Enhance controller to allow a single HTTPbackend ref to have targets across multiple clusters

[liwenwu-amazon]

Use Case

Requirement

Customer want to using one single HTTPbackend ref and dynamically scale out targets across multiple cluster(s)

Why

Customer want to automatically scale-out across multiple cluster(s) and want lattice to load balancing them.

Proposal

Define one backendRef and cluster splits traffic across multiple service names.

• all cluster(s) define same K8S service name e.g. foo
• all cluster(s) define same K8S serviceexport name foo
• HTTPbackend ref just point to serviceimport name foo

apiVersion: gateway.networking.k8s.io/v1alpha2
kind: HTTPRoute
metadata:
  name: foo
spec:
  parentRefs:
  - name: my-hotel
    sectionName: http 
  rules:
  - backendRefs:  
    - name: foo
      kind: serviceimport
      port: 80

each cluster will have

apiVersion: multicluster.x-k8s.io/v1alpha1
kind: ServiceExport
metadata:
  name: foo
  annotations:
          multicluster.x-k8s.io/federation: "amazon-vpc-lattice" 

[zijun726911]

all cluster(s) define same K8S serviceexport name foo

I think currently lattice don't support duplicated tg names? to achieve allow a single HTTPbackend ref to have targets across multiple clusters, we need to change tg name convention to add cluster name in the tg name.

(Or another direction: same name serviceExport from multi clusters use one tg, and register different cluster pod ip targets in the tg)

What's the priority for this enhancement?