From: Rick (rick@iptool.net)
Date: Sun Jan 02 2005 - 01:16:05 GMT-3
I have a question regarding next hop self. When you have a bgp peering session
using loopbacks and updatesource lo0 the next hop of the advertised routes
have the loopback address instead of the interface address. Here is a where my
questions comes from
This example is using interface addresses for all bgp peering and R3 is a
route reflector to R2 and R4:
R1---ebgp----R2----ospf & ibgp-----R3-----ospf & ibgp------R4
The link between R1 and R2 is not advertised in BGP so when R3 receives the
routes it will not install as it can't resolve the next hop. to solve this
problem you just use next-hop-self on R2 to R3 and then R3 will install the
routes and advertise them on to R4
This example is using loopbacks for all ibgp peering and interface addresses
for ebgp(R1-R2) and R3 is a route reflector to R2 and R4:
R1---ebgp----R2----ospf & ibgp-----R3-----ospf & ibgp------R4
The link between R1 and R2 is not advertised in BGP, however R3 gets the
routes with R2's loopback as the next-hop. Now since R3 knows how to get to
the loopback it will install the route and advertise the learned routes on to
R4.
So my questions are should we still use next-hop-self with the second example
even though it isn't needed? For lab purposes is there a reason to not peer
with loopbacks unless of course you are instructed not to?
Rick
This archive was generated by hypermail 2.1.4 : Wed Feb 02 2005 - 22:10:18 GMT-3