Skip to main content
A private Google Kubernetes Engine (GKE) cluster runs entirely on internal IP addresses, offering robust security and network isolation. In a private cluster, nodes, pods, and services communicate only over your VPC’s internal network. You can provision private clusters in either Standard mode or Autopilot mode.
The image is an infographic explaining a private GKE (Google Kubernetes Engine) cluster, highlighting components like nodes, pods, services, and features such as Autopilot and Standard modes. It emphasizes the use of internal IPs and private clusters.

​
Default Network Isolation

By default, private clusters block all inbound internet traffic to nodes and pods. This isolation ensures your workloads remain hidden from the public internet.
The image is a diagram explaining a Private GKE Cluster, highlighting nodes, pods, and services with no direct internet access.

​
Internal IP Addressing

  • Node IPs are allocated from the primary IP range of your chosen VPC subnet.
  • Pod & Service IPs use secondary IP ranges within the same subnet.
These unique internal addresses guarantee secure, high-performance communication between cluster components. Nodes are never exposed externally.
The image is an illustration explaining a Private GKE Cluster, showing components like nodes, pods, and services with internal and secondary IPs, and highlighting security with a lock icon.

​
Controlled Outbound Access with Cloud NAT

When your private cluster needs to reach the internet (for pulling container images or accessing external APIs), configure Cloud NAT. Cloud NAT provides secure, managed outbound connectivity while preserving your nodes’ internal IP addresses.
The image is a diagram explaining a Private GKE (Google Kubernetes Engine) Cluster, showing components like nodes, pods, services, Cloud NAT, and private clusters with internet access.

​
Flexible IP Range Support

As of GKE version 1.14.2, private clusters support multiple internal IP range types:
  • RFC 1918 private ranges (10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16)
  • Privately used public IP blocks
This flexibility lets you align cluster addressing with corporate networking policies.
The image explains what a Private GKE Cluster is, highlighting GKE version 1.14.2, internal IP ranges, private ranges defined by RFC 1918, and privately used public IP ranges.

​
Shared Responsibility Architecture

Private GKE clusters split resources between two projects:
Project TypeComponentsResponsibility
Customer ProjectVPC network, cluster nodes, workloadsYou secure nodes, containers, and applications.
Google-Managed ProjectControl plane (API server, etcd, etc.)Google secures and maintains control plane.
The image illustrates a Private GKE (Google Kubernetes Engine) Cluster Architecture, showing components like the customer project with VPC network and cluster nodes, and the Google-managed project with the control plane. It also indicates no direct internet access to the cluster nodes.

​
Control Plane Endpoints

The control plane exposes:
  • A private endpoint reachable only within your VPC network.
  • An optional public endpoint, which you can disable for maximum isolation.
Enable the public endpoint only if you require external administrative access.
The image illustrates the architecture of a Private GKE (Google Kubernetes Engine) Cluster, showing a control plane connected to both private and public endpoints.

​
Node Connectivity and Isolation

Nodes in a private cluster obtain only internal IP addresses from your subnet. They cannot be accessed directly from the internet, ensuring strong workload isolation.
The image illustrates a Private GKE Cluster Architecture, showing a GKE component connected to pods with no external IP and no direct connection.

​
Service Exposure Options

To route external traffic into your private cluster, choose one of these Kubernetes service types:
  • LoadBalancer Service
    Provisions a managed Google Cloud Load Balancer with a public IP.
  • NodePort Service
    Opens a designated port on each node for direct access.
  • Ingress
    Routes HTTP/HTTPS/TCP traffic by host and path rules through a shared load balancer.

​
Private Google Access

Private Google Access lets VMs without external IPs call Google APIs and services over Google’s network:
  • Enabled: Subnet-bound instances can reach Artifact Registry, Cloud Storage, Pub/Sub, Cloud Logging, and more.
  • Disabled: Instances are barred from Google APIs and managed services.
By default, private clusters in the same project as their VPC have Private Google Access enabled on the cluster subnet.
The image is a diagram illustrating "Private Google Access" within a Google Cloud Platform (GCP) VPC network, showing different subnets and virtual machines with varying access settings. It includes components like Google APIs/Services, Internet Gateway, and VPC Routing.

​
Private Google Access in Shared VPCs

For private clusters in a Shared VPC service project, Private Google Access is not enabled by default. A network administrator or project owner in the host project must manually enable it on each relevant subnet.
The image is a flowchart illustrating "Private Google Access for GKE Private Clusters," showing two paths: "Private Clusters With VPC" (Automatic) and "Private Clusters With Shared VPC" (Manual). It includes icons for Google Cloud Platform and a key symbol.

Watch Video