BGP Remote Peering - End-to-End

[netbeginner]

Hi All, 

We have a Scenario where Router-1(Service provider Router) --> Router-2 (HoP Router)--> Router-3 (Customer Router) connected configured with BGP.

Requirement is :  if BGP peering between Service provider(Router-1) and Router-2(HoP Router) goes down , subsequently this should make the BGP peering down between Router-2(HoP Router) and Router-3(Customer End). Which is not happening now and showing UP always due to which traffic is not getting shifted to backup link from customer end and resulting as service outage.

Current Scenario : 

[balaji.bandi]

Quick fix i can think of using IP SLA and EEM and shutdown the interface on router 2

So router 3 can think link failed and use different path ?

i

[Georg Pauwen]

Hello,

you could use conditional advertising. Where is the traffic from Router 3 supposed to go when the link between Router 2 and Router 3 fails ? To another BGP peer ?

[paul driver]

Hello
You don't mention:

• what routes you are receiving from the primary/secondary ISPs (full/partial/default routes etc..) 
• How and where that secondary ISP connection is being made,
• If any traffic engineering is applied so all/specific egress/ingress traffic is preferred via the primary/secondary ISPs or both?

I would imagine even if you lose you upstream primary ISP connection (r1-r2) the ebgp peering from (r2-r3) will still be established however you would lose all advertised ebgp routes originating via the primary ISP (r1) so any alternate path with its least preferred routes being advertised by secondary ISP should then populate the routing table of R3 and re-route traffic , unless the present resiliency setup if incorrect?

So can you elaborate a little bit more on your topology please and how it is currently setup.

[MHM Cisco World]

event manager applet bgp-down
event track 1 state down
action 1.0 cli command "enable"
action 1.1 cli command "conf t"
action 1.2 cli command "router bgp 100"
action 1.3 cli command "neighbor 100.0.0.2 shutdown"
action 1.4 cli command "exit"
event manager applet bgp-up
event track 1 state up
action 1.0 cli command "enable"
action 1.1 cli command "conf t"
action 1.2 cli command "router bgp 100"
action 1.3 cli command "no neighbor 100.0.0.2 shutdown"
action 1.4 cli command "exit"
!
ip sla 1
icmp-echo 110.0.0.2 source-interface FastEthernet0/0
ip sla schedule 1 life forever start-time now
!

track 1 ip sla 1 reachability


I make this EEM using IP SLA and it work. 

[netbeginner]

Hi MHM, 

On which router You recommending to configure EEM , I believe it's R2 ? and what exactly You are reffering here with IP : 100.0.0.2  as per my diagram scenario.

[Georg Pauwen]

Hello,

what is the backup link on the customer Router 3 ? Is that a BGP peer as well ? I am configuring the conditional advertising option, but that requires a backup BGP peer. Is that peer existing ?

[balaji.bandi]

in your diagram does the R3 have another backup link ?

suggest to run EEM on R2 - when the Link come up and BGP come up you can run the EEM again to bring up Interface and BGP between R2 and R3

[MHM Cisco World]

this is my lab, the R1 must detect R2-R3 link failed to remove R2 neighbor from BGP router. 

[Joseph W. Doherty]

As @paul driver also describes, basically, a configured dynamic routing topology (e.g. BGP), ideally, would withdraw routes for using a "broken" path.

However, if, for example, static routes (and/or some aggregate routes) are being injected into a dynamic routing protocol, they often are not withdrawn when a path is broken.

If that's the situation, you might review how routes are managed, and determine if they can reflect actual availability.  If not, a solution like @balaji.bandi's, @Georg Pauwen's, @MHM Cisco World's might be used to "test" for reachability and if the test fails, modify either routes or the state of a peering relationship.

Of the two approaches, I would suggest trying to get your dynamic routing protocol to work the way it's intended to work, dynamically.