Quickly Configure Kubernetes Using Ansible

5-minute read
Mukkund
ansible • provisioning • infrastructure • cicd

Learn how to automate cluster provisioning using Ansible Link to heading

1.png

What is Ansible? Link to heading

Ansible is an infrastructure automation tool that allows you to do the following: 

  • Create consistent environments for VMs or servers.
  • Provision servers in a stateless manner.
  • Enable version control in infrastructure management.

Ansible makes for a great tool to automate manual and error-prone jobs. One such example is creating and configuring a Kubernetes cluster. 

While configuring a Kubernetes cluster, one has to repeat the process over several nodes which when done manually might lead to inconsistencies. Hence, in this guide, I use Ansible to install and deploy Rancher K3S on my Rasberry Pi cluster.

What will we be doing in this guide? Link to heading

2.png

  • Prepare your server or VM to be managed by Ansible.
  • Install required packages.
  • Install K3S from Rancher.
  • Configure the master and worker nodes in your cluster.

Note: This guide assumes that you are familiar with the basic concepts of Ansible. If not, then here is a quick refresher.

Pre-requisite steps Link to heading

For the controller to access the nodes, we need to enable passwordless, SSH access to them. Therefore, go ahead and add your SSHs keys in the authorized_keys file of the rootuser in your remote host.  

Prepare your inventory Link to heading

The inventory file contains the list of hosts you wish to manage via the worker.

[master]
micro-host-1 ansible-host=xxx.xxx.xxx.xxx

[workers]
micro-host-2 ansible-host=xxx.xxx.xxx.xxx

[microcluster:children]
master
workers

[microcluster:vars]
ansible_user=root


[nodes:children]
master
workers

[nodes:vars]
ansible_user=worker

In my inventory file, I have 2 hosts that I divide into master and worker groups. I have 2 parent groups called microcluster and nodes for which I use different users. The reason will be apparent in the next step.

Create and configure users Link to heading

It is recommended to create a user with sudo rights when provisioning your node. Therefore, we run the first set of tasks using the root user (or using the microcluster group of hosts defined in the inventory file).

---
- name: Make sure the group is enabled
  group:
    name: "{{ user_group }}"
    state: present

- name: Allow group to have passwordless sudo
  lineinfile:
    dest: /etc/sudoers
    state: present
    regexp: "^%{{ user_group }}"
    line: "%{{ user_group }} ALL=(ALL) NOPASSWD: ALL"
    validate: "visudo -cf %s"

- name: Add sudoers users to "{{ user_group }}" group
  user:
    name: "{{ user_name }}"
    groups: "{{ user_group }}"
    append: yes
    state: present
    createhome: yes
    shell: /bin/bash

- name: Set up authorized keys for the user
  authorized_key: user="{{ user_name }}" key="{{item}}"
  with_file:
    - ~/.ssh/id_rsa.pub

The tasks are described as follows:

  • Create a group to which the desired user belongs. In my case, I call it cluster.
  • Provide the group with passwordless sudo access.
  • Create and add users to the newly-formed group.
  • Copy the SSH keys of the controller into the remote machine.

Install updates and required packages Link to heading

The following tasks are self-evident; the packages are updated and new ones are installed. The difference is that these steps are now run using the newly-created user instead of using the root user (see the node group in the inventory file).

---
- name: Install updates
  apt:
    update_cache: yes

# Installing standard apps
- name: Installing standard development toolchain
  tags: dev-tools
  apt:
    name: "{{item}}"
    state: present
    update_cache: yes
  loop:
    - npm
    - vim

Installing K3S on the master Link to heading

We need to follow a slightly different procedure when installing K3S on the master and the worker nodes. Let’s first install it on the master node.


---
- name: Install K3s on the master node
  become: true
  shell: "curl -sfL https://get.k3s.io | K3S_KUBECONFIG_MODE='644' sh -s"
  when: inventory_hostname in groups['master']
  tags: k3s-master

- name: Verifying the installation of k3s on workers
  include_tasks: verify_k3s.yml
  when: inventory_hostname in groups['master']
  tags: k3s-master

Configuring workers Link to heading

---
- name: Fetching the master node token
  become: true
  ansible.builtin.fetch:
    src: "{{ NODE_TOKEN_DIR }}"
    dest: "{{ TMP_NODE_TOKEN_DIR }}/{{ NODE_TOKEN_FILENAME }}"
    flat: yes
  when: inventory_hostname in groups['master']
  tags: k3s-workers

- name: Ensures tmp dir exists dir exists
  file: 
    path: "{{ TMP_NODE_TOKEN_DIR }}"
    state: directory
  when: inventory_hostname in groups['workers']
  tags: k3s-workers

- name: Copy the node token file to workers
  ansible.builtin.copy:
    src: "{{ TMP_NODE_TOKEN_DIR }}/{{ NODE_TOKEN_FILENAME }}"
    dest: "{{ TMP_NODE_TOKEN_DIR }}/{{ NODE_TOKEN_FILENAME }}"
  when: inventory_hostname in groups['workers']
  tags: k3s-workers

- name: Installing and configuring K3S in worker nodes
  shell: "curl -sfL https://get.k3s.io | K3S_TOKEN={{ lookup('ansible.builtin.file', '{{ TMP_NODE_TOKEN_DIR }}/{{ NODE_TOKEN_FILENAME }}') }} K3S_URL='https://{{ ansible_default_ipv4.address }}:6443' K3S_NODE_NAME='{{ inventory_hostname }}' sh -"
  when: inventory_hostname in groups['workers']
  tags: k3s-workers

- name: Verifying the installation of k3s on workers
  include_tasks: verify_k3s.yml
  when: inventory_hostname in groups['workers']
  tags: k3s-workers
  • Before we install K3S on the worker nodes, we need to fetch the installation token from the master.
  • The first 2 steps involve copying the token from the master to the controller.
  • In the 3rd step, we copy the token from the controller to the worker nodes.
  • We then use the token to install and configure K3S on the hosts in the worker group.

Verifying installation of K3s Link to heading

The following tasks are carried out to verify if K3S has been properly installed on all of the managed nodes.

---
- name: Getting installation directory of {{K3S_BINARY}} 
  shell: "which {{K3S_BINARY}}"
  register: kubeout
  failed_when: "kubeout.stdout == ''"
  tags: 
    - k3s-master
    - k3s-workers

- name: "{{K3S_BINARY}} installation status"
  ansible.builtin.debug:
    msg: "Kubectl installed in {{ kubeout.stdout }}"
  tags:
      - k3s-master
      - k3s-workers

And….voilà! Link to heading

And just like that, you have configured Kubernetes for your cluster by running just two Ansible playbooks.

Want to Connect? Link to heading

Thank you for reading my article. You can also find me on LinkedIn and my work on GitHub.