Walk 4, clickable — one route, two re-originations

Site A (domain 65100:1) → DCI (65099:1) → site B (65030:1). Toggle the route type, then click a hop or step through. ✦ marks a gateway — the only places anything is rewritten.

rewritten at this hop deliberately preserved ✦ gateway = new RD + next-hop-self + D-path stamp (+ router MAC)

Two loopbacks per device: Lo0 = router-id and BGP peering — every overlay session here runs Lo0-to-Lo0 (update-source Loopback0, ebgp-multihop 3). Lo1 = the VTEP — what next hops point at and where tunnels terminate. Spines and the DCI have Lo0 only: no VTEP, and after re-origination, never a next hop. Site A addresses abbreviate 10.255.255.x to .x; site B uses 10.255.31.x.

What never changes: the RT (1:10 / 1:50000) is static, ASN-free, and domain-free — it imports identically in all three domains with zero rewriting, in both route types, at every hop.

What never travels at all: Type-3 IMET routes are not relayed across domains — each gateway originates its own per domain, which is why the three flood domains in Walk 5 are independent.

Why 65030:1 never appears: D-path is stamped on exit with the domain the route was learned in — a domain's own ID enters the D-path only when a route leaves it. An incoming route already carrying your own ID is a loop, and the gateway drops it.