The changes that we have been doing to /etc/defaults/kubelet are reset on node reboots, as is apparently the whole /etc directory — which also means that /etc/cni/net.d/05-cilium.conf is removed.

This would not be a problem if the assumption we made that the node taint we recommend placing on the nodes would come back upon reboots held true, but in practice it doesn't.

Besides this, it seems that containerd will re-instante its CNI configuration file, and it will do so way before Cilium has had the chance to re-run on the node and re-create its CNI configuration, causing pods to be assigned IPs by the default CNI rather than by Cilium in the meantime.

This commit attempts at preventing that from happening by observing that /home/kubernetes/bin/kubelet (i.e. the actual kubelet binary) is kept between reboots and executed concurrently with containerd by systemd. We leverage on this empirical observation to replace this file kubelet with a wrapper script that, under the required conditions, disables containerd, patches its configuration, removes undesired CNI configuration files, re-enables containerd and becomes the kubelet.

Additionally, to prevent situations in which the GKE node is forcibly stopped and re-created from causing unmanaged pods, and building on the observation that the node comes back with the same name and pods are already scheduled there, we change the recommended taint effect from NoSchedule to NoExecute, to cause any previously scheduled pods to be evicted, preventing them from getting IPs assigned by the default CNI.This should not impact other environments due to the nature of NoExecute, so we recommend it everywhere.

Prevent unmanaged pods in GKE's containerd flavors.
*Important:* Users should update their node taints from `node.cilium.io/agent-not-ready=true:NoSchedule` to `node.cilium.io/agent-not-ready=true:NoExecute`.
*Important:* During the first node reboot after the fix is applied pods may still get IPs from the default CNI as cilium-node-init is only run later in the node startup process. The fix will then be in place for all subsequent reboots.