The tear down section of the official documentation says:

To undo what kubeadm did, you should first drain the node and make sure that the node is empty before shutting it down.

Talking to the master with the appropriate credentials, run:

kubectl drain <node name> --delete-local-data --force --ignore-daemonsets 
kubectl delete node <node name>

Then, on the node being removed, reset all kubeadm installed state:

kubeadm reset