Cisco UCS Chassis PSU Issue

[jon-alec.morris7]

Hi,

i have a chassis that has 2 PSU's that are faulty, we have 7 blades but only 2 in use.

Power supply is Grid.

PSU3 and 4 keep coming up with the voltage-problem and then power-problem. the power problem happens every minute alternatively (psu3 power problem - resolves- then a minute later psu4, then a minute later psu3 again and so on and so forth) every half hour or so i get the voltage-problem on both of them, but not at the same time. the cables are in firmly to the psu and the supply, i want to reseat them, but i don't want the chassis to die. Seeing as it is a low load, will PSU 1 & 2 (PSU 2 is in power saving mode) be able to support the chassis correctly? 

if i remove psu 3 (this seems to have more voltage problems) first, and psu 4 errors, does psu 2 automatically come out of power saving mode when it detects a missing psu?

Thanks for reading

[Steven Tardy]

The docs for PSU redundancy are okay but not great.

https://www.cisco.com/c/en/us/td/docs/unified_computing/ucs/hw/chassis-install-guide/ucs5108_install/ucs5108_install_chapter_01.html#ID476

I think of UCS Chassis PSU redundancy as:

• Non Redundant (N)
• N+1 (N+1)
• Grid (N+N)

If your servers are using less than 2500W then N will be 1.

If your servers are using more than 2500W (and less than 5000W) then N will be 2.

UCS CLI commands to see what is going on:

scope chassis 1
show psu
show psu detail
show power budget

Look for the  "Power State" line on the detail command to know if a PSU status:

• "On" (participating in "N") or
• "PwrSave" (NOT participating in "N")

With only 2 blades in use N would likely be 1 and that the single PSU1 should hold up the chassis.

When removing PSU3, and PSU4 errors. PSU2 would NOT come online unless using more than 2500W (N).

If PSU2 were to come online when PSU3 and PSU4 are off then that would be (N+1)+(N+1) which isn't a configuration option because you couldn't ever use all of the servers in the chassis.

Now as to why PSU3/PSU4 aren't happy, I don't know.

Many "OMG UCS Chassis PSUs are failing" TAC cases come down to input power.

Are your PSU1/PSU2 on "wall" power but PSU3/PSU4 are on "battery" power or UPS?

Some UPS systems don't output a sine wave, but a jagged square wave which makes the PSU unhappy.

Hope that helps.

[richy-app]

Thanks for your help here, I have kinda similar issue but, in my case, the PSU slot that has the amber light if replaced with a working PSU still turns amber. That is something I am trying to understand. Could this be related to the PSU module in that slot that is having issues? Hopefully, my question is clear enough to help here. Thank  you

[jon-alec.morris7]

Hi Steven,

we are pressing ahead with this tonight, all the blades are empty so it should go nicely, we are just going for a reseat first, and then raise a TAC case after if this doesnt sort this issue.

all the racks are on the same UPS's (one A and one B UPS's) so the issue shouldn't be a UPS as no other chassis are exhibiting these symptoms.

thank you for your help!