General

One of the Neuxs 3548 across vPC peer-link can not reach the direct connected host behind the vPC port-channel. All IP unicast traffic that generated from CPU is involved. L2 traffic , L3 multicast and L3 broadcast works fine. We do see ARP is resolved and HSRP works fine.L2 forwarding for data packet that came outside the switch to the impact port-channel is also fine TX counter did not increase for the impact Port-channel when we send the ping LCND show INVALID packet and it constantly increasing.

Symptom

One of the Nexus across vPC is not reachable to direct connected device. All Control-plane generated IP unicast packet are impacted. L2 forwarding is fine on data plane. TX counter never increased.

Conditions

Two vPC domain interconnect via back to back vPC setup. All four device across two peer-link are nexus 3548. For each domain, Multiple sever and few network device are directly connected via vPC port-channel to each of the Nexus. The issue is triggered after downgrade from 9.3.3 to i7(8). Upgrade to 9.3.5 will not help. Only host devices are impacted

Workaround

If we ping from the end host to the Nexus once, and everything for that specific Port-channel will start working again. ERSPAN with source interface as the peer-link and apply and than delete the static mac entry with exit port as peer-link will also bring the device out of the broken state. After reload, we will back to the broken state again.

Further Problem Description