RKE self-hosted k8s?

Does RKE setup k8s as a “self-hosted” cluster?

Meaning, services for the cluster are run within the cluster (like etc.d).

I know that is one way Bootkube / Kubeadm are evolving too, as you end up using the same tools to manage the cluster, as you do any service in the cluster. Rolling cluster upgrades, downgrades, etc. It looks really interesting.

From the RKE announcement page:

“…RKE packages all system services as containers…” It seems like it is setup as self-hosted, but I’d like to verify that.

Yes everything is running inside the K8s cluster setup by RKE and everything is running inside a container. And I think that is how most K8s setup/management tools do it now days, definitely the way Kops does it.

Does RKE, or Rancher 2.x, have a backup/restore function? For example, if your controller plane / etc.d nodes are all down?

The specific case I am thinking of is power outages in a home lab. Gracefully shutdown the cluster. Then shutdown the nodes. Then turn back on whenever power is restored.

It’s one of the features I’ve seen in alpha in Kubeadm / Bootkube working towards beta in k8s 1.10 or 1.11.

It could also be a useful feature for other situations. For example, saving money from cloud providers for test clusters you only use on the weekends (or week days).