UCS-C220-M3L plus LSI 9272-8i controller = fan speed 16k rpm

[nixx-kstu]

Hello.

I'm trying to use the LSI 9272-8i controller on the Cisco UCS-C220-M3L server (fully updated with ucs-c220-huu-3.0.3e.iso).
The server in its CIMC checks the controller and turns on a very high fan speed (16-17k rpm), because the controller issues a temperature of 255 degrees Celsius on the ROC, but in fact the controller is still cold.
I check the temperature of the controller using the LSI StorCLI utility and get also 255 degrees (storcli /c0 show all). I have five of those controllers, and they all show the same heat. I understand that this is a controller problem, but updating the firmware on it to the latest version (23.34.0-0019) is without result.

So can I somehow turn off the RAID-controller temperature check in the CIMC?

[bsvzpua]

Hi.

I have the same problem with this card on Cisco UCS C240 M3.

Did you find any solution?

[Wes Austin]

Can you confirm the controller is 9272 and not 9271. I do not see 9272 on the spec sheet, so CIMC is most likely not recognizing the card and spinning the fans up as a safety precaution to cool the server.

https://www.cisco.com/c/dam/en/us/products/collateral/servers-unified-computing/ucs-c-series-rack-servers/c220m3-lff-specsheet.pdf

[bsvzpua]

Yes it's 9272, it's the same controller and the same chip that is in 9271, it is it's oem for NEC. I also flash to card firmware from 9271 and have the same problem.

CIMC recognize it, i can make raid and virtual drives from CIMC. Only one problem - CIMC read 255C temp on chip from controller, so all fans run at 16K rpm... But server and controller works correctly.

I know that it's not Cisco problem, but i have problems with my small homelab and try to find the decision.

P.S.

Sorry for my bad English.

[Kirk J]

When Cisco puts items on the spec sheet, it is after evaluating many aspects of equipment especially including manageability.

While the 9272 may have a similar core ROC chip on it, it may be missing other features that Cisco worked out with LSI/Broadcom including customized firmware.

Like Wes says, the 9272 wasn't QA'd or tested by Cisco, so the CIMC isn't going to have all the cooling and power profiles for the unknown card, and runs the fans in max warp  mode in order to accommodate a maximum cooling requirement scenario.

Kirk...

[jhuybers]

I have had the same issue with a couple of different LSI cards, both NEC branded models. A SAS9267-8i only showed the problem in a C220 M3 after a firmware update, unfortunately I do not have the original firmware it was on to set it back and test.

The 9272-8i had the 255C issue out of the box.

This does not appear to be limited to the C220 as I can put the card in other servers and workstations and the card reads 255C.

I have tried multiple firmware and driver combinations. I have an open ticket with LSI about the issue, they initially said that the card is faulty but this is different cards with the exact same issue. Maybe it is something NEC do differently.

I have a 9271-8i on the way.

[Kirk J]

Greetings.

Your scenario is expected.

The CIMC has been programmed via firmware to know about Cisco specific cards, their thermal profile/cooling requirements, power requirements, and have the correct firmware to communicate via OOB realtime stats like ROC temp, etc

If you want the optimal config that includes correct cooling, fan speeds, etc, then you need to stick to the server's specsheet equipment list for specific Cisco peripherals/adapters.

Kirk...