[j-nsp] Troubleshooting BGP

[Kevin Oberman]

> From: Jason Lixfeld <jason at lixfeld.ca>
> Date: Thu, 8 Oct 2009 09:07:41 -0400
> Sender: juniper-nsp-bounces at puck.nether.net
> 
> 
> On 2009-10-07, at 10:24 PM, Stefan Fouant wrote:
> 
> > Yep, the OSPF route is preferred, you can see that in your display by
> > looking at the * symbol next, which indicates which route is  
> > active.  Also
> > you can see that the BGP route has an "Inactive reason: Route  
> > Preference",
> > which makes sense in this case because as you indicate the default
> > preference value for OSPF externals is lower than BGP routes.  By  
> > default,
> > BGP will only advertise active BGP routes.  You'll need to either  
> > put in a
> > policy to redistribute the OSPF route to BGP, or use the advertise- 
> > inactive
> > command to advertise the inactive BGP route.
> >
> > -- 
> > Stefan Fouant
> 
> We decided that instead of using advertise-inactive, we'd instead just  
> prevent this prefix from being redistributed into OSPF in the first  
> place.  It's a customer prefix, so there's really no reason why it  
> would need to be in our IGP.

I think this is the right answer. We have always kept IGP and BGP routes
entirely separate and it has prevented a wide variety of problems. In
general, keep RIBs as separate as possible. If it's in BGP, don't put it
in OSPF/ISIS and vice-versa.

Sites may, in some cases, find it desirable to keep all routes in both
protocols, but providers probably want to keep the IGP very light with
only backbone addresses and have filters to make sure that the routes
don't leak form one to the other.
-- 
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman at es.net			Phone: +1 510 486-8634
Key fingerprint:059B 2DDF 031C 9BA3 14A4  EADA 927D EBB3 987B 3751