Engineering experience in BGP Route reflectors

Some experiences in man will be shared with you. The approximate topology is shown in (physical connection diagram). The two downlink devices (R1 and R2) and the two core devices both establish IBGP neighbor relationships with the two route reflectors RR, in the BGP of R1 and R2, use the network to publish the network segment 192.168.1.0/24 at the same time. 1. If no multiple paths (maximum-paths) are set on the two route reflectors, note: Cisco divides IBGP and EBGP. If the ibgp parameter is not followed, only allow many equal-cost paths in ebgp, but not from Huawei). RR selects only one optimal route and adds it to the routing table (without any other parameters set, select the one with a small router-id) and then reflect it to the two cores. 2. If multiple paths are set on the two route reflectors, the two RR will select multiple equal paths. However, note that when viewing the BGP table, only one route entry is displayed *> (the route entry with a small router-id), but multiple routes are shared when you look at the route table. At the same time, when the two RR reflection routes are routed to the two core routers, only the optimal route is reflected, that is, the route with *>, therefore, only one route entry is displayed on the two core routers. 3. For the next-hop-self command, when using some commands on the two route reflectors, the next hop will not be changed to itself when the reflection routes to the core, that is, when next-hop-self is used as the route reflector, the reflection route will not change the next hop. 4. For devices of Huawei and Cisco, the next hop generated during network is different: When Cisco devices publish a route learned from IGP on the network (or redistribution, by default, the next hop is the next hop address of IGP. Similarly, when a Huawei device publishes a route entry learned from IGP in network (or redistribution), by default, next Hop is your own address, that is, the address used to build BGP neighbors)

This article is from the "Esc end" blog