BGP <-> OSPF default route.

[tomek0001]

<p><br />

I am trying to figure out what would be the best way to provide redundant Internet access across two sites running BGP full routing tables at the edge and OSPF between the sites. My main goal is to be able to have fail-over to the other site's Internet access in case the Internet circuit goes down. <br />

<br />

Since I won't be getting a default route from the ISP I can't redistribute that default route to OSPF. One way is to just originate a default route from the OSPF process running on the edge provider, but my question is how to take it away if the Internet circuit fails and reroute it to the other site. <br />

<br />

Topology:<br />

(ISP1 ) - Site A (BGP-OSPF)– OSPF Backbone - Site B (OSPF-BGP) - (ISP 2)<br />

<br />

If the connection to ISP1 fails, how to I reroute uses from Site A to use Site B's Internet access?<br />

<br />

I looked at using “default-information originate always route-map DEPENDENT-ROUTE”  where my dependent-routes would be something like a RR route from the ISP or the ip address of the interface connected to the ISP.3<br />

<br />

What do you think is the best way to accomplish this?<br />

<br />

Is there anything within iBGP I should look at?<br />

<br />

<br />

Thank you.</p>

[Giuseppe Larosa]

Hello Tomasz,

I did this in the past the command default-originate uses the route-map as an alternative option to alwasys

router ospf 10

default-information originate route-map check-bgp-routes metric-type 1

in the route-map you can:

set the metric type to O E1 (recommended if you have multiple exit points)

set a seed metric

check the presence of a network

match ip address

check the BGP next hop of the route with

match ip next-hop acl-number

if you haven't an iBGP full mesh you can use the method described above.

Hope to help

Giuseppe

[tomek0001]

Hello,

What could you do if you had a full mesh iBGP?

Thank you,

[paolo bevilacqua]

Well... in reality full mesh iBGP is necessary only when doing transit, or otherwise wanting full control on any aspect of your routing.

Understandably, with just the need to inject certain specific routes into igp for optimized exit, you can get away without.

Regarding your specific case, I'm not sure if would help, as the suggestions given above should be enough. The conditional to route map is also a very good one.

[paolo bevilacqua]

Hi, if you point the default route to next hop or interface used for BGP, if the circuit fails, it will go away.

If you're concerned that the circuit can stay up but no connectivity, you can use SLA to monitor an object and have the route track that.

The alternative is taking full tables and not have a default.

You should have already iBGP across your BGP routers.