I have a weird connectivity issue for the ports on a pair of NX9K, which seems like one way traffic problem. The setup is:
NX7 --- NX7 - has HSRP gateway for the VLAN
| \/ |
| /\ |
NX9 --- NX9 < VPC peer-link between
HOST on VLAN 2
So the host on VLAN 2 cant establish IP connectivity to the gateway of the VLAN. If I ping from the host towards the gateway, then the NX7 switches will create an ARP entry for the right MAC address and IP, but no ICMP reply reaches the host. If I ping from the NX7 to the host IP, then it doesn't find the host. If I create an SVI on the 9K in VLAN 2 with an IP address, then IP connectivity between the NX7 and NX9 on VLAN 2 works just fine. If I create an SVI on the 9K, then I can ping between the host and the 9K on the VLAN. So there is some layer 2 problem here. The MAC address of the host is available on all the switches. The port connected to the host is not a port-channel/VPC. The 9Ks and 7Ks are not running back to back VPC, STP is just blocking one link.
Is there anything to be aware of on the 9K platform in regards to FEX access ports or something else? I've stared myself blind on this problem, so I could use some input ideas.
At this moment very difficult to say where is the problem resides, if we can see you configuration 7K - 9K and Also Fex port configuration you have connected.
Make sure Vlans are announced in vPC peer-link. also give more inputs do you have any IGP running here ?
The 9Ks and 7Ks are not running back to back VPC, STP is just blocking one link.
I would not say that it will not work with this but it is not a recommended design. For troubleshooting your problem, I would look at the STP states, vPC VLAN forwarding, ensuring vPC loop avoidance rule is not dropping packets.If possible I will shut down one N9K and test this scenario. However it would be best if you can have a back to back vPC between the N7K and N9K which is the recommended design
Here is snippet of vPC loop avoidance rule :