OSPF not load balancing over MPLS network

[dayo.olatunji]

I have an MPLS network with 4 PE routers running ospf btw them. Btw 2 of them, I have 2 parallel links with equal cost. Sometime ago, I noticed an upsurge in traffic over one of the links while the other remained stable. I shut the link and the traffic moved to the other leg and remained even when I opened up the 1st leg. What could be the problem?

[swaroop.potdar]

The load balancing happens in Cisco by the hashing derived by CEF. Can u check whether your routers are configured for per packet load balancing.

As if its at Per destination load balancing then you will see traffic for some certain session stuck to one link and uneven traffic distribution.

Here is the link to configure per packet load balancing.

http://www.cisco.com/en/US/products/ps6350/products_configuration_guide_chapter09186a0080430ac3.html

HTH-Cheers,

Swaroop

[dayo.olatunji]

Hello Swaroop,

Thanks. Another quick question. Is there any way i can trace where this traffic is coming from? I'm using the uBR 7246 for my PEs with many vrfs within. I'd be grateful for your continued assistance.

Rgds,

Dayo

[swaroop.potdar]

1) You can enable MPLS Aware Egress Netflow.

Here is the link to configure the same.

http://www.cisco.com/en/US/products/sw/iosswrel/ps1612/products_feature_guide09186a008008081e.html

I am not sure whether this feature is supported by the IOS/HW you have currently.

Please verify the same.

2) Second option is since you have onlu 2 of PE's involved in this uneven behavior you can check on the destination PE where the Egress traffic is uneven. Then you have to verify interface by interface throughput for all the dropped VRF's. Tedious hardwork if MPLS netflow support isnt there on the HW/SW you have now.

HTH-Cheers,

Swaroop

[swaroop.potdar]

Here is a good doc which you can read for your load balancing issue.

It has a example for when you have a pop Label involved.

http://www.cisco.com/application/pdf/en/us/guest/netsol/ns172/c654/cdccont_0900aecd802e18af.pdf

HTH-Cheers,

Swaroop