Creating an MPLS VPN

By stretch | Monday, May 16, 2011 at 1:17 a.m. UTC

Today we're going to look at the configuration required to create a basic MPLS VPN servicing two customers, each with a presence at two physical sites. If you're unfamiliar with the concepts of MPLS switching and VRFs on Cisco IOS, you may want to check out a few of my past articles before continuing:

Our lab topology looks like this:

topology.png

As a review, recall that

  • P (provider) routers are ISP core routers which don't connect to customer routers and typically run only MPLS
  • PE (provider edge) routers connect to customer sites and form the edge of a VPN
  • CE (customer edge) routers exist at the edge of a customer site; they have no VPN awareness
  • an IGP running among all P and PE routers is used to support LDP and BGP adjacencies within the provider network
  • MP-BGP is run only among PE routers
  • an IGP (typically) is run between each CE router and its upstream PE router

In our lab, OSPF is already in operation as the provider network IGP. OSPF processes have also been preconfigured on the CE routers; however, these OSPF topologies will remain separate from the provider OSPF.

There are five core tasks we need to accomplish to get an MPLS VPN up and running:

  1. Enable MPLS on the provider backbone.
  2. Create VRFs and assign routed interfaces to them.
  3. Configure MP-BGP between the PE routers.
  4. Configure OSPF between each PE router and its attached CE routers.
  5. Enable route redistribution between the customer sites and the backbone.

Although plenty of CLI outputs are shown below, you may want to grab the finished router configurations if you'd like to duplicate the lab on your own.

Enable MPLS

First we need to enable MPLS on all P-P and P-PE links with the mpls ip interface command. MPLS is not enabled on any CE-facing interfaces; CE routers do not run MPLS, just plain IP routing. LDP is enabled automatically as the default label distribution protocol (versus Cisco's legacy TDP). LDP typically runs between loopback addresses not directly reachable by LDP peers, which is why it's important to configure an IGP in the core before enabling MPLS.

We can verify the configuration of MPLS interfaces with show mpls interfaces.

P1(config)#    interface f0/1P1(config-if)# mpls ipP1(config-if)# interface f1/0P1(config-if)# mpls ipP1(config-if)# do show mpls interfacesInterface IP Tunnel OperationalFastEthernet0/1 Yes (ldp) No Yes FastEthernet1/0 Yes (ldp) No Yes 
P2(config)#    interface f0/1P2(config-if)# mpls ipP2(config-if)# interface f1/0P2(config-if)# mpls ip
PE1(config)#    interface f1/0PE1(config-if)# mpls ip
PE2(config)#    interface f1/0PE2(config-if)# mpls ip

LDP adjacencies can be verified with the command show mpls ldp neighbor:

P1#    show mpls ldp neighbor Peer LDP Ident: 10.0.0.2:0; Local LDP Ident 10.0.0.1:0TCP connection: 10.0.0.2.45114 - 10.0.0.1.646State: Oper; Msgs sent/rcvd: 12/13; DownstreamUp time: 00:02:43LDP discovery sources:FastEthernet0/1, Src IP addr: 10.0.9.2Addresses bound to peer LDP Ident:10.0.9.2 10.0.9.9 10.0.0.2 Peer LDP Ident: 10.0.0.3:0; Local LDP Ident 10.0.0.1:0TCP connection: 10.0.0.3.20327 - 10.0.0.1.646State: Oper; Msgs sent/rcvd: 12/12; DownstreamUp time: 00:02:25LDP discovery sources:FastEthernet1/0, Src IP addr: 10.0.9.6Addresses bound to peer LDP Ident:10.0.9.6 10.0.0.3 

Create and Assign VRFs

Our next step is to create customer VRFs on our PE routers and assign the customer-facing interfaces to them. We need to assign each VRF a route distinguisher (RD) to uniquely identify prefixes as belonging to that VRF and one or more route targets (RTs) to specify how routes should be imported to and exported from the VRF.

We'll use a route distinguisher for each VRF in the form of <ASN>:<customer number>. For simplicity, we'll reuse the same value as both an import and export route target within each VRF (though we are free to choose a different or additional route targets if we prefer). VRF configuration must be performed on both PE routers.

PE1(config)#    ip vrf Customer_APE1(config-vrf)# rd 65000:1PE1(config-vrf)# route-target both 65000:1PE1(config-vrf)# ip vrf Customer_BPE1(config-vrf)# rd 65000:2PE1(config-vrf)# route-target both 65000:2
PE2(config)#    ip vrf Customer_APE2(config-vrf)# rd 65000:1PE2(config-vrf)# route-target both 65000:1PE2(config-vrf)# ip vrf Customer_BPE2(config-vrf)# rd 65000:2PE2(config-vrf)# route-target both 65000:2

The command route-target both is used as a shortcut for the two commands route-target import and route-target export, which appear separately in the running configuration.

