Currently I cannot find any docs about dependency-watchdog. Currently it seems to be:

1. probing the kube-apiserver and scaling down the kube-controller-manager to 0 replicas when the kube-apiserver is reachable internally but unreachable externally
2. restarting control plane components in CrashloopBackoff once etcd is again available

Expected Behaviour

After provisioning and coming back later, an inlets exit-node will be known mainly by its IP rather than the internal ID used by the cloud provider.

We co

What would you like to be added:
Avoid draining of nodes in NotReady state.

Why is this needed:
Draining of nodes in a NotReady state sometimes ends up waiting for up to drain timeout as the Kubelet might be stuck. So in cases where the node isn't ready, a force delete might be preferred.

Otherwise developer wouldn't understand why job isn't starting. This info should be in Environment section of the Dashboard & displayed as warning icon in the top menu.

At the moment, the godoc for Packngo doesn't read very well:
https://godoc.org/github.com/packethost/packngo

This would need improving the comments in the code, and add some summary, like in
https://github.com/gorilla/websocket/blob/master/doc.go

And then some examples
https://blog.golang.org/examples
https://github.com/gorilla/websocket/blob/master/example_test.go

Operations like create() and terminate() block execution until completed. There should be a way of polling for state changes as we do with EC2. Not sure if there is a precedent for this in libcloud or if we would have to abandon libcloud.

While yes technically alumni are upperclassmen, an alumni signature on packet is not the same as an active member's signature. I am proposing the that /upperclassmen endpoint list alumni separately from active members.

Summary

Document upstream firmware upgrade procedure

Why do we need this ?

In order to have a clear documentation on how