QuestionQ82

VerifiedConfiguring and implementing hybrid and multi-cloud network interconnectivity

Your organization has run Google Kubernetes Engine (GKE) clusters in several distinct Virtual Private Clouds (VPCs) for the last year, with one GKE cluster in each VPC. Although the GKE nodes use unique IP address space for every cluster, Pod IP ranges have been reused among the VPCs. The VPCs are spokes in a Network Connectivity Center hub that uses mesh topology. Pod IP ranges have been excluded from the Network Connectivity Center VPC spokes.

Although this setup appears to work and Compute Engine instances in the VPCs can access services, Pods cannot access services in other VPCs. You need to enable communication between Pods in separate VPCs. What should you do?

Explanation

When Pod IP ranges are reused across clusters, they cannot be uniquely routed between VPCs. GKE’s IP masquerade agent can apply source NAT to Pod traffic leaving a cluster, replacing the overlapping Pod source address with the cluster’s unique node IP address. Return traffic can then be routed correctly while the Pod CIDRs remain excluded from Network Connectivity Center advertisements.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!