Looking for some advice on what others are doing with NFS over 10Gb. The setup will have 2 x 10Gb ports per filer with each port into a none vPC Nexus 5548. Each ESX server again has dual 10Gb ports terminated on each on the N5Ks. The ESX servers are running Nexus 1000v with the uplinks based on mac-pinning (due to the none vPC config upstream).
The initial thought is to configure a single mode vif with the active interface favouring one of the N5K, the vmkernal interface would also be pinned to a pin group on the uplink attached to the same N5K.
This design wastes bandwidth (active passive on the filers). However, the major concern is traffic being black holed if the physical interface / cable etc fails on either the esx servers or filers on the active interfaces... Ie, if the cable fails on the filer active interface traffic will failover to the other N5K, however, the ESX server will still be sending NFS via the original pinned uplink...
PS – I would prefer not to link to the 2 x N5Ks, also, vPC is not being used due to issues with routing via the N5Ks.
