Nexus 3000 layer 2 or 3 uplink to Nexus 9396

[daider.lim]

Hello all,

I am using a pair of Nexus 3000 switches as TOR switches (access layer). Both Nexus 3000 switches will be configured for vPC peer setup

and using its management ports for vPC peer keepalive.

Next, I intend to connect both Nexus 3000 switches using their 40Gb uplink ports to a pair of Nexus 9396 switches. Both Nexus 9396 will

be acting as aggregation switches with vPC peer setup and its management ports for vPC peer keepalive and running OSPF routing protocol.

My question is which best practices for this kind of setup:

a) since Nexus 3000 support EIGRP protocol, should I run inter-vlan routing at the Nexus 3000 switches and configured its uplink to

    Nexus 9396 switches using layer 3 and redistributing EIGRP into OSPF at Nexus 9396 switches ? Any advantages?

b) simply using Nexus 3000 for layer 2 connection and its uplink to Nexus 9396 also layer 2, so all routing done at Nexus 9396? Any advantages?

Appreciate some kind advice. Thank you

[Reza Sharifi]

Hi,

I would keep the TOR switches (3000) as layer-2 only.  Create the SVIs for all your vlans on the 9396 and do the inter-vlan there.  This keep your design pretty simple.

HTH

[daider.lim]

Hi,

Thanks for the advice. That looks a lot simpler to setup.

[sbhadrav@cisco.com]

In a high scale setup with many BGP routes getting redistributed into EIGRP, modify the EIGRP signal timer to ensure that the EIGRP convergence time is higher than the BGP convergence time. This process allows all the BGP routes to be redistributed into EIGRP, before EIGRP signals convergence.
When OSPF is used as a protocol between PE and CE devices, the OSPF metric is preserved when routes are advertised over the VPN backbone. The metric is used on the remote PE devices to select the correct route. Do not modify the metric value when OSPF is redistributed to BGP and when BGP is redistributed to OSPF. If you modify the metric value, routing loops might occur.