Case study
Transit change across two sites without renumbering
Summary
The problem
Two sites, a primary and a backup, had to move to a new transit provider and new edge hardware in the same change, without renumbering any of the organisations connected behind them.
What we did
Designed the new edge: eBGP to the provider and the local IX route server at both sites, iBGP between four routers, an OSPF ring between the sites, VRRP for customer gateways, and inbound traffic steered with local preference and AS-path prepending. Wrote and lab-tested the configurations, then wrote the cutover runbook, the failover design and the handover documents for the client’s team.
Outcome
Cut over at both sites, with every connected organisation still on its existing addresses.
In more detail
Failover in both directions
Failover between the sites is designed to be predictable. Outbound, a local-preference ladder decides which exit wins and in what order the others take over. Inbound, AS-path prepending on the backup site tells the rest of the internet which way to come in, and the BGP sessions use RFC 9234 roles so a misconfigured neighbour can’t turn the network into an accidental transit path.
Addresses stay where they are
The organisations connected behind the edge keep their prefixes and their gateways. VRRP keeps the gateway address alive across a router failure, and the same prefixes are announced through the new provider, so only the path changes at cutover.
Written for the people who run it
The handover is a set of documents the client’s own team can work from: the configuration for every device, the failover behaviour in plain language, and a cutover runbook with a check after every step and a way back if one fails.