpi.lan
The Traefik dashboard is now available at http://traefik.rpi.lan
