I am learning k8n and want to clarify that I am understanding the traffic flow correctly.
I setup Rancher on domain
a.com (single node for now) and create a new Cluster (RKE) via it (master, 2 workers) on bare metal. I can download the KubeConfig and use
kubectl etc to connect to the new cluster.
When I now setup an Ingress for domain
b.com, is the idea that traffic first goes through the rancher server and then flows to the new cluster? Or are clients directly communicating with the new cluster?
In other words:
To who's IP should the DNS for
b.compoint? Am I correct that this should not be the rancher server and neither the master of the new cluster? From my current knowledge I would guess it should be pointed to a L7 Load Balancer that itself then routes to the worker nodes of the new cluster that expose the desired service. And further that L7 LB is normally provided by the Cloud Provider. Now on bare metal I am confused - there is no L7 LB out of the box which is why people started MetalLB? And since this is beta, I could use traefik as Ingress/L7 LB for this (but not its SSL management when I want HA, because traefik is stateless and I would need a cert-manager dealing with this) - correct? But anyways what is the IP that I should point
b.comto in order to reach the Ingress?
If the Rancher server goes down, I understood that I can no longer connect via
kubectlto the new server, since this traffic does flow through Rancher, correct? But clients should still be able to connect to services running in the new cluster as long as only the rancher server goes down - is this correct?