Now we need to assign the appropriate interfaces to each VRF and reapply their IP addresses. (Assigning an interface to a VRF automatically wipes it of any configured IP addresses. Your version of IOS may or may not inform you of this when it happens.) The command show ip vrf interfaces can be used to verify interface VRF assignment and addressing.

PE1(config)#    interface f0/0PE1(config-if)# ip vrf forwarding Customer_A% Interface FastEthernet0/0 IP address 10.0.1.1 removed due to enabling VRF Customer_APE1(config-if)# ip address 10.0.1.1 255.255.255.252PE1(config-if)# interface f0/1PE1(config-if)# ip vrf forwarding Customer_B% Interface FastEthernet0/1 IP address 10.0.1.5 removed due to enabling VRF Customer_BPE1(config-if)# ip address 10.0.1.5 255.255.255.252PE1(config-if)# ^ZPE1# show ip vrf interfacesInterface IP-Address VRF ProtocolFa0/0 10.0.1.1 Customer_A up Fa0/1 10.0.1.5 Customer_B up 
PE2(config)#    interface f0/0PE2(config-if)# ip vrf forwarding Customer_A% Interface FastEthernet0/0 IP address 10.0.2.1 removed due to enabling VRF Customer_APE2(config-if)# ip address 10.0.2.1 255.255.255.252PE2(config-if)# interface f0/1PE2(config-if)# ip vrf forwarding Customer_B% Interface FastEthernet0/1 IP address 10.0.2.5 removed due to enabling VRF Customer_BPE2(config-if)# ip address 10.0.2.5 255.255.255.252PE2(config-if)# ^ZPE2# show ip vrf interfacesInterface IP-Address VRF ProtocolFa0/0 10.0.2.1 Customer_A up Fa0/1 10.0.2.5 Customer_B up 

Configure MP-BGP on the PE Routers

This is where things start to get interesting. In order to advertise VRF routes from one PE router to the other, we must configure multiprotocol BGP (MP-BGP). MP-BGP is a little different from legacy BGP in that it supports multiple address families (e.g. IPv4 and IPv6) over a common BGP adjacency. It also supports the advertisement of VPN routes, which are longer than normal routes due to the addition of a 64-bit route distinguisher (which we assigned under VRF configuration).

MP-BGP runs only on the PE routers: P routers rely entirely on the provider IGP and MPLS to forward traffic through the provider network, and CE routers have no knowledge of routes outside their own VRF.

Minimal MP-BGP configuration is pretty straightforward. Both PE routers exist in BGP AS 65000.

PE1(config)#    router bgp 65000PE1(config-router)# neighbor 10.0.0.4 remote-as 65000PE1(config-router)# neighbor 10.0.0.4 update-source loopback 0PE1(config-router)# address-family vpnv4PE1(config-router-af)# neighbor 10.0.0.4 activate
PE2(config)#    router bgp 65000PE2(config-router)# neighbor 10.0.0.3 remote-as 65000PE2(config-router)# neighbor 10.0.0.3 update-source loopback 0PE2(config-router)# address-family vpnv4PE2(config-router-af)# neighbor 10.0.0.3 activate

If we look at the running configuration of the BGP process on either PE router, we notice that a bit more configuration than we provided has appeared:

PE1#    show running-config | section router bgprouter bgp 65000no synchronizationbgp log-neighbor-changesneighbor 10.0.0.4 remote-as 65000neighbor 10.0.0.4 update-source Loopback0no auto-summary!address-family vpnv4neighbor 10.0.0.4 activateneighbor 10.0.0.4 send-community extended  exit-address-family!address-family ipv4 vrf Customer_B  no synchronizationexit-address-family!address-family ipv4 vrf Customer_A  no synchronizationexit-address-family

In addition to our VPNv4 address family, address families for the two customer VRFs have been created automatically. Also, support for extended community strings has been added to the VPNv4 neighbor configuration.

Verify that the MP-BGP adjacency between PE1 and PE2 was formed successfully with the command show bgp vpnv4 unicast all summary:

PE1#    show bgp vpnv4 unicast all summaryBGP router identifier 10.0.0.3, local AS number 65000BGP table version is 1, main routing table version 1Neighbor V AS MsgRcvd MsgSent TblVer InQ OutQ Up/Down State/PfxRcd10.0.0.4 4 65000 12 12 1 0 0 00:06:05 0

Currently, there are no routes in the BGP table, because we have not specified anything to be advertised or redistributed, but we'll get to that after this next step.

