[quagga-users 13577] Re: BGP Routes from R4 to R1

Alexis Rosen quagga-users at alexis.users.panix.com
Tue Feb 18 21:46:48 GMT 2014


On Feb 17, 2014, at 3:34 AM, "BIRLASEKARAN, DINESH (DINESH)" <dinesh.birlasekaran at alcatel-lucent.com> wrote:
> Thank you for pointing out the basic iBGP full requirement.  Now I changed so that they all are in a mesh.  

Bah. I must be blind. As Jan pointed out, you're not running iBGP, and the full mesh isn't needed.

> Further to this.
> 
>> R1 is connected to R2 and R3
>> R2 is connected to R1 and R4
>> R3 is connected to R1 and R4
>> R4 is connected to R3 and R2.
> 
> R1 is also connected to R4
> R2 is also connected to R3 
> 
> To give it a full mesh config.

Let's assume that you revert to not doing this, since presumably the real-life scenario you want to implement won't have that.

> If I wanted the forward traffic to go via R1 to R2 to R4 and reverse via R4 to R3 to R1, which local preference should I change on every single router R1, R2, R3, R4?

Do you want all traffic from R1 to go out via R2, or just traffic to R4? And do you want just R4's traffic inbound to R1 to prefer R3, or all traffic? If you want this for all traffic, you should be padding AS paths, not playing with local prefs. But assuming it's just R1<->R4 traffic, then on R1:

> # Begin configuration for R2
> neighbor 10.91.33.63 remote-as 49
> neighbor 10.91.33.63 soft-reconfiguration inbound
> neighbor 10.91.33.63  description ISP R2
> neighbor 10.91.33.63  route-map rm_peer_1_in in
> neighbor 10.91.33.63  route-map rm_peer_1_out out

This is where you'll prefer R2 as a path to R4.

First of all, you don't need rm_peer_1_out. Did you think you need to prepend your AS on all sessions? You don't, BGP will automatically insert your AS into every route it announces.

Also, to prevent confusion on larger configs, I'd suggest changing the inbound map name:
> neighbor 10.91.33.63 route-map R2_bgp_in in

Now, change that route map like this:

> route-map R2_bgp_in permit 5
> ! match R4 routes, note RMs and PLs can have the same name
> match ip address prefix-list R4_routes
> set local-preference 200


Now all you need is the prefix-list:

ip prefix-list R4_routes seq 5 permit 10.91.51.0/24

If you have a lot of routes coming from R4, instead of filtering by ip address prefix list, you may find it easier to filter by AS number.

So, that's it. On R4 you can do the same thing. Now your only problem is that you can't see R4's routes from R1 in the first place, and I don't yet know why. After you tear down the new sessions between R2/R3 and R1/R4, if it's still not working, show the BGP route from R4 on R2, and also show ip bgp advertised-routes so we know what R2 thinks it's doing.

/a


More information about the Quagga-users mailing list