From: Peter van Oene (pvo@xxxxxxxxxxxx)
Date: Tue Mar 12 2002 - 16:07:01 GMT-3
A route reflector should not change the BGP next_hop when reflecting
whereas next-hop-self should. Further, next-hop-self in now way helps
alleviates the full IBGP full mesh issue. For example, if A sets
next-hop-self toward B, A will not reflect B's routes to the rest of the
IBGP mesh, thereby creating the potential for black holes.
In general, the BGP next hop is not usually re-written on IBGP connections.
Pete
At 08:41 AM 3/12/2002 -0600, Michael Snyder wrote:
>I've been reading up on BGP lately, and have come up with a question.
>
>Is there a functional difference between setting a neighbor as a route
>reflector client, or setting a neighbor as next hop self? Assuming 'no
>sync', to my mind these commands seem to do the same thing. The same
>network topology and limits seem to apply to each command.
>
>
>
>Michael Snyder
>Lead Engineer
>Revolution Computer Systems
>CCNP/DP MCSE NT/2000
>(270) 443-7400
>(Fax) 443-7070
>
>
>-----Original Message-----
>From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
>Peter van Oene
>Sent: Monday, March 11, 2002 9:12 PM
>To: Stephen Oliver; ccielab@groupstudy.com
>Subject: Re: BGP and OSPF synchronization
>
>BGP synch is long dead and gone. Further, it may have been dead before
>route reflection became commonplace. Given 1403 (bgp to OSPF
>interaction)
>predates 1863 (first hint of RR technology) by two years, it occurs to
>me
>that possibly interaction with BGP synch wasn't a tested feature of RR
>code, particularly wrt to 1403 operation. This would make sense since
>one
>wouldn't run into IBGP scaling issues if one were running a non full
>IBGP
>mesh transit AS. The two features indirectly solve the same problem.
>
>It's a sad state of affairs when intelligent folks have to waste quality
>
>time learning useless features.
>
>
>At 07:13 PM 3/11/2002 +0000, Stephen Oliver wrote:
> >I have 3 routers in a frame relay hub and spoke ospf configuration.
> >
> >The ospf is working fine and I have set the router IDs to the router
> >number. R1 is 1.1.1.1 etc.
> >
> > r1 ---------r5-----------r2
> >
> >I have configured BGP in AS 13 on all the routers with r5 as a
> >route-reflector to both r1 and r2. Next I add a loopback on r1 and r2
>and
> >include them in OSPF. They are reachable everywhere. I then add the
> >loops into BGP. When they get to R5 they are unsync because of the
> >OSPF/BGP router ID issue so I change the router IDs on R1 and R2 to
>match
> >OSPF and hey presto R5 syncs the loopbacks.
> >
> >Now when I look at the spoke routers even though they get each others
> >loopbacks into their BGP tables they have marked them as unsync because
>
> >they are sourced from R5 for BGP and from the other end router for
> >OSPF. Another RID mismatch.
> >
> >I can overcome this by simply adding a network statement for the
>unsynced
> >prefixes on each spoke router.
> >
> >Is this a valid solution to the problem ?
> >
> >The whole scenario is just a test to see OSPF/BGP interactions.
> >
> >Thanks, Stephen.
> >
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:57:02 GMT-3