Configure PE-CE OSPF

We just configured MP-BGP between the two PE routers. Now, let's configure an IGP between each PE router and its attached CE routers to exchange routes with the customer sites. We're going to use OSPF for this lab, but we could just as easily use another IGP like EIGRP or RIP.

Single-area OSPF has already been configured on the CE routers; all CE interfaces are in area 0. Remember that although we're using OSPF between each of the CE routers and its upstream PE router, these OSPF processes are isolated from the provider OSPF topology. The overall routing topology will look like this:

routing_topology.png

The provider OSPF process has already been configured on the PE routers as process 1. We'll configure an additional OSPF process for each CE router on each PE router. Each PE router will then have three OSPF processes total: one for the provider network, and one for each CE router. Whereas the provider OSPF process exists in the global routing table, the two CE processes will each be assigned to their respective customer VRFs.

PE1(config)#    router ospf 2  vrf Customer_APE1(config-router)# router-id 10.0.1.1PE1(config-router)# interface f0/0PE1(config-if)# ip ospf 2  area 0PE1(config-if)# router ospf 3  vrf Customer_BPE1(config-router)# router-id 10.0.1.5PE1(config-router)# interface f0/1PE1(config-if)# ip ospf 3  area 0
PE2(config)#    router ospf 2  vrf Customer_APE2(config-router)# router-id 10.0.2.1PE2(config-router)# interface f0/0PE2(config-if)# ip ospf 2  area 0PE2(config-if)# router ospf 3  vrf Customer_BPE2(config-router)# router-id 10.0.2.5PE2(config-router)# interface f0/1PE2(config-if)# ip ospf 3  area 0

We should see each PE router form an OSPF adjacency with both of its attached CE routers, and the customer routes should appear in the VRF tables on the PE routers.

PE1#    show ip route vrf Customer_ARouting Table: Customer_A...172.16.0.0/16 is variably subnetted, 2 subnets, 2 masksO 172.16.1.0/24 [110/11] via 10.0.1.2, 00:04:21, FastEthernet0/0O 172.16.0.1/32 [110/11] via 10.0.1.2, 00:04:21, FastEthernet0/010.0.0.0/30 is subnetted, 1 subnetsC 10.0.1.0 is directly connected, FastEthernet0/0PE1# show ip route vrf Customer_BRouting Table: Customer_B...172.17.0.0/16 is variably subnetted, 2 subnets, 2 masksO 172.17.1.0/24 [110/11] via 10.0.1.6, 00:03:03, FastEthernet0/1O 172.17.0.1/32 [110/11] via 10.0.1.6, 00:03:04, FastEthernet0/110.0.0.0/30 is subnetted, 1 subnetsC 10.0.1.4 is directly connected, FastEthernet0/1

Configure Route Redistribution

We're almost done! We have our MPLS and MP-BGP backbone up and running, and our CE routers are sending routes to our PE routers within their VRFs. The last step is to glue everything together by turning on route redistribution from the customer-side OSPF processes into MP-BGP and vice versa on the PE routers.

First we'll configure redistribution of CE routes in each VRF into MP-BGP. This is done under the BGP IPv4 address family for each VRF.

PE1(config)#    router bgp 65000PE1(config-router)# address-family ipv4 vrf Customer_APE1(config-router-af)# redistribute ospf 2PE1(config-router-af)# address-family ipv4 vrf Customer_BPE1(config-router-af)# redistribute ospf 3
PE2(config)#    router bgp 65000PE2(config-router)# address-family ipv4 vrf Customer_APE2(config-router-af)# redistribute ospf 2PE2(config-router-af)# address-family ipv4 vrf Customer_BPE2(config-router-af)# redistribute ospf 3

This enables redistribution of OSPF routes into BGP for transport across the provider network between the two sites. We can verify that the routes learned from the customer sites (the 172.16.0.0/16 and 172.17.0.0/16 networks) now appear in the BGP tables for their respective VRFs.

