Routing

Hi,

Here is a topology: R1(e1/1) ----------(e1/1) R2 (e1/0) ---------(e1/0) R3

R1' AS is 1, R2' AS is 2 and R3' AS is 3

e1/0 in R2 and R3 are vrf DOC. e1/1 in R2 and R1 are normal interface

Can you give some suggestion how to configure the bgp between R1 and R2 in order for traffic to go from R3 to R1? Thank you

Hi,

______________________________________________________________________________

e1/0 in R2 and R3 are vrf DOC. e1/1 in R2 and R1 are normal interface

______________________________________________________________________________

Do you mean e1/1 in R1 and R2 are in inet.0 [default instance].

I see the question is more or less same as here.

IMO, solutions are same too, i.e one of the available methods to exchange routes between vrf on R2 to inet.0 on R2.

To summarize:

1. if vrf to vrf route exchange: auto-export

2. vrf/virtual-router to inet.0 / virtual router : 

    a. rib-groups

    b. lt-interface - Requires dynamic routing protocol to exchange routes between the 2.

My preferred option is rib-groups for scenario 2 which is only sharing routes between 2 tables. This may require interface routes to be exchanged as well for traffic forwarding.

lt- interface may introduce some complexities with dynamic routing protocols. For instance if BGP, additional ASN in as-path, in case this is ISP transit peering.  If IGP, other existing IGP in the design needs to be considered,

Cheers,

Ashvin

Thank you so much for your reply. I get confused when I see a lot juniper command because I am new to Juniper.

In Cisco device, my understanding is two vrf route-targe exchange so that the traffic can pass the two vrf. But in my case, it  is one vrf and the other is normal one(global). that is, R1 and R3 are Cisco device and R2 is Juniper MX960. 

Based on Cisco device, we need to know what is global vrf value? If we know it, we can just exchange route-target between the two vrf. am I right? 

From your query , looks like the purpose is to leak the routes from vrf instance to global instance .

If yes , pls do it with rib groups.and i hope it is one way(from vrf to global ) leak

The config looks something like this:

set routing options rib-groups VRF TO GLOBAL import-rib vpn1.inet.0 import-rib  inet.0

(VPN1 indicate vrf name please replace it  with your vrf name)

under routing instance :

set routing-instance vpn1 protocol bgp group x family inet-unicast rib-group VRF TO GLOBAL

You can use the below link for reference:

https://mellowd.co.uk/ccie/?p=3691

Hi,

________________________________________________________________________________

In Cisco device, my understanding is two vrf route-targe exchange so that the traffic can pass the two vrf

________________________________________________________________________________

AFAIK, in Cisco 2 vrfs exchange routes by default if the required route-targets are being imported.

In Juniper "auto-export" needs to be enabled on the participating vrfs for the route exchange between them:

set routing-instances xxx routing-options auto-export

___________________________________________________________________

Based on Cisco device, we need to know what is global vrf value?

___________________________________________________________________

I am not sure there is any route-target [vrf value] for the global unicast routing table even on Cisco or Juniper. Please share if you have some info on this.

There is a caveat with scenario 2 [vrf MP-BGP l3vpn routes to inet.0] using rib-groups. Rib-groups can export PE-CE routes from the vrf to inet.0, but not MP-BGP l3vpn routes learnt from a remote PE. This is because of different BGP address families.

To export vrf l3vpn routes to inet.0, you could use the lt interface to bridge the vrf to inet.0.

How are the routes being learnt on R2 from R3. Is that over an MPLS L3VPN PE-PE?

Could you share your sample configs to have a better understanding.

Cheers,

Ashvin

Hi We already changed topology. So the issue have been solved. Thank you all