Each node runs the kubelet, a systemd service that starts Pods and reports the node's status. If it stops reporting, the control plane marks the node NotReady (Ready Unknown in kubectl describe node, "Kubelet stopped posting node status"), taints it and, after a grace period, moves its Pods away.
The fix is on the node itself. Connect with ssh, then check the service: systemctl status kubelet says whether it is running, failed or disabled, and journalctl -u kubelet says why it failed, often a missing or broken file such as /var/lib/kubelet/config.yaml, an expired certificate or a stopped container runtime. Service changes need root: sudo systemctl restart kubelet, or sudo systemctl enable --now kubelet when it was disabled and should start at boot.
Always confirm which node is NotReady with kubectl get nodes before you touch a machine.
From NotReady to the kubelet's own words.