I'm currently learning about Kubernetes. While the abstractions are great I'd like to understand what is actually happening under the hood with Ingress set up.

In a cloud context and using nginx ingress controller, following an external request from the load balancer to a pod, this is what I believe is happening:

1. The request arrives at the load balancer and, using some balancing algorithm, it selects an ingress controller. Many instances of the ingress controller will likely be running in a resilient production environment.

2. The ingress controller (nginx in this case) uses the rules exposed by the ingress service to select a node port, selecting a specific node to route to. Is there any load balancing happening by nginx here?

3. The kubelet on the node receives this request. Depending on the set up (iptables vs ipvs) this request will be further load balanced and using the clusterip a specific pod will be selected to route to. Can this pod could exist anywhere on the cluster, on a different node to the kubelet routing it?

4. The request is then forwarded to a specific pod and container.

Is this understanding correct?