kubernetes-install-cleanup — README

This repository contains a helper script to remove previous Kubernetes/container runtime state and install Kubernetes components on a node.

Purpose

• Clean old kube/k8s/container runtime state (packages, data directories, iptables rules).
• Install containerd and kubeadm/kubelet/kubectl.
• Initialize a control-plane (master) or join a worker node (using kubeadm join).

Prerequisites

• Ubuntu 18.04 / 20.04 / 22.04 (tested).
• SSH access or direct console to each node.
• Internet access for apt package downloads.
• Ensure swap is disabled on all nodes (the script will also attempt to disable it).
• Edit host and cluster config files in the config directory if needed:
  • config/hosts.txt
  • config/k8s-config.yaml
• Optional: run the prerequisites script first: scripts/prerequisites.sh

Usage

1. Make the script executable: sudo chmod +x install_and_cleanup.sh

2. Initialize the first control-plane (run on the node you choose as first master): sudo ./install_and_cleanup.sh --role master [--k8s-version 1.27.0] [--pod-cidr 192.168.0.0/16]

   • The script runs kubeadm init and applies a Flannel CNI manifest by default.
   • It prints a worker join command (kubeadm join ...) — copy this for worker joins.
   • For additional control-plane nodes (HA), follow the kubeadm output and run the control-plane join command (includes --control-plane and --certificate-key).

3. Join worker nodes (run on each worker or via SSH): sudo ./install_and_cleanup.sh --role worker --join-cmd "kubeadm join API:PORT --token --discovery-token-ca-cert-hash sha256:"

Options

• --role master|worker (required)
• --k8s-version X.Y.Z (default in script; installs kubeadm/kubelet/kubectl of this version)
• --pod-cidr CIDR (defaults in script; used during kubeadm init)
• --join-cmd "..." (required for role=worker)

Post-install verification

• On the control-plane (after exporting kubeconfig or using $HOME/.kube/config): kubectl get nodes kubectl get pods -A

Notes / Caveats

• The script is destructive: it removes /etc/kubernetes, /var/lib/etcd and other runtime data. Back up anything you need before running.
• For production HA with multiple control-planes, use a load-balancer in front of API servers or follow kubeadm HA docs for stacked control-planes.
• The script applies Flannel by default. Replace the CNI manifest in the script if you prefer Calico/Weave/etc.
• Tokens expire; if a token expires, generate a new worker join command on the master: sudo kubeadm token create --print-join-command

Support

• Repo layout / scripts are in: rv-kubernetes/kubernetes-install and rv-kubernetes/multinode-kubernetes
• Contributions are welcome for automated multi-master join or a remote-run wrapper (SSH/parallel).