PE1#    show ip bgp vpnv4 vrf Customer_A...Network Next Hop Metric LocPrf Weight PathRoute Distinguisher: 65000:1 (default for vrf Customer_A)*> 10.0.1.0/30 0.0.0.0 0 32768 ?*>i10.0.2.0/30 10.0.0.4 0 100 0 ?*> 172.16.0.1/32 10.0.1.2 11 32768 ?*>i172.16.0.2/32 10.0.0.4 11 100 0 ?*> 172.16.1.0/24 10.0.1.2 11 32768 ?*>i172.16.2.0/24 10.0.0.4 11 100 0 ?PE1# show ip bgp vpnv4 vrf Customer_B...Network Next Hop Metric LocPrf Weight PathRoute Distinguisher: 65000:2 (default for vrf Customer_B)*> 10.0.1.4/30 0.0.0.0 0 32768 ?*>i10.0.2.4/30 10.0.0.4 0 100 0 ?*> 172.17.0.1/32 10.0.1.6 11 32768 ?*>i172.17.0.2/32 10.0.0.4 11 100 0 ?*> 172.17.1.0/24 10.0.1.6 11 32768 ?*>i172.17.2.0/24 10.0.0.4 11 100 0 ?

The last step is to complete the redistribution in the opposite direction: from BGP into the customer OSPF processes. If you're accustomed to route redistribution, there's nothing new here. (We don't have to specify any VRF information in the redistribution statement because each customer OSPF process is already assigned to a VRF.)

PE1(config)#    router ospf 2PE1(config-router)# redistribute bgp 65000 subnetsPE1(config-router)# router ospf 3PE1(config-router)# redistribute bgp 65000 subnets
PE2(config)#    router ospf 2PE2(config-router)# redistribute bgp 65000 subnetsPE2(config-router)# router ospf 3PE2(config-router)# redistribute bgp 65000 subnets

Testing and Confirmation

If has gone well, we should now have end-to-end connectivity between the CE routers within each VRF. Both routers for each customer should now have complete routing tables. Here are customer A's routes:

CE1A#    show ip route...172.16.0.0/16 is variably subnetted, 4 subnets, 2 masksC 172.16.1.0/24 is directly connected, Loopback1C 172.16.0.1/32 is directly connected, Loopback0O IA 172.16.2.0/24 [110/21] via 10.0.1.1, 00:03:50, FastEthernet0/0O IA 172.16.0.2/32 [110/21] via 10.0.1.1, 00:03:50, FastEthernet0/010.0.0.0/30 is subnetted, 2 subnetsO IA 10.0.2.0 [110/11] via 10.0.1.1, 00:03:50, FastEthernet0/0C 10.0.1.0 is directly connected, FastEthernet0/0
CE2A#    show ip route...172.16.0.0/16 is variably subnetted, 4 subnets, 2 masksO IA 172.16.1.0/24 [110/21] via 10.0.2.1, 00:02:49, FastEthernet0/0O IA 172.16.0.1/32 [110/21] via 10.0.2.1, 00:02:49, FastEthernet0/0C 172.16.2.0/24 is directly connected, Loopback1C 172.16.0.2/32 is directly connected, Loopback010.0.0.0/30 is subnetted, 2 subnetsC 10.0.2.0 is directly connected, FastEthernet0/0O IA 10.0.1.0 [110/11] via 10.0.2.1, 00:02:49, FastEthernet0/0

You may notice that OSPF routes sent between two sites belonging to the same customer appear as inter-area routes. Remember that although OSPF area 0 is being used at both sites, each site exists as a separate link-state topology connected by the MPLS VPN.

We should be able to ping from one CE router to the other. (Remember that we don't need to specify a VRF when doing so because CE routers have no knowledge that they're in a VRF.)

CE1A#    ping 172.16.0.2Type escape sequence to abort.Sending 5, 100-byte ICMP Echos to 172.16.0.2, timeout is 2 seconds:!!!!!Success rate is 100 percent (5/5), round-trip min/avg/max = 12/21/32 ms

We can perform a traceroute to verify the path taken as well as the MPLS labels used to traverse the provider network.

CE1A#    traceroute 172.16.0.2Type escape sequence to abort.Tracing the route to 172.16.0.21 10.0.1.1 4 msec 4 msec 8 msec2 10.0.9.5 [MPLS: Labels 19/22 Exp 0] 16 msec 12 msec 24 msec3 10.0.9.2 [MPLS: Labels 19/22 Exp 0] 24 msec 20 msec 16 msec4 10.0.2.1 [MPLS: Label 22 Exp 0] 20 msec 16 msec 24 msec5 10.0.2.2 16 msec * 36 msec

Here's a packet capture of the above traceroute if you're interested in how the MPLS label information is returned. And again, here are the the finished router configurations if you'd like to replicate the lab yourself.

(Thanks to Ivan Pepelnjak of Cisco IOS Hints helping revise this article!)