How to manage MP-BGP relationships with MPLS

[John Peterson]

Hi, I'm in the process of adding a third customer site to our MPLS core. The first two sites work perfect with the MP-BGP neighbor commands pointing to each other.

The client has requested a third site to be added to the MPLS. I have configured the vrf etc... but when setting up the neighbors should the statement point to each of the other two neighbours and vice-versa creating a full mesh? If so, how would you manage a multiple site MPLS as sites would increase for a customer the management would become a nightmare?

Thanks.

address-family vpnv4
  neighbor 10.0.21.3 activate

[blau grana · Accepted solution]

Hi John

When adding new site for customer, CE will peer only with PE to which it will be connected.

In your MPLS backbone you need to distribute vpnv4 routes to every PE, you can do it by creating full mesh (which with many PEs will be almost impossible) or you can use route reflectors (PEs will be peering only with RR, so no full mesh will be required)

You can use also bgp confederations, but I think that ISPs usually use primary RR and secondary RR.

some links:

http://tinyurl.com/b7h2on7

https://supportforums.cisco.com/thread/250252

[John Peterson]

Ah perfect, I was under the impression that the RR were only for iBGP relationship.

Sent from Cisco Technical Support iPhone App

[blau grana]

I am not sure if I understand you correctly.

I supose that your core network consists of one ASN, so every PE has same AS number. If your network consists of many PE, you need to configure RR to limit number of iBGP sessions.

I can not imagine reason why use RR for eBGP peers because eBGP peers do not suffer same symptoms as iBGP peers (iBGP learned prefix is not advertise to another iBGP peer).

I do not say that it is not possible or no reason exist, I just can not think of any...

[John Peterson]

Thanks when configuring RR I keep on getting,
% Activate the neighbor for the address family

Sent from Cisco Technical Support iPhone App

[blau grana]

There is sample config in links which I provided to you earlier.

To configure RR in your backbone you can configure peer group which will shortens configuraition if number of PEs wil be increasing.

router bgp 100
 neighbor PEER_GROUP peer-group
 neighbor PEER_GROUP remote-as 100
 neighbor PEER_GROUP update-source Loopback0
 no auto-summary
 !
 address-family vpnv4
  neighbor PEER_GROUP send-community both
  neighbor PEER_GROUP route-reflector-client
  neighbor 2.2.2.2 peer-group PEER_GROUP
 exit-address-family
This way only vpnv4 prefixes will be reflected.

Best Regards

Please rate helpful posts

[John Peterson]

Thank you.

It seemed that for some reason address-family ipv4 got added in. Once I removed this and configured your method above it all worked.

We currently have 4 routers in our core, very small. We are planning to have more pop areas shortly. I would be grateful if you could help on the below:

# We want to also provide Internet connectivity to our customers via the MPLS core, is it best practice to keep your MPLS core and Internet core different?

# I guess you could configure RR under 'router bgp 100' when you want to push out networks from eBGP or other BGP network, is this correct?

# Is we were to configure RR under 'router bgp 100' and address-family vpn4 would there be an issue?

Thanks

[blau grana]

Hi John, glad I could help.

address-family ipv4 is used for exchanging ipv4 prefixes.

1)

mpls has many advantages, one of them is to provide different services on one infrastructure. you can provide internet connection to customers on same boxes as vpns, for example: on global table you will keep public prefixes and in vrfs you will keep private prefixes of different customers.

there are many design scenarios how you can do it

2)

here is configuration guide for RR:

http://www.cisco.com/en/US/docs/ios/12_0/np1/configuration/guide/1cbgp.html#wp5155

here is some case study:

http://www.cisco.com/en/US/tech/tk365/technologies_tech_note09186a00800c95bb.shtml#routereflectors

Also I recommend this book, a lot of useful info and examples:

http://www.amazon.com/BGP-Design-Implementation-Randy-Zhang/dp/1587051095/ref=sr_1_2?s=books&ie=UTF8&qid=1360789918&sr=1-2&keywords=bgp

3) configuring RR under vpnv4 family will cause only reflecting vpnv4 prefixes, it means distributing only private prefixes of your customers

if you want to reflect also ipv4 prefixes (f.e. public BGP table) you need to configure RR under ipv4 family

Best Regards

Please rate helpful posts

[John Peterson]

Thanks, I'm reading on RR and it's seems the client of each cluster can only be one hop away from the RR server, is this correct?

Sent from Cisco Technical Support iPhone App

[Giuseppe Larosa]

Hello John,

iBGP sessions have IP TTL 255, there is no need to have the client one hop away from the RR server. They can be several hop aways if routers in the middle have the necessary knowledge of routes for setting up the BGP sessions.

Loopback IP addresses have to be published in an IGP like OSPF or ISIS to create a scalable solution.

MPLS services remove the problem of internal black holes inside an AS, you can have a BGP free core layer without any issues when MPLS provide LSPs between iBGP endpoint addresses.

Hope to help

Giuseppe