Manipulate traffic in OSPF

Unanswered Question
Sep 7th, 2010

Hello all,

I have come to a dead end and I would really appreciate your insight.Our network is running OSPF and we want to achieve symmetric routing with our branches (anything that comes through routerA should exit from routerA) as can be seen in the attached file. I know that the solution would be to run eBGP with the provider and with BGP attributes (AS prepend, local preference) manipulate the traffic, but my company wants to create an OSPF neighborship with the PEs. The incoming traffic will be routed by the providers MPLS, so no worries there. The only solution I can think of to route the outgoing traffic to either routerA or routerB is to use static routes. My company is yet again not fond of static routes, but everything I have read about OSPF being a link-state protocol has driven me to a dead end. Is there a way to influence the metric of the specific routes?

Any suggestions or reading material would be highly appreciated.

PS. The attached file does not show the topology behind the 2 CPEs. Each is connected to a 6509 and the 6509 have a connection between them. So it looks like a square with one side missing.

Thank you in advance,


I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Giuseppe Larosa Wed, 09/08/2010 - 04:11

Hello Katerina,

if no firewalls are on the path, the asymmetric paths are not an issue and you can live quietly with this.

to see if MPLS L3 VPN SP is using two devices you can check the OSPF router-ids of neighbors in show ip ospf neighbors.

if you see the same OSPF router-id on both of your CE nodes you can bet that there is only one PE.

Having two PE nodes is important for fault tolerance and you may paying for it.

You can only have one CE primary and used until it is alive and one CE secondary used if first one fails but the price to pay is in terms of performance.

if the SP is doing well its job you should see the remote branch IP subnets as OSPF O IA (inter route), if so increasing the ospf cost on CEB interface to PE would make return path via CEA.

int type x/y

ip ospf cost 1000

For receiving traffic in a primary/secondary schema you could increase OSPF cost on links between CEB and internal C6500 nodes.

if SP is giving you only O E2 routes for the remote sites ( not a good job) you may have more troubles attempting to choice the paths.

But I would not change current setup as there are no real issues with asymmetric paths if no FWs are on the path.

about having half sites using RA and half sites using RB:

only SP could do something about this if it is using two different PE nodes playing with BGP local preference in BGP address family vpnv4.

in OSPF you cannot achieve this as you have already noted.


it may be helpful for your current scenario the use of

show ip cef exact-route

that allows on the C6500 to understand what interface and next-hop is used by a flow with a specific source address, destination address.

This may be acceptable for your management as you can demonstrate to be able to track the path that is used by a specific flow.

Hope to help


katerina.dardoufa Wed, 09/08/2010 - 05:41

Thank you very much for your answer Guiseppe.

It really helps to know that assymetric routing won't be such a worry.

My question regarding your answer is that if I change the cost on the interfaces towards PE and 65xx then one link will always be preferred. Correct?

The good thing is that my company is beginning to understand the benefits of BGP, so I hope that we will implement my original suggestion. Furthermore it turns out that our provider is pretty unwilling to form OSPF neighborship with us!!!

I have been dying to implement BGP, so here is my chance

Chetan Kumar Ress Wed, 09/08/2010 - 06:29

Hi Katerina,

As  Guiseppe told that you won't face any issue ultill an unless you have firewall in current senario.

And in existing senario you have two options:

1] To keep Router A as a primary & Router B as a secondary.

2] Or keep as same your current network.

So for 1st option you can play with OSPF cost to set as router A as a primary & Router B as a secondary.

And if you want to distribute traffic with both router A & router B then you need to configure PBR in your Core switch saying that few network next hop is router A & other remaning network use router B that use default becasue of OSPF cost , so only few netork need to divert to other router . so in this way  you can distribute traffic from HO to location & From Locaiton to HO you can tell ISP to use same  PBR list to set next hop as a router A or B ( in this senario you need to configure first as option 1 mention above so only one will be active & second will be backup ,So using PBR on both  i.e. Core & PE you can divert traffic to other path that is unused)

But i dont know if your company don't have any firewall in current senario or any issue  then why need a change or this is what above solution i don't prefer to do becasue to achive what you want you have special dedicated protocol i.e. BGP.

but still we are engineers we need to follow the orders given by company.


Chetan Kumar


This Discussion