Stop a route from being advertised via BGP

[graham smart]

Hi Guys,

Quick Q

I have BGP running ontop of my OSPF igp.

Now..

OSPF is advertising its loopback and uplink routes to the network and BGP advertises connected and static routes.

Because BGP is advertising the same loopback and uplink IP's as OSPF, i get RIB failures in the BGP routing table for those ranges.

This doesnt cause an issue because OSPF has the routes as expected.

But for neatness id like to block a few routes from being advertised via BGP..

Whats the best way to do this? / Best practise for this?

Or shall I just live with 3 routes in my BGP table downstream that are rib failures?

Obviously its pointless having them in BGP anyway as if OSPF dies BGP would also so its not like theres any resillience here.

Thoughts?

-G

[Peter Paluch]

Hello Graham,

My personal opinion: Having routes in BGP marked as rib-failure is quite normal if you perform redistribution from OSPF into BGP, making the BGP basically carry the same set of routes as OSPF. Filtering them out is possible on various stages - you could prevent these routes from being redistributed into BGP in the first place (which would be the best thing to do), or you could filter them out using prefix-lists or other filtering mechanisms on a per-neighbor basis.

However, please note that if you have an eBGP peering with other ASes and you prevent these "offending" routes from being injected into BGP, you will not be able to advertise them to other ASes, as they won't be advertised via BGP at all.

So removing these routes from BGP and not having them marked as rib-failures depends on what your routing policy actually is. It may be a bit offensive to the general sense of network aesthetics (as Jeff Doyle would eloquently put it ) but in case of external BGP peerings, it is necessary - and tolerated.

I hope Milan Kulik and other BGP gurus will offer their own views on this issue!

Best regards,

Peter

[graham smart]

Hi,

Thanks for the info.

I have the main supernets configured on our border routers so the ranges still get advertised externally ( even though these ranges would never be connected to via an external AS )

This is merely to remove them from the BGP table as rib failures annoy me

-Graham

[Peter Paluch · Accepted solution]

Hi Graham,

This is merely to remove them from the BGP table as rib failures annoy me  

Well, in that case, if there is no point in having these networks advertised in BGP at all, I suggest not injecting them into BGP in the first place. If you are injecting them via the network command then simply remove it from appropriate routers. If you are using redistribution, use route-maps to select which networks should be redistributed into BGP.

Best regards,

Peter

[graham smart]

Currently its picking the routes up because of the redistribute-connected & static ( Got loads of other connected networks etc that need to be in bgp )

Ive setup a prefix-list to control these now so they have gone.

Thanks for your time

At least I know im not being completely stupid wanting to get these off my BGP table. Must be my "OCD" coming out.

[Peter Paluch]

Hi Graham,

Thanks for your time 

You are sincerely welcome!

At least I know im not being completely stupid wanting to get these off my BGP table. Must be my "OCD" coming out. 

Oh, don't worry about that I know the feeling myself

Best regards,

Peter