Component: docs, kubernetes, etcd, systemd, containers, ...

Why this is needed:

Recently, a failed expansion in production led to a very broken cluster, and wiping and reinstalling new machines was out of the question, so we needed a manual clean-up procedure.
Such a procedure doesn't exist in our documentation today: that would have saved both developers and support teams much time to have it somewhere.

What should be done:

Describe procedures for:

• rolling back a failed expansion on a node (remove manifests, certificates, disable services, reboot...)
• resetting a cluster back to bootstrap-stage
• removing a failed etcd member
• troubleshoot Unauthorized in kubelet journal (and more examples of logs when something is broken)

Implementation proposal (strongly recommended):

Write all this in a Troubleshooting guide, reference it throughout Installation and Operation guides.