07-07-2009 01:25 PM - edited 03-04-2019 05:21 AM
Is it possible to associate object tracking IP SLA with BGP conditional advertisement?
If I only use the BGP conditional advertisement then I will still need to wait for the route timeout but the user application can't wait that long so I am wondering if anyway to associate the object tracking with conditional advertisement.
Or if there is any work around to withdraw the BGP route when remote route not reachable immediately?
Solved! Go to Solution.
07-08-2009 08:59 AM
The problem I've seen with MPLS and CSC is broken labels in the path.
The BGP session (PE<->PE) will remain up and the routes will be in the table but there is a blackhole. Nasty scenario.
What I've seen done to workaround this dilemma is to advertise loopbacks from the CE to the MPLS Backbone via BGP and create another BGP peer CE<->CE with those loopbacks. As soon as the loopbacks aren't reachable, the BGP session will go down.
HTH,
__
Edison.
07-08-2009 05:38 AM
Hello Frank,
I'm afraid this is not possible however, using aggressive timers on the primary BGP session and neighbor ... fall-over should help to minimize the detection time.
By default in modern IOS images ebgp fast external fallover is enabled and as soon as the outgoing interface is detected down the session is torned down and this should trigger also BGP conditional advertisement
Hope to help
Giuseppe
07-08-2009 05:49 AM
Giuseppe,
Thanks for the reply.
We are facing some outage inside the MPLS cloud and is not local circuit outage and the customer want to fail to their backup circuit ASAP instead of waiting for BGP timeout... We have tuned the BGP advertisement interval but still not good enough for their apps.
So we are trying to use some tracking mechanism to re-route the traffic ASAP.
I am thinking to use conditional injection instead of advertisement and will test it in our lab...
07-08-2009 08:14 AM
Hi...
Have you looked into BGP BFD (Bi-directional Forwarding Detection) ?
Thanks...
Vyto
07-08-2009 08:38 AM
Yes, we think about that but the PE from carrier doesn't allow the BFD and it is not local outage but someing inside the cloud so we are looking for end-to-end tracking to influence the routing.
07-08-2009 08:59 AM
The problem I've seen with MPLS and CSC is broken labels in the path.
The BGP session (PE<->PE) will remain up and the routes will be in the table but there is a blackhole. Nasty scenario.
What I've seen done to workaround this dilemma is to advertise loopbacks from the CE to the MPLS Backbone via BGP and create another BGP peer CE<->CE with those loopbacks. As soon as the loopbacks aren't reachable, the BGP session will go down.
HTH,
__
Edison.
07-08-2009 09:19 AM
Edison,
Yeap, what a coincident that one of our guys talk about GRE end-to-end then we go further for EBGP multi-hops end-to-end and tune the timer very low...
I think it is a better way instead of doing tracking with conditional injection.
I will test that in our lab then propose to the client.
Thanks.
07-08-2009 09:37 AM
GRE is another approach but you will face the common issues associated with GRE, fragmentation and such.
07-08-2009 09:31 AM
Hello Edison,
very good idea!
Best Regards
Giuseppe
07-08-2009 09:39 AM
Giuseppe,
grazie :)
07-08-2009 08:43 AM
Frank,
This has to do with Ebgp updates, Object tracking is not applicable here neither Fast external failover feature.
The Conditional Advertisment relies on the BGP routing table, therfore, relies on the ebgp routing update which is 30sec by default.
If you need to have rapid routing convergence, you could change the BGP Scanning time for that update although its not recommended as this could seriously impact your memory resources of the router.
HTH
Mohamed
07-08-2009 09:06 AM
Edison,
Good that you have observed that, At least the LSPs , CSC , and PE-PE.
looks like you have been tackling the passed CCIS service provider track....
Regards,
Mohamed
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: