Multicluster networking connects services and workloads across multiple Kubernetes clusters, whether those clusters share one infrastructure environment or span public cloud, on-premises, and multiple clouds. It forms the cluster-to-cluster layer within a broader Anthos networking architecture. section 10.0
Choose the connectivity model according to cluster placement:
- Use a flat network with a single or shared VPC for VPC-native GKE clusters on GCP; Pod-to-Pod connectivity can work across regions without additional configuration. section 10.5
- Configure additional networking for multiple-VPC GCP architectures, using special gateways or an Ingress model to connect Pods and Services. section 10.5
- Use gateways or Ingress for non-GCP clusters, whose cluster networks and address ranges are isolated by default. section 10.5
- Deploy Anthos Service Mesh east-west gateways across participating clusters when multicluster service discovery and cross-cluster communication are required. section 10.5
Use the four-layer networking diagram to distinguish cloud or hybrid connectivity, cluster networking, multicluster networking, and service or client connectivity. section 10.0
