Our 2 FIs are 6296UPs with fully-populated expansion slots, 8 chassis (8 x 10Gigabit fiber optic cables per side per chassis), FC to our VMAX SAN (no blades have any local storage).
We have 28 x B200 M4s, and 18 x B420 M3s, leaving only a couple of full-width blade slots empty in the entire UCS domain.
We are planning an upgrade that will migrate UCS from 2.2(3c) to 3.1(2e).
I need to coordinate with our DBAs to take databases down beforehand and bring them back up afterward, and they will probably appreciate some sort of estimate from me as to how long they will have to wait to bring the databases back up afterward.
We tentatively plan to take all production running on this system down and acknowledge each blade reboot as soon as they request it, as opposed to the "sliding tile game" of shuffling VMs off each blade (host) just before rebooting each one.
How long should I expect each stage to take (1. UCS infrastructure upgrade, then 2. UCS server upgrade)? I do plan to use autodeploy for infrastructure and server phases of the upgrade.
Focusing on the server upgrade stage for a minute: 1. How long for the server upgrade stage if I do actually take all production down beforehand, and can reboot each blade as soon it requests a reboot, and 2. How much longer would you guesstimate it might take if I opted to play the "shuffle" game of only rebooting 1-3 hosts at a time because of moving production VMs to another blade before rebooting? We have at least 3 surplus hosts of compute power according to vSphere HA.
Thanks for any help with this. Any information from people with experience is better than a wild shot in the dark by me.
Oh, and I understand it can potentially take as long as having Cisco replace both FIs if they both get bricked. I am more interested in better case scenarios and how long it typically takes in situations like this.
Thanks in advance,
Andrew
