Solved: Re: UCS mini upgrading infastructure firmware on FI

tnrdit001 · ‎01-25-2022

We have a Cisco UCS mini. I'm reviewing the instructions for upgrading the infrastructure firmware. ( 3.2(3n) - 4.1(3f) ) I have a question regarding the fabric interconnects during the upgrade process.

It's been many years since I have been involved with updating this system in a production environment. I understand there will be a data interruption while the fabric interconnects reboot and transfer roles. I'm reading about the FI traffic evacuation. I'm I understanding this correctly that this will route the traffic to the other FI during the restart so I wont have to power off the hypervisor/VM/OS?

Wes Austin · ‎01-27-2022

There should be no unexpected data plane interruption during the upgrade process nor doing the failover/role transfer. If you are using auto-install this process will be conducted for you. The only impact you will see is loss of connectivity to the UCSM GUI briefly.

Fabric evacuation can be used, but is not required. It will shutdown all the host interfaces (HIF) on the IOM connecting to your blade servers on either the A or B side.

For UCS installations you typically have paths to both A and B side fabrics in ESXi. So when you reboot FI-A or FI-B during the upgrade, the path that is not upgraded just stays online and serves the traffic. There is no "failover" or "routing of the traffic to the other FI". Both paths are active/active in UCS always and one goes down per fabric as its upgraded.

Now you could be using Fabric Failover (you can check on the vNICs in UCSM) and if this is the case, UCSM would failover the vNIC from A > B or vis versa during a reboot or upgrade.

View solution in original post

Wes Austin · ‎01-27-2022

There should be no unexpected data plane interruption during the upgrade process nor doing the failover/role transfer. If you are using auto-install this process will be conducted for you. The only impact you will see is loss of connectivity to the UCSM GUI briefly.

Fabric evacuation can be used, but is not required. It will shutdown all the host interfaces (HIF) on the IOM connecting to your blade servers on either the A or B side.

For UCS installations you typically have paths to both A and B side fabrics in ESXi. So when you reboot FI-A or FI-B during the upgrade, the path that is not upgraded just stays online and serves the traffic. There is no "failover" or "routing of the traffic to the other FI". Both paths are active/active in UCS always and one goes down per fabric as its upgraded.

Now you could be using Fabric Failover (you can check on the vNICs in UCSM) and if this is the case, UCSM would failover the vNIC from A > B or vis versa during a reboot or upgrade.

tnrdit001 · ‎01-27-2022

Thank you for this. Very helpful.