Last reviewed: May 2026
If GCP returned this error, the fix is mechanical once you can see which existing range you collided with. This page shows how to find it and how to choose a range that will be accepted.
The primary or secondary IP range you supplied intersects a range already assigned to a subnetwork in the same VPC network. Because GCP VPCs are global rather than regional, the conflicting subnetwork may be in a completely different region from the one you are deploying into.
List what is already allocated before guessing at a new range:
gcloud compute networks subnets list \ --filter="network:my-vpc" \ --format="table(name, region, ipCidrRange, secondaryIpRanges[].ipCidrRange)"
The GKE trap: a GKE cluster consumes three ranges — nodes (primary), pods (secondary) and services (secondary) — and the pod range is large, commonly a /14. Engineers plan the primary range carefully, let GCP auto-allocate the secondaries, and then collide months later when a second cluster is created. Always plan pod and service ranges explicitly.
This error is almost always a symptom of address space that was never planned centrally. Allocating each environment a documented block from one supernet, and checking new ranges against that record before deployment, removes the whole class of failure. The Cloud VPC & VNet CIDR Planning Guide sets out that approach.
The primary or secondary IP range you supplied intersects a range already assigned to a subnetwork in the same VPC network. Because GCP VPCs are global rather than regional, the conflicting subnetwork may be in a completely different region from the one you are deploying into.
Not within the same VPC or virtual network. Providers require subnets to be disjoint so routing is unambiguous. Overlap is also what blocks VPC peering, VNet peering and Transit Gateway attachments between networks.
List every existing range in the network, then compare each one against the range you are trying to create. The bulk subnet calculator on this site accepts CLI JSON and CSV exports directly and reports every overlapping pair, which is faster and more reliable than comparing by eye.