Thanks, all, for the replies and help. I was able to resolve this by configuring the CE to peer with the *primary* ip address on the PE loopback interface. It would not peer with a secondary ip address on that interface, even with "disable-connected-check" and "ebgp-multihop 3" on the CE and "update-source loopback 0" on the PE.
I discovered as well that the CE "ebgp-multihop" command was not even necessary in this scenario. Perhaps it is because I am using "disable-connected-check". On the PE end, I did not need to use a static route to the looopback interface, but I'm thinking that may be because the ATM interface over which the link to the customer is established is configured with "ip unnumbered loopback 0". Thanks again. Adam ----- Original Message ----- From: "Gerald Krause" <[EMAIL PROTECTED]> To: <[email protected]> Sent: Thursday, November 15, 2007 3:29 PM Subject: Re: [c-nsp] BGPoPPPoEoA ?! > Ok, but I need none if the GW IP address from the PPP negotiation is the > loopback IP address on the PE in question. That's why I have no configured > static route on the CE. > > On Thursday 15 November 2007 20:47:56 Aaron wrote: >> Don't forget the static for the loopback >> >> On Nov 15, 2007 2:30 PM, Gerald Krause <[EMAIL PROTECTED]> wrote: >> > On Thursday 15 November 2007 17:40:54 Adam Greene wrote: >> > > Lots of o's in that subject line ... >> > > >> > > I'm trying to set up a BGP session over a PPPoEoA DSL line. This is >> > > in >> > > the context of setting up redundant DSL lines to a single provider >> > > router. I control both ends (PE and CE). PE is 7200 NPE 200, IOS >> > > 12.3(15b). CE is 1841, IOS 12.4(17). >> > > >> > > I can't establish the BGP session. Both sides are in active state, >> > > but >> > > won't go further. >> > > >> > > The PE ATM interface is configured as IP unnumbered pointing to >> > > Loopback 0. The CE BGP neighbor thus points to the PE Loopback IP >> > > address. The PE BGP neighbor points to the IP address assigned to the >> > > CE Dialer (a /32 from the /23 block on the PE Loopback). >> > > >> > > I saw that the CE was reporting that the external BGP neighbor is not >> > > directly connected, so I issued "neighbor A.B.C.D >> > > disable-connected-check" to no avail. >> > > >> > > I tried specifying the update-source interface on both ends (loopback >> > > 0 >> > > on PE, Dialer1 on CE) again to no avail. >> > > >> > > I'm wondering if I have to do something at the neighbor A.B.C.D >> > > transport level, like disabling path-mtu-discovery (this is a wild >> > > shot >> > > in the dark). >> > > >> > > Has anyone else successfully established BGP over PPPoEoA before? >> > >> > Yes, and it (still) works. As Aaron and Peter already mentioned you >> > should use "ebgp-multihop" on both systems. Our config looks like this: >> > >> > CE config: >> > ========== >> > ! >> > interface Dialer1 >> > ! local IP address is always 10.250.250.50/32 and >> > ! remote-GW IP address 10.255.255.255 - from PPP/RADIUS >> > ip address negotiated >> > encapsulation ppp >> > ... >> > ! >> > ! >> > router bgp 65534 >> > neighbor 10.255.255.255 remote-as 123 >> > neighbor 10.255.255.255 ebgp-multihop 2 >> > ... >> > ! >> > >> > PE config: >> > ========== >> > ! >> > interface Loopback101 >> > ip address 10.255.255.255 255.255.255.255 >> > ! >> > ! >> > router bgp 123 >> > ... >> > neighbor 10.250.250.50 remote-as 65534 >> > neighbor 10.250.250.50 ebgp-multihop 2 >> > neighbor 10.250.250.50 update-source Loopback101 >> > ... >> > ! > > > -- > Gerald (ax/tc) > _______________________________________________ > cisco-nsp mailing list [email protected] > https://puck.nether.net/mailman/listinfo/cisco-nsp > archive at http://puck.nether.net/pipermail/cisco-nsp/ > > > > > _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
