I recently performed an NX-OS upgrade on a pair of 5596's running in vPC mode. Due to some lingering spanning-tree in the datacenter, this was a non-ISSU upgrade. We upgraded one switch at a time using the "install all..." command, with the idea that dual connected devices would still have connectivity via the switch not being upgraded at the moment. That proved to be partially true. During the upgrade of the primary vPC switch (SW1) servers still had the necessary connectivity. It appears that when SW1 came back online from it's upgrade and before the upgrade process was initiated on SW2, systems lost their communications. My questions is, what happens in this scenario to a vPC pair when there is a mismatch in the NX-OS code levels? Could this have caused all forwarding to stop until both switches were complete with their upgrades?
