GKE Networking Fundamentals
What You'll Learn
You'll understand how GKE clusters handle IP addressing through VPC-native networking, IP aliasing, and distinct CIDR ranges for pods and services.
The Core Concept
GKE uses VPC-native clusters as the modern networking standard. Unlike legacy route-based clusters, VPC-native clusters assign IP addresses directly from your VPC's subnet ranges using a feature called IP aliasing.
Think of it like organizing a neighborhood: your main street (VPC subnet) has house numbers (node IPs), but each house also has apartment numbers (pod IPs) and mailbox numbers (service IPs). All three come from pre-planned address blocks that don't conflict.
How It Works
When creating a GKE cluster, you define three separate CIDR ranges:
- Primary subnet range: IP addresses for nodes themselves
- Pod CIDR (secondary range): IP addresses allocated to pods across all nodes
- Service CIDR (secondary range): Virtual IP addresses for Kubernetes Services
GKE automatically handles IP allocation from these ranges. Pods receive IPs directly routable within your VPC without extra NAT or routing tables. This means pods can communicate directly with other GCP resources like Cloud SQL or Compute Engine VMs using their pod IPs.
Why VPC-Native Matters
VPC-native clusters enable better security (firewall rules work on pod IPs), simpler troubleshooting (no hidden routing layers), and native GCP integration (Private Google Access, VPC peering, Shared VPC).
The alternative—route-based clusters—require complex routing tables and aren't recommended for new deployments.
IP Planning Considerations
Size your CIDR ranges carefully. A /14 pod CIDR supports roughly 250,000 pods, while /20 supports about 4,000. Service CIDR is typically smaller (/20 is common) since you have fewer services than pods.
Key Takeaway: GKE VPC-native networking uses IP aliasing with separate CIDR ranges for nodes, pods, and services, enabling direct VPC routing and seamless integration with Google Cloud services.