Question

A retail company is deploying a new GKE cluster into an existing custom-mode Shared VPC subnet. The platform team must plan IP addressing so the cluster can scale to 500 nodes, each supporting up to 110 Pods, plus roughly 2,000 Services. The subnet's primary range (10.10.0.0/24) is already used by on-prem-routed VM workloads and cannot be resized. The networking lead insists Pod and Service IP allocation must not consume addresses from that primary range. Which approach correctly satisfies these requirements?

Accepted Answer

Create a VPC-native (alias IP) cluster and assign two secondary IP ranges on the subnet — one sized for Pods and one for Services — separate from the primary range.

Answer

Create a route-based cluster so Pod traffic uses custom static routes and does not require secondary ranges on the subnet.

Answer

Expand the subnet's primary range to a /16 and let the cluster carve Pod and Service IPs from that primary range.

Answer

Create a VPC-native cluster but place both Pods and Services into the subnet's existing primary range using alias IPs to conserve address space.