Back to Back VRF with redundant PEs

Unanswered Question
Jul 4th, 2008


Anyone done a back to back VRF with redundant PEs?

I am considering having eBGP between the PEs... have also checked the Multipath option, doesn't fit my case as I don't want to load balance... I want to use one PE as Primary and the other only as backup incase my Main PE fails... any unusual routing behaviour etc., to expect?





I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
n.nandrekar Sat, 07/05/2008 - 03:55

hello sultan!

Can you please elaborate on the scenario, with a small diagram perhaps? I am unable to understand the requirement. Are you talking about inter-as back-to-back vrf option? Or is it something else?



sultan-shaikh Sun, 07/06/2008 - 22:09

Hi Niranjan/Devang,

Thanks for your replies...

Here are the details:-

Redundant PE Configuration

1) PE A, B, 1 and 2 - Main Path

2) PE C, D, 4 and 3 - Backup, only used when Main goes down.

3) Suppose traffic is flowing on Main path and one of the NNI Main Back to Back VRF PE fails, how will the connection switch to Backup path what kind of redundant configuration is suggested, I am thinking about eBGP sessions over different VRFs on PE B and D and 1 and 2, and sharing the route targets.

4) How will the routing work in this scenario

I have done this on a non redundant PE, just want to know expert opinion on how redundant PE configuration will work.

Thanks for your replies...



devang_etcom Mon, 07/07/2008 - 07:31


assuming that you have same rd and route-target for VRF on link between PEB-PE1 and PED-PE4.

so here you have CE connected to PEA on site A and CE connected to PE2 and PE3 on site D.

as a normal behavior you will have two path to reach from one CE to other end CE. and that will be via main path and other is backup path. but BGP will put only one path as a best path as per is normal behavior right...

so if you will do multipath option then bgp will put both back up and main path in routing table.

so what i did in my lab is i created topology where

CE1 is connected with PEA and PEA is connected with PE1 and PE2... that is my site 1

CE2 is connected with PEB and PEB is connected with PE3 and PE4... that is my site 2...

now PE1 of site 1 is connected with PE3 or site 2 as well as PE1 and PE2 has link between them. same setup for PE2 and PE4.

so here I created vrf on PEA for customer, as you are talking about back-to-back vrf i created same vrf with same rd and rt on PE1 and PE2 and assigned that vrf to interface of PE1 and PE2 facing towards the PE3 and PE4... and did same thing for site 2...

now i am running eBGP on back to back vrf using ipv4 address family... and redistributed connected subnets...

PE1--eBGP for back to back vrf --PE3

PE2--eBGP for back to back vrf --PE4

now PEA, PE1 and PE2 are iBGP nei as they are in same AS1. and PEB, PE3 and PE4 are iBGP nei as they are in same AS2.

With this set up under "sh ip bgp vpnv4 all" under PEA and PEB i found two path to reach each other CE's subnets and that is from main path which is CE1-PEA-PE1-PE3-PEB-CE2 and back up path which is CE1-PEA-PE2-PE4-PEB-CE2...

and under "sh ip route vrf ce" on PEA and PEB i found main path in vrf routing table which is CE1-PEA-PE1-PE3-PEB-CE2. now if you shut down link between PE1-PE3 then you will find back up path in your vrf routing table... so this is the way you can achieve redundancy...

sorry for posting such a big post :-)... if it looks confusing to you then please feel free to ask me any question or query...


Devang Patel

sultan-shaikh Thu, 07/17/2008 - 22:40

Hi Devang/Niranjan,

Thanks for your response, I was doing this test and am facing some redundancy issues.

I tried but have been unable to finish it as I have hit upon some conceptual issues and am just unable to think any further... :)

I am attaching one more setup diagram which gives some more detail w.r.t. my configuration.

What I am thinking is a redundant configuration, I am running HSRP for each redundant customer on my PE1 and PE1a in Mumbai site and similarly my Partner will too run the same in Delhi at his PEs.

By Back to Back VRFs how can I achieve similar kind of Active/Backup configuration for each L3VPN configured across this interconnect.

The path over PE1 - PE2 - PE3 - PE4 is always Main, while the other is always Backup, (in my network ABC I have FRR configured, so has Provider XYZ), now what should I do to ensure that the path switches to Backup incase of Main NNI (Back to Back) link failure.

Should there be HSRP/VRRP or any other configuration between Back to Back connected interfaces?

I completed non PE redundant configuration, this redundant PE is proving to be a pain.

Has anyone done something similar to this?

Awaiting your response.




devang_etcom Sun, 07/06/2008 - 21:08

Hi Sultan,

It will be great if you can provide us more information about the topology... as you are talking about back to back VRF means are you talking about ASBR Routers facing towards the other AS or PE routers facing towards the CE routers...?



sultan-shaikh Wed, 07/09/2008 - 15:58

Hi Devang,

Thanks for your response and sorry for my late reply.

I have tested single PE back to back and it works, it is really simple, I wanted to get some response on real life scenarios/issues faced by our peers on this forum.

I will be testing this out sometime next week and will update the forum.

I will post some more details of the topology soon.

Really appreciate your response.





This Discussion