Had to toss together quickly a high level “What is going to happen” email regarding updating code on a UCS. Put the below together in a blazing 15 minutes. Isn’t perfect but was fun to make. Thought I would share my primitive diagram skills. Feel free to crib any of this for your purposes.
UCS has two controllers FIA & FIB. 50% of connectivity for each server on UCS goes through the Active/Active controllers. Each FI is rebooted for code update, one at a time.
If the a server has 4 paths to SAN, 2 are on each FI. So 2 will fail each FI update, but the updates are staggered so connectivity is never lost completely. Multipathing should handle this server side.
When networking fails over Bonded NIC’s will see one side fail at a time but not both at once. If only one NIC is present on server, UCS will handle failover and server will not see it (other than tcp/ip session reconnect but interface will not go down).
Most servers live through this without any issues. ESX, Linux, and Windows bare metal servers usually do fine.
Can only give a 99% guarantee that a server will have no issues.
Link to source here http://realworlducs.com/stuff/Drawing4.vsd
Graphical interpretation of upgrade below