BGP Peering with HSRP Virtual Address

netservices · ‎11-23-2007

Are there any caveats associated when peering in BGP with an HSRP virtual address?

I have a national ethernet circuit which terminates on a layer 2 switchport. The vlan of this switchport is then trunked to 2 Layer 3 gateway routers running HSRP between them for resiliency.

I want to set up eBGP between the CE on one end of the link and the PE routers at the other end.

Is it possible for the CE to peer with the HSRP virtual address? I'm hoping that whichever PE router is the active router for the standby group will peer with the CE and in the event of a PE router failure, the standby router will take up the responsibility of peering with the CE.

Thoughts anyone?

royalblues · ‎11-23-2007

You cannot peer with the HSRP virtual address

The TCP packets for BGP peering is allowed to source only from an interface, not from a virtual address.

You may be able to configure the neighbor pointing to the HSRP address but the remote connection will be refused since it will be sending the packet from an invalid IP address.

HTH

Narayan

Harold Ritter · ‎11-28-2007

Narayan,

It is actually possible to run BGP session to the VIP address. This might require to configure the HSRP routers to passively accept the BGP session (nei x.x.x.x transport connection-mode passive".

Bear in mind that the convergence will be slower than if you had a BGP session between the CE and both PEs since the BGP session will need to reinitialize when a fail over occurs.

Regards,

Harold Ritter
Sr Technical Leader
CCIE 4168 (R&S, SP)
harold@cisco.com
México móvil: +52 1 55 8312 4915
Cisco México
Paseo de la Reforma 222
Piso 19
Cuauhtémoc, Juárez
Ciudad de México, 06600
México

AJAZ NAWAZ · ‎12-31-2015

implementing aggressive timers would minimise failover period required for reconvergence.

good shout harold !

Alexei Monastyrnyi · ‎11-23-2023

HSRP is a LAN technology, a First Hop Redundancy Protocol, and should be used on the LAN side of things IMO. Of course, we can hack things and make them work in a twisted way (some folks treat it as CCIE way :-), but would we want such a solution in a production environment? One should put together extensive redundancy testing with such implementation and there is still there is a chance for a FFFF-up simply cause the technology in use is not designed for the use-case.