Skip to content

Scripts Reference

KubeAuto uses five Bash scripts for cluster provisioning. All scripts use set -euo pipefail for strict error handling.

Script Execution Order

Order Script Runs On Phase
1 01-common.sh All nodes 1
2 02-kubernetes.sh All nodes 2
3 03-controlplane.sh Control plane 3
3 04-worker.sh Workers 3
— 05-test.sh Control plane Manual

01-common.sh — OS Preparation

Runs on: All nodes (Phase 1)

Purpose: Prepares the operating system for Kubernetes installation.

Actions:

  1. Disables swap permanently — swapoff -a and comments out the swap entry in /etc/fstab
  2. Updates APT package index
  3. Installs prerequisite packages: apt-transport-https, ca-certificates, curl, gnupg, lsb-release
  4. Loads kernel modules: overlay and br_netfilter (persisted to /etc/modules-load.d/k8s.conf)
  5. Applies sysctl settings: enables bridge-nf-call-iptables, bridge-nf-call-ip6tables, net.ipv4.ip_forward (persisted to /etc/sysctl.d/k8s.conf)
  6. Persists PRIMARY_IP to /etc/environment

02-kubernetes.sh — Kubernetes Installation

Runs on: All nodes (Phase 2)

Purpose: Installs the container runtime and Kubernetes tooling.

Actions:

  1. Installs containerd via APT
  2. Configures containerd — generates default config and patches SystemdCgroup = true
  3. Detects Kubernetes version — fetches the latest stable minor version from https://dl.k8s.io/release/stable.txt
  4. Adds Kubernetes APT repo with GPG keyring for the detected version
  5. Installs kubelet, kubeadm, kubectl and holds them at the installed version with apt-mark hold
  6. Configures crictl to use the containerd socket
  7. Sets KUBELET_EXTRA_ARGS in /etc/default/kubelet to bind kubelet to PRIMARY_IP
  8. Enables and reloads kubelet via systemd

03-controlplane.sh — Cluster Initialisation

Runs on: Control plane only (Phase 3)

Purpose: Initialises the Kubernetes cluster and deploys the CNI.

Actions:

  1. Runs kubeadm init with:
    • --pod-network-cidr=$POD_CIDR
    • --service-cidr=$SERVICE_CIDR
    • --apiserver-advertise-address=$PRIMARY_IP
  2. Sets up kubeconfig for both vagrant user and root
  3. Installs Calico CRDs from calico/operator-crds.yaml
  4. Deploys Tigera operator from calico/tigera-operator.yaml
  5. Polls Tigera operator for up to 3 minutes until the pod is Running
  6. Downloads and patches calico/custom-resources.yaml (replaces default CIDR with POD_CIDR)
  7. Generates join command and writes to /vagrant/join-command.sh
  8. Polls kube-system until at least 6 pods are Running
  9. Prints summary of nodes, pods, and Tigera status

04-worker.sh — Worker Join

Runs on: Worker nodes only (Phase 3)

Purpose: Joins the worker node to the existing cluster.

Actions:

  1. Polls for /vagrant/join-command.sh — checks every 5 seconds for up to 5 minutes (60 attempts)
  2. Executes the join command using bash /vagrant/join-command.sh
  3. Exits with error if the join command is not found within the timeout

05-test.sh — Test Deployment

Runs on: Control plane only (manually)

Purpose: Validates the cluster by deploying a sample workload.

Actions:

  1. Deploys Apache httpd (alpine) as a 2-replica Deployment
  2. Waits up to 2 minutes for the Deployment to reach Available
  3. Exposes as NodePort service on port 80
  4. Queries worker IPs dynamically from the cluster
  5. Prints access URLs for every worker node using the assigned NodePort

Invocation:

vagrant ssh controlplane
sudo bash /vagrant/scripts/05-test.sh