From: Jim Brown (Jim.Brown@xxxxxxxxxxxxx)
Date: Mon Jul 15 2002 - 12:00:33 GMT-3
Creating a default route could create problems and routing loops. Imagine
this....
There are two exit points from AS1 using BGP routers R5 and R3
The non-BGP router R2 points towards the R5 exit point as the default route.
A BGP learned route appears more desirable via the R3 exit point instead of
the R5 exit point.
If R5 sends a packet towards R3 via R2 because R3 is the more desirable exit
point for this particular packet based on BGP information. R2 would
immediately send it back towards R5 based on the configured default route
and around we go.
In essence, you have created a routing loop between R5 and R2 for BGP only
routes which should exit via R3 if you use a default route.
-----Original Message-----
From: Jonathan V Hays [mailto:jhays@jtan.com]
Sent: Sunday, July 14, 2002 8:34 PM
To: 'Yakout esmat'; 'Groupstudy'
Subject: RE: BGP and OSPF redistribution necessity
Yakout,
I find this an interesting problem. But I doubt if redistributing BGP
into OSPF would be allowed as a solution in the CCIE Lab. And I don't
think it would help. The problem stems from the fact that when you ping
a host in the remote AS, the hub router drops the packet because it is
not participating in BGP and has no route to that host.
Here is what I just set up in my lab:
R4(AS4)--EBGP--(AS1)R5/IBGP--R1/IBGP--R2--R3/IBGP
I configured AS1 with various 10.1.x.x/24 networks
R2 is a non-BGP router in the role of HUB, in a frame-relay cloud.
R1,R2,R3, and R5 are all running IBGP in AS1 and all in OSPF area 0.
Every router in AS1 can ping every other router in AS1.
R4 is peering via EBGP with R5.
Let's call R4 the "remote AS."
I had R4 advertise 144.44.0.0 into AS1 and injected 10.1.x.x into AS4.
After configuring a full mesh (I didn't use route reflectors) and
turning off synchronization on all the IBGP routers the 144.44.0.0 route
appeared in both the BGP table and the routing table on all routers in
AS1.
As you say, R5 and R1 have no problem pinging host 144.44.4.4, which is
in remote AS4. But R3 (on the other side of the hub router R2) fails to
ping, getting host unreachable ICMP messages from R2. Of course R2 has
no knowledge of 144.44.0.0 since it is not participating in BGP.
Your idea of redistributing BGP into OSPF would be a solution. The other
solution would be to configure R2 with a default gateway, pointing
towards R5. I understand that default and static routes are generally
forbidden in the CCIE Lab. The next-hop-self command briefly flicked
through my mind, but that doesn't help this situation.
Q1. Did I recreate your problem accurately?
Q2. Any comments from the experts? Is redistribution of BGP into OSPF a
legitimate approach to this lab scenario?
Jonathan
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
Yakout esmat
Sent: Saturday, July 14, 2001 8:59 AM
To: Groupstudy
Subject: BGP and OSPF redistribution necessity
Gents,
I ahve a doubt here about BGP and OSPF redistribution necessity.
Scenario:
If I have three routers in one AS, one of them is running as Route
Reflector and the other two are Clients. Becasue of Hub and Spoke
topolgy, traffic between the Route Reflector and one of the clients has
to go through another NON-BGP router.
In scenario like this one, if I don't redistribute BGP into OSPF (IGP
running) and vica vers, the NON-BGP router will drop the packts if I am
trying to ping one of the external BGP networks, because this router
doesn't have this network in Routing Table.
So, can we say, as a rule of thumb, whenever we have non-bgp router in
the middle between two bgp routers in the same AS, we HAVE to
redistribute bgp into igp and vica versa??
Cheers
This archive was generated by hypermail 2.1.4 : Sat Sep 07 2002 - 19:36:30 GMT-3