PuppyIP Resource Center
Network And Account Troubleshooting 5 min Published 2026-11-20

How to verify a fixed namespace egress IP with GKE Granular NAT

Before registering a GKE egress IP with a partner, contact the same test endpoint from a matching namespace and a control namespace. Check the public source IP at the receiver. Enabling Granular NAT Preview on an existing cluster recreates all node pools, so plan the test and change window first.

GKE Granular NAT Fixed egress IP Kubernetes

Service eligibility and regional restrictions

PuppyIP serves only compliant overseas businesses and their authorized personnel. Proxy services are not available in mainland China. The service may only be used for lawful business activities outside mainland China. Use of this service within mainland China is prohibited.

Hosting a proxy IP or server overseas does not change these restrictions. The service must not be provided to end users in mainland China through relaying, forwarding, sharing or resale. Before use, read the Terms of Service.

Key Takeaways

  • GKE Granular NAT is in Preview. It supports VPC-native Autopilot and Standard clusters with Dataplane V2 and version 1.37.1-gke.2611000 or later.
  • Public NAT requires a regional Cloud Router, reserved public IPs in that region, and VPC routes covering the destination CIDRs with the default internet gateway as next hop.
  • Verify the source IP at the receiver from both the target and control namespaces. For cleanup, remove user-created TrafficSelector resources, then CloudNATGateway resources, and disable the feature last.

Check cluster eligibility and the effect on node pools

Google Cloud announced GKE Granular NAT Preview on November 18, 2026. It supports VPC-native Autopilot and Standard clusters, whether regional or zonal, with GKE Dataplane V2 enabled and version <code>1.37.1-gke.2611000</code> or later.

Google's setup guide says enabling the preview on an existing cluster recreates all node pools and can disrupt workloads. To validate the address a partner sees, you can first check routing and the source IP in a test cluster, then use your team's change process for the existing cluster.

Match the gateway, namespace, and destination network

A <code>CloudNATGateway</code> describes the gateway. A <code>TrafficSelector</code> uses namespace labels and destination CIDRs to match traffic. Preview policies apply at namespace scope. If just one Pod needs different egress, plan the namespace layout first. Selectors in this preview cannot select a single Pod or ServiceAccount.

For a partner's public service, check the regional Cloud Router and reserved public IPs. GKE does not automatically create the Cloud Router for a new gateway. Verify that a static route in the cluster VPC covers each destination CIDR and uses the default internet gateway as next hop. For an existing gateway, check that its region and VPC match. Private NAT serves supported NCC or hybrid connectivity. This public-address test applies to Public NAT.

Verify the actual source address at the receiver

Use a test receiver you are authorized to access and log the client source IP. Connect to that same destination from the matching namespace and an unmatched control namespace. Record the time, destination, and observed source. The target namespace should use the selected NAT address. Interpret the control result using the node and VPC's existing egress. If Cloud NAT translation logs are enabled, cross-check <code>nat_ip</code>, <code>dest_ip</code>, and the gateway identifier.

If the actual IP differs, check the namespace label, whether the destination IP matches the intended CIDR, the selected gateway and address, and the VPC route's next hop. If the receiver sees no request, check connectivity and network policies. Register the IP with the partner after the receiver observes the expected source.

Let the controllers clean up before disabling the feature

To remove the configuration, delete the <code>TrafficSelector</code> resources you created by name and preserve system selectors. Wait for deletion, then remove the <code>CloudNATGateway</code> resources. Disable Granular NAT last. Disabling first can stop the controllers, leave resources stuck in <code>Terminating</code>, and leave NAT rules behind. For shared gateways, check Google's documentation for which gateways and Pod IP ranges remain.

Sources

Frequently Asked Questions

Can I assign egress by Pod or ServiceAccount?

The preview matches by namespace. Check the namespace labels and the destination CIDR in the TrafficSelector.

Can I submit an IP to the partner after creating the selector?

First contact an authorized test endpoint from the matching namespace and verify the observed public source IP at the receiver. Submit the verified address.

Can enabling the preview on an existing cluster interrupt workloads?

Google documents a recreation of all node pools that can disrupt running workloads. Plan a change window and verify on a test cluster first.