cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1376
Views
0
Helpful
3
Replies

OSPF design issue

etxnreg
Level 1
Level 1

Hi,

I have a deign issue that I have problem to solve.

Background

Picture:See attachment

Our customer has a mpls network and using different vpn's for different traffic types.

Between the two sites 10 ge links is used and intra 1 ge links is used.

OSPF is used in between GW, R1 and R2, (also between R1 and R2 to prevent GW to be transit )

During normal situations the traffic from Gateway to VRF A R2 goes via R2.

Problem:

If the link B goes down the traffic should go via the other site, i.e GW to VRF 1 via R1 and R4.

That's is not the case because the ospf distance is better so it choose R1 => R2=> VRF A R2.

An idea I hade was to change ospf distance so it is higher then BGP.

After a test it works well during a faulty link (B).

However during normal operation (both link up)the traffic wil not use the right path i.e GW=>R1=>VRF-A R2.

Instead it will use GW=>R1=>R2=> VRF-A R2.

Why is this happend when I changing Distance?

Has somebody and idea how I can solve this?

Thanks

Niklas

1 Accepted Solution

Accepted Solutions

Hello Niklas,

the OSPF sham link is actually built by using BGP multiprotocol to carry specialized route targets that convey OSPF LSA information so that it can be rebuilt on remote PE.

In this way from the backbone point of view you are comparing MP BGP routes and this should help.

The sham link allows to rebuild O routes instead of O IA that is the normal type of rebuilded routes when using the MPLS superbackbone.

So seeing this from the customer side if you have a primary path that is O type and a secondary type that is O IA you should be fine.

Verify all fault and restore events to be sure that it works in any case.

It can be a possible solution.

Hope to help

Giuseppe

View solution in original post

3 Replies 3

Giuseppe Larosa
Hall of Fame
Hall of Fame

Hello Niklas,

it is not totally clear to me.

My understanding is the following:

the customer has two multihomed VRF sites.

GW is the CE of site1.

VRF A R2 is the CE of site2.

R1 and R2 are the PE nodes that advertise in MP BGP the IP subnets of site1. OSPF is the PE-CE protocol running on R1, R2 (VRF A instance) and on VRF A GW.

R1 and R2 have MPLS connections to R3 and R4 with OSPF/LDP/BGP.

R2 has also a connection 10 GE to a L2 network that allows to reach VRF A R2. From the point of view of R2, VRF A R2 is another VRF site that can communicate with site1 VRF A or it is the same VRF?

Putting the link 'E' in a separate VRF may be a way to have a cleaner design.

Also I see R2 and R4 providing a VRRP def GW to VRF A R2. What PE-CE protocol is used between VRF A R2 and its two PE nodes?

At level of PE nodes like R1 and R2, using OSPF means moving inside VRF A routing table, when using MP BGP it involves going to other PE nodes to other VRF sites that can communicate with VRF A site 1.

I don't understand why R2 is PE node for both site1 and for VRF A R2 this is the key point here.

How have you configured the 'E' link on R2? why is R2 also providing VRRP together with R4?

The design rules should be: have multihomed sites but use backbone links to connect between sites. I may be wrong but your current design is a mix.

Moving admin distance has too big effects as you have seen.

Hope to help

Giuseppe

Hi,

Thanks for the quick response and hints.

I have done following changes:

Removed the ospf (vrf) instance between R1 and R2.

Added a sham-link ospf (vrf) instance between R1 and R2.

After the changes all fauiler situation seems to work.

Comments?

Thanks Niklas

Hello Niklas,

the OSPF sham link is actually built by using BGP multiprotocol to carry specialized route targets that convey OSPF LSA information so that it can be rebuilt on remote PE.

In this way from the backbone point of view you are comparing MP BGP routes and this should help.

The sham link allows to rebuild O routes instead of O IA that is the normal type of rebuilded routes when using the MPLS superbackbone.

So seeing this from the customer side if you have a primary path that is O type and a secondary type that is O IA you should be fine.

Verify all fault and restore events to be sure that it works in any case.

It can be a possible solution.

Hope to help

Giuseppe

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: