Hi,
This is possibly due to the loop avoidance mechansim built into vPC. Essentially a Nexus device running vPC will not forward trafifc on a vPC member port if the traffic has crossed the peer link.
When the traffic is sent from the PC attached to the Catalyst switch, that switch will make a hashing decision at Layer-2 to decide which physical link of the port-channel to use. If the traffic destined to the IP address 10.234.100.1 takes the physical path to N9K2, it will then have to cross the peer-link to get to N9K1. That's OK, but consider the return traffic.
The MAC address table on N9K1 knows the MAC for the PC via the peer-link and will forward the return traffic on that link. Now, based on the loop avoidance mechansim i.e., a switch will not forward traffic on any vPC member port if that traffic has crossed the peer-link, N9K2 will drop the return traffic.
You may also be able to solve the problem by adding the peer-gateway command within the vpc domain <domain-id> context. For example:
!
vpc domain 1
[..]
peer-gateway
[..]
!
This commands allows a switch to route traffic destined to the MAC address of the peer i.e., N9K1 can route traffic destined to the MAC address of N9K2. This is actually considered a best practice anyway for vPC.
In reality you probably wouldn't have this configuration, but instead use a First Hop Router Procotocol such as HSRP which would be your default gateway IP. In a vPC environment both routers are active from a data plane perspective and so you'd avoid this issue. The use of HSRP with vPC is explained in the HSRP/VRRP active/active with vPC section on page 78 of the Design and Configuration Guide: Best Practices for Virtual Port Channels (vPC) on Cisco Nexus 7000 Series Switches.
The above design guide is a good source of information, but a long read. If you're new to vPC then an excellent starting point is the Quick Start :: Virtual Port Channel (vPC) guide. This shows typical configurations, but also has a number of recomendations and best practices documented.
Regards
