Hey,
I’m trying to understand the difference between deployment Rancher in a docker image (docker … rancher/rancher deploy cluster with UI…) and the solution with RKE where you deploy kube in cli command.
Is it possible to deploy after a RKE install a docker image and import the rke cluster in the Rancher UI ?
There is no schema to understand the possibilities …
Olivier
I think you’re conflating a couple different things…
RKE is a kubernetes distribution that creates clusters. It has no UI and is not “running” anywhere all the time. It’s a CLI you occasionally manually invoke to apply changes to make the actual state of the nodes/cluster match the desired state described in a config file.
K3s, RKE2, and many other Kubernetes distros work in a more or less similar manner.
Rancher is a platform that creates/imports/manages clusters and the stuff in them. This is the UI.
You can import any existing k8s clusters (such as above) into a running Rancher installation and manage the stuff in the cluster through Rancher.
You can use Rancher to create new clusters, which works by talking to a provider to create VMs (or you register your own for custom) and then basically runs RKE internally for you to assemble them into a cluster.
Separately there is the different ways you can install Rancher.
One way is a standalone docker container (docker run … rancher/rancher). The other main way is as a set of k8s resources deployed into an already-running k8s cluster.
The “helm install” way is the primary way to create a production installation. You create a cluster of some live someone (could be RKE, or a hosted EKS cluster, etc) and then use Helm to install Rancher into that. That cluster (should) have multiple workers to handle failures, backups of etcd, etc to make this reliable. You generally manage this cluster outside of Rancher to avoid chicken-egg scenarios, but it does show up as a cluster called “local” and you can manage the stuff in it and to in some cases the cluster itself, depending on what created it.
The “docker run” version is meant mostly for experimenting/demos; there is no HA if it fails, etc. What it is actually doing internally is running a k3s cluster inside there and running Ranchrr on top of that, because Rancher only actually runs as a k8s workload. It uses CRDs in that cluster as the “database” backing everything.
(There’s also a “rancherd” installation but I’m ignoring that for now)
Ok, thanks for the answer, it’s more clear.
I can now understand the Rancher Docs: Install/Upgrade Rancher on a Kubernetes Cluster
Olivier