Skip to content

Kubernetes Reset (kubeadm)

This section describes how to cleanly reset a Kubernetes cluster installed with kubeadm.

A reset may be required in the following situations: - failed or incomplete cluster initialization - incorrect Pod CIDR or network configuration - CNI installation issues - certificate or kubeadm configuration errors - reinstallation from scratch

The goal is to return the system to a clean, pre-Kubernetes state without reinstalling the operating system.

⚠️ Important warning

This procedure:

  • removes all Kubernetes state
  • deletes etcd data
  • removes CNI configuration
  • does not preserve workloads or cluster data

Use it only when you intend to rebuild the cluster.

Reset the control plane and worker nodes

The following steps must be executed on all nodes
(control plane and workers).

Run kubeadm reset

sudo kubeadm reset --force

This command:

  • stops Kubernetes services
  • removes static pod manifests
  • deletes etcd data (on control plane)
  • resets kubeadm state

Remove Kubernetes configuration files

rm -rf $HOME/.kube
sudo rm -rf /etc/kubernetes

Cleanup CNI configuration

Remove CNI configuration and state:

sudo rm -rf /etc/cni/net.d
sudo rm -rf /var/lib/cni

If Calico was used, also remove its state:

sudo rm -rf /var/lib/calico

Kubernetes and CNI plugins modify iptables rules. Reset them to avoid conflicts.

sudo iptables -F
sudo iptables -t nat -F
sudo iptables -t mangle -F
sudo iptables -X

Restart container runtime and kubelet

sudo systemctl restart containerd
sudo systemctl restart kubelet

Verify system state

Ensure Kubernetes is no longer active:

kubectl get nodes

Expected:

  • command fails or reports no configuration

Check that no Kubernetes ports are listening:

sudo ss -lntp | grep 6443

Expected:

  • no output

Optional: reboot the node

A reboot ensures:

  • clean network state
  • cleared virtual interfaces
  • clean kernel routing tables
sudo reboot