Blog

Overlapping VPN Subnets: Should You Use NAT or Renumber?

Choose renumbering or NAT for a site-to-site VPN with overlapping subnets, using a worked address map and clear gateway, DNS, and application responsibilities.

When two sites use overlapping private subnets, choose between giving them distinct address ranges and translating addresses at the connection boundary. Renumbering is usually the cleaner long-term choice for networks being integrated extensively. Network address translation (NAT) can preserve existing addresses when renumbering is impractical, provided the required applications, routes, and gateway capabilities support the design.

For an IT manager connecting an acquired office through a site-to-site virtual private network (VPN), the decision starts with which systems must communicate. A single shared application may justify a limited translation arrangement. Combining the offices into one broadly connected environment makes a coordinated addressing plan more valuable.

Why the same private subnet creates ambiguity

Private Internet Protocol (IP) addresses can be reused by separate organizations. RFC 1918 explains the collision risk when those networks are joined: an address that was unique inside each business may identify different machines in the combined environment.

Suppose both offices use 10.20.0.0/24, meaning the same 256-address block. A computer at headquarters cannot use the bare address 10.20.0.50 to distinguish its local device from the acquired office's server at that address. Encrypting a connection between the sites does not supply that missing distinction.

For ordinary routed connectivity between these sites, each remote destination needs an unambiguous address from the caller's perspective. Renumbering changes the actual address assignment. NAT creates a translated address that the other network can use.

Choose the operating cost you can sustain

Renumber when the networks need lasting integration

Renumbering moves one overlapping network into a range that is unused across the intended connected environment. Review branches, cloud networks, partner connections, and remote-access pools before choosing it; a different range can still collide elsewhere.

Our recommendation is to favor this approach when many applications and administrators will need access across both sites. One coordinated address plan reduces the number of alternate identities that operators must maintain. The transition requires work on address assignments, routing, firewall policies, name resolution, monitoring, and applications with fixed addresses. Plan that work with the system owners and an agreed rollback.

Translate when preserving local addresses has a clear purpose

NAT can be appropriate when a partner controls the remote network, a packaged system cannot readily change addresses, or a limited connection is needed during a longer integration. It adds an operating obligation for as long as it remains in use: maintain the mappings and keep applications, policies, and diagnostic records consistent with them.

Microsoft's Azure VPN Gateway documentation illustrates why platform details matter. It describes static and dynamic translation, requires non-overlapping translated ranges, and limits which connection types support NAT. Its dynamic rules also restrict which side may initiate a connection. Those are Azure-specific capabilities and constraints, not evidence that every VPN gateway behaves the same way.

Network Box USA's Managed Unified Threat Management (UTM+) service includes NAT, advanced routing, Internet Protocol Security (IPsec) VPN, stateful firewall policy, and logging, with managed design, validation, and tuning. That work helps coordinate the permitted application path with its gateway controls. Confirm the proposed translation mode, peer compatibility, and policy behavior for the actual deployment; the general feature list does not establish every overlapping-subnet design.

Work through one connection in both directions

Consider Northline Parts, a hypothetical distributor acquiring a warehouse. Its headquarters staff need to query the warehouse's inventory server. Both locations use 10.20.0.0/24, and the warehouse cannot renumber before the connection is needed. Its IT manager, Dev, proposes translation for this application while planning the wider integration.

Assume the gateway design supports the required source and destination translations. Reserve 172.20.10.0/24 to represent headquarters remotely and 172.20.20.0/24 to represent the warehouse remotely. For this example, both ranges are unused throughout the connected environment. They are planning examples, not ranges to adopt without checking.

  1. Headquarters sends the request. Its client at 10.20.0.25 connects to the warehouse server's translated address, 172.20.20.50.
  2. The warehouse server receives it. After the required translations, the destination is its real address, 10.20.0.50. The source appears as 172.20.10.25, representing the headquarters client.
  3. The server replies to that remote identity. It sends the response toward 172.20.10.25 through the agreed gateway path. It does not mistake the client for a machine on its own overlapping subnet.
  4. Headquarters receives the response. Reverse translation delivers it to 10.20.0.25, with the server appearing as 172.20.20.50, the destination the client originally contacted.

The useful design artifact is this pair of identities for each endpoint, together with the route in each direction. Cisco's overlapping-network example demonstrates why translating the client source as well as the server destination can be necessary. The sequence above describes the required endpoint views; the gateway operator must establish where translation occurs and how it interacts with VPN routing and firewall-rule evaluation.

Make the application use the intended address

For the warehouse example, headquarters' name resolution or application configuration must direct the inventory connection to the translated server address. Returning the warehouse's original 10.20.0.50 to that client would reintroduce the ambiguity. The relevant Domain Name System (DNS) and application owners need to implement that part of the plan.

Also check whether the application sends IP addresses inside its messages or opens additional connections. RFC 3022's discussion of NAT limitations explains why changing packet-header addresses may be insufficient for such protocols. Test the complete inventory transaction, including any callbacks, against supplier guidance. A successful initial query does not exercise a later server-initiated connection.

Translation supplies addressing; firewall policy still decides which connections are permitted. In this example, approve only the required headquarters-to-inventory service. Verify that unrelated warehouse services remain inaccessible and establish which original or translated addresses the deployed policy expects. Keep the mappings available to authorized investigators; our guide to tracing NAT records explains why address context matters when interpreting an alert.

The customer and its application teams retain endpoint renumbering, DNS, and application changes unless separately agreed. Before approving the connection, give the gateway operator the original ranges, proposed translated ranges, required services, connection initiators, and long-term integration plan. Contact our team to review that managed gateway scope.