Supermicro H8DGi-F change fan policy from ipmitool

[bellocarico]

Hi there

I have a SuperMicro H8DGi-F

I can see in the BIOS a setting to change the fan policy like power saving/balanced/performance/full speed

I have noticed that changing these settings affect two factors:
- the minimum speed the fans will spin to
- the reactivity to temperature increase

Essentially what I'm trying to do is to change this setting without having to reboot the server. I would ideally looking to use e.g. Performance in summer and perhaps Power Saving in winter, mainly due to the noise.

I had a look at the IPMI documentation: https://www.supermicro.com/manuals/other/IPMI_Users_Guide.pdf
I can see (and have performed) a calibration of the fan alerts but but I don't see how I can change the "fan policy" while the server is running. Unless I'm missing something of course.

Can anybody please point me in the right direction?

Many thanks!!

 

[EffrafaxOfWug]

I know that on current versions of supermicro IPMI you can read the current fan policy through the following raw:

ipmitool raw 0x30 0x45

...but I've no idea if this will work on IPMI as old as the ones in the H8 series, I suspect it might not as they use some quite different IPMI syntax IIRC.

P.S. PigLover's guide to some of the IPMI raws for the X9/X10/X11 series is here and the associated forum topic is here.

 

[bellocarico]

I know that on current versions of supermicro IPMI you can read the current fan policy through the following raw:

ipmitool raw 0x30 0x45

...but I've no idea if this will work on IPMI as old as the ones in the H8 series, I suspect it might not as they use some quite different IPMI syntax IIRC.

P.S. PigLover's guide to some of the IPMI raws for the X9/X10/X11 series is here and the associated forum topic is here.

wow that was quick

Do you mind if I ask what the 0x30 and 0x45 represent?

root@centos7~/ipmitool# ipmitool -H 10.10.10.97 -U xxxxxxx -P xxxxxx raw 0x30 0x45
 00

root@centos7~/ipmitool# ipmitool -H 10.10.10.97 -U xxxxxxx -P xxxxxx raw
Not enough parameters given.
RAW Commands:  raw <netfn> <cmd> [data]

Network Function Codes:

  VAL   HEX     STRING
==============================================
  0     0x00    Chassis
  2     0x02    Bridge
  4     0x04    SensorEvent
  6     0x06    Application
  8     0x08    Firmware
  10    0x0a    Storage
  12    0x0c    Transport

(can also use raw hex values)

Many thanks!!

 

[EffrafaxOfWug]

They don't really represent anything I'm aware of, they're just the numbers that SM picked to use for their IPMI raw values for fan control. Raws are rarely standardised - for example, Supermicro and ASRock both make heavy use of the AST2x00 BMC chips but ASRock's raw numbers are completely different to those of SM even on the same BMC chips. I'm not aware of finding out what any particular set of random numbers might do when run as a raw except from information gleaned from observation or released by the vendor.

Anyway, interesting to see you got a 0x00 response out of that first raw - is your IPMI web GUI set to have the fan control mode set to "standard", as that's what this particular raw is meant to represent?

If you feel like trying to change it you can try running `ipmitool raw 0x30 0x45 0x01 0x01` - after running that, the fans should all run at maximum rpm, and if you run `ipmitool raw 0x30 0x45 0x01 0x01` it should put them back in to standard operation.

To explain:
ipmitool raw 0x30 0x45 0x00 - read the IPMI value
ipmitool raw 0x30 0x45 0x01 - set the IPMI value (except it won't work as you haven't provided a new value)
ipmitool raw 0x30 0x45 0x01 0x02 - set the IPMI value to 2

I'm not making any promises though as I'm not sure which "version" your IPMI chip is using; the above commands might make your house catch fire etc etc.

 

[lpallard]

@bellocarico : did you manage to remotely control fan speed/mode via ipmitool? I am trying to achieve the same with a H8DCL-iF and so far I've only hit brick walls....

 

[bellocarico]

As per today I still don't understand why one server (identical motherboards) answers via IPMI and the other not. I guess I would need to cpmpare the full ipmi output to find out...