From: Mike Gutknecht (mgutknec@xxxxxxxxx)
Date: Tue Feb 20 2001 - 19:32:08 GMT-3
Hi Andrew,
I agree with you on the description below and that adding the next-hop-self
makes it a more 'bullet-proof' config in that the BGP configuration is not
relying on a specific OSPF configuration detail. On the second point,
however, I don't really get what you mean.
The IBGP peer is not "communicating" with the EBGP peer at all.
Say router 5 (IBGP peer) had traffic for a network learned via BGP that
originated out at router 3 or 4. The BGP_NEXT_HOP is either router 3 or
router 1 (depending on whether you use the next-hop-self). If it is using
router 3 as the BGP_NEXT_HOP, then it will recurse to find a route to router
3.
Using the ospf network (P-MP) statement, router 5, has a host route for
router 3 whose next hop is router 1. So router 5 sends the packet to router
1. If there were no ospf network statement, we would need a frame-map
statement, policy routing, etc on router 5 to get the packet to router 1
since router 1 is the only router with a PVC built to router 3.
Given that router 5 and router 3 do not have layer 2 connectivity, they will
never be directly "communicating."
my $0.02
-Mike G
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]On Behalf Of
Andrew G. Mason
Sent: Tuesday, February 20, 2001 12:10 PM
To: Mike Gutknecht; ccielab@groupstudy.com
Subject: RE: cciebootcamp #2, #4 - Next hop self
Hi,
I found that everything works fine without the next-hop-self command but
only because the next-hop for the BGP routes is already in the IP routing
table as a host OSPF route (due to the ip ospf network type). If these
routes were not there, the BGP route would not make it to the IP routing
table, hence the need for the next-hop-self command.
Also, you do not really want an IBGP peer communicating directly with an
EBGP peer when no direct peering exists between them both.
Using the next-hop-self allows you to control the egress points (and policy)
for the network.
HTH
Andrew..
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]On Behalf Of
Mike Gutknecht
Sent: 20 February 2001 19:49
To: 'Lachlan Kidd'; 'Scott Schneidewind'; ccielab@groupstudy.com
Subject: RE: cciebootcamp #2, #4 - Next hop self
Not quite. The "ospf network point-to-multipoint" command on R1 inserted
/32 s for every host on the NBMA cloud and they got distributed either via
OSPF or EIGRP.
I don't think the "next-hop-self" command really needs to be there. I ended
up with full connectivity without it.
-Mike Gutknecht
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]On Behalf Of
Lachlan Kidd
Sent: Monday, February 19, 2001 3:50 PM
To: Scott Schneidewind; ccielab@groupstudy.com
Subject: RE: cciebootcamp #2, #4 - Next hop self
Hi Scott,
The reason for this is that R1 is an the same AS as R2 and R5. Without
next-hop-self, R3 will see routes from R5 and R2 with a next hop of R5/R2's
IP address. Due to the frame/ospf setup, I'm pretty sure that R5/R2's s0 IP
address is unreachable and hence the route will not be published in the
routing table. By setting next-hop-self, the routes will appear with a next
hop of R1 which is directly connected to R3 and hence reachable....at least
I think that's how it goes <grin>
Regards,
Lachlan
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]On Behalf Of
Scott Schneidewind
Sent: Tuesday, 20 February 2001 6:23:AM
To: ccielab@groupstudy.com
Subject: cciebootcamp #2, #4 - Next hop self
Hello All.
Taking a look at the cciebootcamp labs #2 and #4, router 1 has the
neighbor <router 3> next-hop-self
Could somebody please explain why this router needs the command and no other
routers do?
Thanks,
Scott
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:28:54 GMT-3