[nsp] ospf default routes and bgp injection

[Lim Fung]

Christopher J. Wolff wrote:

> Hello,
> 
> In the example below, Router A receives a full routing table via BGP.
> Router A, Router B and Router C are members of ospf area 0.  There are no
> static routes in this network at this time.
> 
> ISP---Router A---Router B---Router C
> 
> If I understand correctly, Router C can connect to the ISP through Router B
> if one of the following conditions are met:
> 
> 1. Router A redistributes its BGP table into the ospf area 0.
> 2. Router B advertises its default route to router C using
> default-information originate.
> 3. Router A advertises 0.0.0.0 to all members of area 0 since its
> configuration specifies default-information originate always.
> 
> Accordingly,
> 
> If we choose option 1, Router A redistributes its BGP table into the ospf
> area, all routers need to have magnificent amounts of memory.

You are better off running iBGP on all the routers, instead of 
redistributing BGP into OSPF. Your IGP is supposed to be kept small and 
clean, and is supposed to carry only the infrastructure prefixes.

Redistributing BGP to your IGP is a general no-no :)

> With option 2, If router B advertises its default route, it becomes default
> gateway for all current and new ospf area 0 members, which is not desirable.

Assuming that that is the only egress path, it should not be an issue.

> With option 3, If router A advertises 0.0.0.0, this example would be broken
> when I add router D and E to area 0, right?
> 
> ISP-A/BGP---Router A---Router B---Router C
> 		| (igp)
> ISP-B/BGP---Router D---Router E
> 
> Any advice will be appreciated.  Thank you for your time.

Assuming that you run iBGP between Router A and Router D, it should not 
matter, since even if both Router B and E defaults to Router A and D 
respectively, the more specific routes learnt via eBGP (propagated via iBGP) 
will ensure that the shortest path to the destination is used.

regards,
Lim Fung
--