cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
336
Views
0
Helpful
3
Replies

BGP next-hop/routing issues

jackson.a
Level 1
Level 1

Scenario :

5 routers connected via a single vlan. See text diagram below (i've had to use dot's as spaces don't format correctly) :

< ----ASN2---->..............<ASN3>

Rtr C......Rtr D................ Rtr E

___|_____|__________|___ Ethernet LAN

.........|............... |

......Rtr A........Rtr B

....<-----ASN1------>

Configs are :

Rtr A -

!

router bgp 1

neighbor <rtr B> remote-as 1

neighbor <rtr C> remote-as 2

neighbor <rtr D> remote-as 2

neighbor <rtr E> remote-as 3

!

Rtr B -

!

router bgp 1

neighbor <rtr A> remote-as 1

neighbor <rtr C> remote-as 2

neighbor <rtr D> remote-as 2

neighbor <rtr E> remote-as 3

!

Rtr C

!

router bgp 2

neighbor <rtr D> remote-as 2

neighbor <rtr A> remote-as 1

neighbor <rtr B> remote-as 1

!

Rtr D

!

router bgp 2

neighbor <rtr C> remote-as 2

neighbor <rtr A> remote-as 1

neighbor <rtr B> remote-as 1

!

Rtr E

!

router bgp 3

neighbor <rtr A> remote-as 1

neighbor <rtr B> remote-as 1

!

Issue :

If you are connected to rtr C or D then routes for networks connected to rtr E have the correct next-hop ip address of the rtr E ethernet port (as per the BGP documentation on how the protocol should operate on "Multi-access" networks such as ethernet)

If you are connected to rtr E then routes for dual-homed networks connected to rtr C & D have an INCORRECT next-hop ip address of the rtr A or B ethernet port. One strange observation....... During testing if you reset the BGP connections on Rtr C and check the routing before these have been re-established then Rtr E sees the ethernet interface of rtr D as the next-hop - which is correct (As soon as the neighbor connections are re-established the routes revert to the INCORRECT next-hop ip address)

I hope the above is clear enough to understand.

Thanks,

Adam

3 Replies 3

thisisshanky
Level 11
Level 11

Please note you have EBGP peering from E only to A and B. So routes from C and D are advertised by A or B to E. So E will not have the next hop of C or D for those networks. THey will have the next hop of A or B.

This is the behaviour of EBGP.

IF, A B and E were in same AS, the peering between them would have been IBGP. in thsi case, the next hop of C or D would have been preserved.

Hope its clear.

Sankar Nair
UC Solutions Architect
Pacific Northwest | CDW
CCIE Collaboration #17135 Emeritus

See link: http://www.cisco.com/warp/public/459/14.html#A13.0 under section "BGP Nexthop (NBMA)".

It seems in your example the nexthop is changing only one way (ie C&D have the correct nexthop for E but E doesn't for C&D).

Try a debug ip bgp or debug ip tcp transaction to see what is going on.

Steve

Thanks for the link above, this is one of many I've looked at for an answer to this. You are correct in your statement of the nexthop is changing in only one direction.

I've had debug ip bgp events / updates running (on Rtr E) and cleared all the BGP connections from Rtr C. At this point the only device in ASN2 is router D and therefore you see updates from RtrA/B changing the next-hop to router D (at this brief point everything then looks ok)

The debug shows :

BGP(0): rcvd 10.x.x.x/24

BGP(0):Non-multipath->multipath for 10.x.x.x/24 from 98.39.123.64

BGP(0):rcvd UPDATE w/attr:nexthop origin i, path ASN1 ASN2

(These updates repeat from RtrB - the end subnet is 10.x.x.x/24 in the above)

As soon as Rtr C re-establishes its BGP neighbor connections (and ASN2 now contains 2 devices) you see the next-hop change back to RtrA/B

BGP(0): Revise route installing 10.x.x.x/24 -> to main IP table

Any ideas ???

Adam

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: