OSPF dual-homed to an MPLS SP backbone

systemasfabbro · ‎06-10-2013

Hello all, my name is Mariano from Buenos Aires, Argentina. I work at an ISP and I'm responsible for the integration of customers networks to the MPLS enabled backbone.

In this case, the topology is as follows:

PE1 - Loopback0 1.1.1.1

PE2 - Loopback0 3.3.3.3

RR - Route Reflector - Loopback0 2.2.2.2

Situation:

OSPF is running in this PE-CE topology. In PE's side, OSPF is enabled under the desired customer VRF. Customer network 10.84.56.0/24 is reachable from the MPLS backbone in its VRF:

PE1#sh ip ro vrf vpn_Int 10.84.56.0

Routing entry for 10.84.56.0/24

Known via "ospf 16", distance 110, metric 74, type intra area

Redistributing via bgp 11664

Advertised by bgp 11664 match internal external 1 & 2

Last update from 10.1.2.2 on Serial0/0, 00:10:49 ago

Routing Descriptor Blocks:

* 10.1.2.2, from 10.84.56.2, 00:10:49 ago, via Serial0/0

Route metric is 74, traffic share count is 1

PE2#sh ip ro vrf vpn_Int 10.84.56.0

Routing entry for 10.84.56.0/24

Known via "ospf 16", distance 110, metric 74, type intra area

Redistributing via bgp 11664

Advertised by bgp 11664 match internal external 1 & 2

Last update from 10.1.2.6 on Serial0/0, 00:13:04 ago

Routing Descriptor Blocks:

* 10.1.2.6, from 10.84.56.2, 00:13:04 ago, via Serial0/0

Route metric is 74, traffic share count is 1

So far, so good.

My question is regarding CE's routing table:

PUB_A#sh ip ro | b resort
Gateway of last resort is not set

     9.0.0.0/26 is subnetted, 1 subnets
O E2    9.9.9.0 [110/1] via 10.1.2.1, 00:14:01, Serial0/0
     10.0.0.0/8 is variably subnetted, 9 subnets, 3 masks
O       10.1.2.8/30 [110/65] via 10.1.2.1, 00:14:11, Serial0/0
O E2    10.49.51.0/25 [110/1] via 10.1.2.1, 00:14:01, Serial0/0
O E2    10.48.50.0/25 [110/1] via 10.1.2.1, 00:14:01, Serial0/0
C       10.1.2.0/30 is directly connected, Serial0/0
O       10.1.2.4/30 [110/74] via 10.84.56.2, 00:14:11, FastEthernet0/0
O       10.1.2.16/30 [110/75] via 10.84.56.2, 00:14:11, FastEthernet0/0
O E2    10.90.2.0/24 [110/1] via 10.84.56.2, 00:14:01, FastEthernet0/0
O E2    10.92.2.0/24 [110/1] via 10.1.2.1, 00:14:01, Serial0/0
C       10.84.56.0/24 is directly connected, FastEthernet0/0

=============

PUB_B#sh ip ro | b resort
Gateway of last resort is not set

     9.0.0.0/26 is subnetted, 1 subnets
O E2    9.9.9.0 [110/1] via 10.1.2.5, 00:15:26, Serial0/0
     10.0.0.0/8 is variably subnetted, 9 subnets, 3 masks
O       10.1.2.8/30 [110/75] via 10.84.56.1, 00:15:36, FastEthernet0/0
O E2    10.49.51.0/25 [110/1] via 10.84.56.1, 00:15:26, FastEthernet0/0
O E2    10.48.50.0/25 [110/1] via 10.84.56.1, 00:15:26, FastEthernet0/0
O       10.1.2.0/30 [110/74] via 10.84.56.1, 00:15:36, FastEthernet0/0
C       10.1.2.4/30 is directly connected, Serial0/0
O       10.1.2.16/30 [110/65] via 10.1.2.5, 00:15:36, Serial0/0
O E2    10.90.2.0/24 [110/1] via 10.1.2.5, 00:15:26, Serial0/0
O E2    10.92.2.0/24 [110/1] via 10.84.56.1, 00:15:26, FastEthernet0/0
C       10.84.56.0/24 is directly connected, FastEthernet0/0

For example, network 10.90.2.0/24 is part of the same customer VRF's (vpn_Int) and it's configured in Loopback5 in PE1. To reach this network from customer PUB_A router, it's next-hop is 10.84.56.2 (PUB_B), and then, it's routed to the MPLS backbone.

PUB_A#traceroute 10.90.2.1

Type escape sequence to abort.
Tracing the route to 10.90.2.1

1 10.84.56.2 36 msec 24 msec 36 msec
2 10.1.2.5 20 msec 28 msec 8 msec
3 10.90.2.1 [MPLS: Label 27 Exp 0] 36 msec * 36 msec

This behavior happens with other routes, but viceversa: PUB_B is pointing to PUB_A for other destinations (for example, 10.48.50.0/25). So I have some routes exiting from customer side via PUB_B and other networks via PUB_A.

How can I configure customer side to unify all reachable networks through MPLS backbone to the same exit router? (for instance, PUB_A)

What other considerations do I have to take into account?

This forum rocks!

Thanks a lot!

Mariano.

Vivek Ruhil · ‎06-10-2013

Hi Mariano

In OSPF, path selection is based on cost and lower the cost better is the route.

So, in order to influence the Customer network to use PUB_A for all networks in MPLS Backbone, you could increase the cost on the link between PUB_B and PE2. The command to do so is:

conf t

!

int ser0/0

ip ospf cost 10

!

This way you will instruct the routers in customer network to use lower cost path which is through PUB_A.

Also, you have created a document, if you want to ask a question in the community simply start a discussion rather than creating a document.

Thanks and let me know if you have more questions.

systemasfabbro · ‎06-10-2013

Hi,

Thanks for your reply! I've tested the config and it works OK!

Two more questions:

1) From service provider MPLS backbone perspective, should I configure something in PE's in order to adapt them to the customer side change?

2) Since this VRF contains several routes... I'm thinking in making the area totally stub in order not to increase CPU load on customer routers. What do you think? Is it recommended routing all traffic via a default route to the MPLS backbone?

Thanks.

Mariano.

Vivek Ruhil · ‎06-10-2013

Hi Mariano

1. As a best practice, you should always have a controlled redistribution between customer routing process and SP process, reason being due to an unwarranted change in the customer routing which leads to a routing spam SP will be affected. By doing controlled redistribution you will be able to handle that.

2. You could make the area totally stub but then you should make sure that the customer only has one exit path out of each site, i.e. towards the MPLS cloud. By the way, which router model are you using and how huge is the routing table ?

systemasfabbro · ‎06-11-2013

Hi, this is the output of the real and production enviornment of the VRF:

Customer side router: 4948.

SP side: 7609

systemasfabbro · ‎06-11-2013

All,

We were able to solve the routing problem creating an OSPF sham-link between PE's. This way, all traffic takes the PUB_A - PE1 link, and PUB_B - PE2 link remains as backup.

Thanks to all!

Mariano.