I have a UCS 5108 chassis with eight B200 M5's, I upgraded the chassis manager from 4.3.4a to 4.3.5a at Christmas time but when upgrading a blade from 4.2.1f to 4.3.5a this week it failed at the following step:
Associate Check Power Availability - Check if power can be allocated to server 1/x(FSM-STAGE:sam:dme:ComputerPhysicalAssociate:CheckPowerAvailability)
The server was powered off at this point so I tried powering it on which went fine although it booted into the UEFI shell rather than the OS on the disk. In the power control monitor for the chassis I noticed the server was allocated 25W and its oper state was unbudgeted whereas the other servers (all identical) were allocated 550W and set to deployed. All including the problematic server were running a no-cap power priority. I came across this Cisco bug report:
https://quickview.cloudapps.cisco.com/quickview/bug/CSCuy52691
It's for an older version and the symptoms are different since I had no issues powering the server on however I tried the workaround changing the power policy to 5 and everything worked normally again, the server was able to complete its upgrade and booted its OS. It has changed the maximum power from 550W to 700W for this server in the Power Control monitor.
I'm wondering if this is the same bug I'm seeing and if I should proactively change all the other servers to the same policy setting of 5 instead of no-cap? The servers have been running for years now without issue so I'm wary of making a change to fix one proble that could cause others.
