From: ccie candidate (ccie1@xxxxxxxxx)
Date: Sat Aug 03 2002 - 14:03:55 GMT-3
you are mixing between two things , the route itself and the next hop of it .
any router which runs BGP , when it learns a certain route from another peer in
the same AS (learnt through IBGP) , two conditions has to be met for the route
r to push this route into the routing table
1-the next hop of the route should be reachable (here where you sometimes need
next-hop self )
2-the route itself should be known through IGP protocol (here where you sometim
es disable synchronization to skip this condition ).
if the synch is disabled , you must get the route into your routing table throu
gh IGP before the BGP will put it in its routing table (again this is only for
routers learnt that route through IBGP not EBGP)
HTH
--On Sat, 3 Aug 2002 12:51:01 Tom Larus wrote: >I am working on a practice scenario in which we cannot disable >synchronization, and I am finding that I am several "no valid path" errors >in my BGP debug output. I thought that when we have to use synchronization, >we should redistribute BGP into the IGP at the edges of the IGP. That would >not, however, make the next hop of each IGP route match up with the next-hop >of the BGP route. Any advice on this. > >Do we redistribute just at the edge of each IGP, or on each router? > >Am I supposed to put "next-hop-self" on every BGP router? Even then, I >might have to tinker with IGP metrics and ADs to make sure the IGP next-hop >and BGP next-hop agree. > >"No syn" makes everything work fine, but that is not an option.
This archive was generated by hypermail 2.1.4 : Sat Sep 07 2002 - 19:48:15 GMT-3