BGP redistribution problem between two sites

[taelon_x7]

I have attached a diagram of what I am trying to explain.

We have two locations, both run an OSPF area 0 that is redistributed to each other by BGP. One location (Site 1) has a WAN aggregation point for all of our remote offices and it runs EIGRP which is redistributed into OSPF. Everything works as it should right now. A WAN route at Site 1 gets redistributed into OSPF and then is redistributed into BGP and forwarded to Site 2 which in turn redistributes the route into its OSPF area 0.

The issue we are having is when a WAN site goes down. For instance, say that one of the remote sites (10.0.0.1) loses its T1 connection for a while, EIGRP stops advertising this route into OSPF. The problem then starts because BGP at site 2 is redistributing that route into OSPF so it thinks that the route is still valid and when the remote site comes back online; site 1 has a route for 10.0.0.1 with an AD of 20 because it is learning it from Site 2. The only way to resolve the issue is to clear OSPF at Site 2 so it will stop advertising the route back to Site 1.

So my question is how to fix it. I am not 100% sure that BGP sync will fix this problem. I am leaning more towards using AS Filters to ensure Site 2 never propagates routes to Site 1 that are Site 1's routes and vice-versa.

[andrew.prince]

Apart from the obvious issue of to many routing protocols, and a bad OSPF design; you have a redistribution loop, which is compounded with slow convergence issues.

What I would do is:-

1) Tune the EIGRP Hello & Hold Timers to be fast i.e 1 and 3

2) Configure a BGP Summary's of Site 1 and Site 2

HTH>

[taelon_x7]

Agreed that we need to eventually get rid of EIGRP and convert the remote sites to OSPF but I am not sure if I understand why you think this is a bad OSPF design.

[milan.kulik]

Hi,

so you are mutually redistributing BGP and OSPF on site 2?

If yes, what about some route tagging?

Using a route-map assigning a tag to a BGP prefix when redistributed to OSPF, e.g.?

And denying those tagged OSPF prefixes by another route-map when redistributing OSPF to BGP?

HTH,

Milan

[taelon_x7]

Thank you for the response! We actually saw the loop happen again yesterday for one of the remote sites and did verify that at site 2, OSPF and iBGP were redistributing the route and causing the loop. We implemented a route-map to block that AS tag to keep it from happening and it seemed to have resolved the issue.

The TAC engineer assigned to this case said all configurations looked fine. Wish I would have posted here sooner.

[Marwan ALshawi · Accepted solution]

Hi

This expected behavior with multiple point of redistribution called mutual redistribution

In ur case what u need is

Create route maps that tag all the route coming from site 1 when get redistributed into ospf of site 2

When u redistribute ospf into bgp back in site 2 deny the route u previously tagged

In this way u will never advertise the route back from site 2 to site 1

Good luck

Sent from Cisco Technical Support iPhone App

[andrew.prince]

A summary-only BGP Summary sent from site 1 to site 2, and the same type of summary from Site 2 to Site 1 should stop the need for route maps & tagging.

[Marwan ALshawi]

The summary will work

But in the case of route flapping as EIGRP is quicker in convergence time this summary might result in some route blackholing!!!

Unless the summary will point out to the local router when the most specific route withdrawn !!!!!

Sent from Cisco Technical Support iPhone App

[andrew.prince]

Route flapping can be controlled with IP Event Dampening or just adjust the EIGRP timers.