Cisco Cisco Nexus 5010 Switch Guida Alla Progettazione

 

 

Design Guide 

© 2010 Cisco and/or its affiliates. All rights reserved. This document is Cisco Public Information. 

Page 26 of 38 

 

At the time of this writing, we recommend the use of Layer 3 links to connect the vPC aggregation layer with the 

Layer 3 core instead of the use of vPC PortChannels for Layer 3 connectivity. 

Figure 17 shows why. The design on the left shows of a router connected with a Layer 3 vPC to Cisco Nexus 

Switches Switch1 and Switch2. At the time of this writing this design does not work. Imagine that client 1 sends traffic 

to server 1. Router 1 has Switch1 and Switch2 as neighbors, so it load-balances the routed traffic to both BIA MAC 

addresses of routers 1 and 2. The PortChannel hashing is independent and may forward the routed frame with the 

BIA MAC address of Switch2 to Switch1 (and Switch1 to Switch2). In this case, the frame would traverse the peer link 

to be then routed to the PortChannel Po2. At this point, the duplicate prevention rule would intervene, and the frame 

would be dropped. 

Thus, at the time of this writing the connectivity between the core and the aggregation layers needs to follow the 

topology depicted on the right side of Figure 17. 

Figure 17.    Interactions Between vPC and Routing 

 

This section discusses the most important interactions between multicast and vPC. 

Layer 2 forwarding of multicast traffic with vPC is based on a modified IGMP snooping behavior that consists mostly 

of synchronizing the IGMP entries between vPC primary and secondary devices. 

In a vPC implementation, IGMP traffic entering a Cisco NX-OS device through a vPC PortChannel triggers hardware 

programming for the multicast entry on both vPC member devices. The synchronization of the IGMP information is 

performed over the peer link (the M1-to-M2 link in Figure 18) using Cisco Fabric Services over Ethernet.