Hi, Sorry for maybe a bit odd question, but I am a beginner and would like to get some clear info (read tons of articles and still can't understand fully). We have VCF9 infrastructure and now we would like to give our teams/devs ability to run microservices on Kubernetes. Thinking of installing Rancher. But here is the main question. Should we install RKE2 cluster with 3 master nodes and 3 worker nodes. And place Rancher on top manage this cluster. Or install Rancher on RKE2 cluster (leave cluster just for Rancher to run) and via Rancher deploy a new cluster in VCF9 for application to run and be deployed?

We have set up Rancher on K3s, three nodes, with a separate MariaDB cluster to act as etcd. From that, we build RKE2 clusters (via Terraform, where we can). The servers for those clusters can be built separately and the Registration command used or can be part of the Rancher Create Cluster configuation (for some cloud provider settings). Three control plane/etcd servers are advised as a minimum, three workers which can then be added to as needed. But I'm just a beginner as well.

So it is better for Rancher management to have dedicated cluster of 3 workers and for teams to run their stuff create a separate dedicated clusters? With their own master and worker structure? If the load will not be big in the begging aren't this an overkill for having hardware wasted on masters? I mean. Can't we just run ups on same cluster workers on which Rancher is installed?

Rancher doesn't have workers, just the control plane and the etcd. At least, that's how it is for me. When looking at the three nodes, they're all control plane. If you want something small, there's the Rancher docker installation, just the one server covers all requirements. The RKE2 clusters, it all depends on how you want to run them. You can have one server with control plane, etcd and worker on the same thing. If you're talking of anything where you want resilience, though, it would be encouraged that you include a few more.

So when you are creating RKE2 cluster for Rancher it is not mean to be for production running workloads and just for Rancher and some internal components. And for production workloads you create to totally dedicated cluster via Rancher management dashboard is that correct?

Rancher is your tool for creating and managing RKE2/K3s clusters. For such, it is fully production-grade. The servers for those RKE2/K3s clusters are completely separate and can be set up as having all of control plane, etcd and worker roles on each server or splitting that up so the roles are on different servers. To just try things out, have one server where you install the docker Rancher, make up a cluster, get the registration command from that and run it on a different server (with network access to the first) with all the roles on it. I'd not advise it for a production load, but it should give you a fine idea of what can be done.

Thanks will try.