Supermicro chassis temperature sensor

[matt_garman]

Does anyone happen to know how Supermicro chassis determine temperature (and in turn fan speed)?

I got a SC836A-R1200B. I like everything, except the fans are crazy loud. But I know they are temperature controlled. When I first built the system and turned it on, after boot, the fans settled down, and it was an acceptable noise level.

Then I turned the system off, shuffled some stuff around in my rack, and turned the system back on. Now the fans are running full-speed, and the temperature light on the case is flashing.

For some reason, the system thinks something is too hot, so it won't let the fans spin slowly. But I can't figure out what that is.

I monitor temperatures pretty closely with Munin, and no temps have gone up... only fan RPM. (I can even see on the Munin graphs a short time where I first powered on the system and the fans ran slowly.)

Any thoughts?

 

[Patrick]

Dis you look at IMPI 2.0? Usually you can see there.

 

[matt_garman]

Dis you look at IMPI 2.0? Usually you can see there.

How does one do that? Are you suggesting using SM's "Super Doctor III" software? If so, that's only available for Windows (well, there's a Linux version, with a fraction of the functionality).

Anyway, I went into my motherboard's BIOS and changed the fan profile from "Balanced" to "Energy Efficient". That dropped the fan speeds significantly, and the temp warning light on the chassis is no longer lit.

So far so good...

 

[Patrick]

What do you have motherboard wise? Setting those BIOS settings helps a lot. Glad to hear that is helping

 

[matt_garman]

What do you have motherboard wise? Setting those BIOS settings helps a lot. Glad to hear that is helping

The motherboard is the (ubiquitous?) Supermicro X8SIL-F.

 

[matt_garman]

Changing the BIOS fan profile didn't seem to be a permanent fix. After a while (hours to days), the thermal sensor gets tripped and the fans start blowing at full speed and the "thermometer" light on the chassis starts flashing. It stays like this indefinitely, until a reboot. After rebooting, the fans stay slow until something randomly trips the sensor again. I'm sure it's not a load issue---the server is idle practically all the time. I also setup all the logging and monitoring I know about, and there's definitely nothing interesting going on when the sensor gets tripped. Anyone other ideas?