need some help with Supermicro BMC and fan speed

[BLinux]

I have a Supermicro X9DR3-LN4F+ system I'm tinkering with. Up until tonight, the system behaved what I thought was normal, where the fan speed was controlled by the motherboard and would quiet down at idle and speed up under load. All was well...

Tonight, I was experimenting with adding an extra fan + custom fan shroud to the PCI slot area. I connected a Noctua 120mm NF-F12 Industrial iPPC-3000 PWM fan. The PWM signal to the fan caused it slow down all the way to near 300RPM, which hit the fault threshold (as seen in 'ipmitool sensor') and caused all fans to speed up. It would then eventually slow down again, hit the threshold, go full blast, rinse & repeat. However, my 1st question isn't about how to adjust the PWM fan signal and/or thresholds. The main problem I'm having now is that after disconnecting the Noctua 120mm fan, the system is constantly running at full blast; it will no longer slow down the fans. Is there something I need to do to have it resume the behavior it did prior to me connecting the Noctua fan? Is there something in the BMC that I have to reset? Not familiar with Supermicro systems that much....

Then, my second question, if I can get the 1st problem fixed, and re-try my experiment, is there a way to adjust how the motherboard is controlling this Noctua fan? Is there a way to adjust the fan speed curve individually?

 

[realtomatoes]

haven't had much supermicro experience, so can't give a fix for it.

but i'd like to help troubleshoot and learn something new too.
are all the fans NF-F12 iPPC-3000 pwm fans?
are they connected to separate fan headers or share some?
are all the fans pwm fans?

 

[i386]

Reseting the bmc helped me when I had similar problems with the fans. (IPMI saw a fan that wasn't there and therefore ran all the fans at 100%)

 

[K D]

Reseting the bmc helped me when I had similar problems with the fans. (IPMI saw a fan that wasn't there and therefore ran all the fans at 100%)

This. Reset the BMC you solve the phantom fan issue.

 

[BLinux]

Reseting the bmc helped me when I had similar problems with the fans. (IPMI saw a fan that wasn't there and therefore ran all the fans at 100%)

how?

 

[BLinux]

haven't had much supermicro experience, so can't give a fix for it.

but i'd like to help troubleshoot and learn something new too.
are all the fans NF-F12 iPPC-3000 pwm fans?
are they connected to separate fan headers or share some?
are all the fans pwm fans?

1. no, only 1 fan was a iPPC-3000 PWM. the others were the standard factory fans of a Supermicro 846 chassis (3 mid-plane, 2 rear).
2. all separate headers, X9DR3-LN4F+ has 8 total.
3. yes, all fans are PWM.

 

[i386]

In ipmi: Maintenance > Unit Reset

 

[BLinux]

In ipmi: Maintenance > Unit Reset

Since I didn't see that in the BIOS IPMI menu, I'm going to guess that's in the IPMI web UI? Is there a CLI method to do the same? Unfortunately, I don't have a wired network where this machine is being worked on; all I have for network is wifi via a USB wifi adapter for network.

 

[BLinux]

for future reference, i think i found my answer:

# ipmitool raw 0x3c 0x40

haven't tested it yet... will do when I have time.

 

[K D]

Or if possible Unplugging and replugging the power cable works too

 

[BLinux]

Or if possible Unplugging and replugging the power cable works too

that was the first thing I did. but it did not work. fans were full blast even after unplugging for 10 min. i think this BMC must save some data somewhere, but I don't know as I'm not that familiar with these systems.

 

[K D]

that was the first thing I did. but it did not work. fans were full blast even after unplugging for 10 min. i think this BMC must save some data somewhere, but I don't know as I'm not that familiar with these systems.

Hmm... Interesting. The Unit Reset just reboots the BMC. Unplugging the power and replugging it in should do the same. It has worked for me,

 

[realtomatoes]

if there was still enough juice to power the ipmi (might have a battery) after the plug was pulled, it would take some time before it went offline.

 

[Aestr]

if there was still enough juice to power the ipmi (might have a battery) after the plug was pulled, it would take some time before it went offline.

This isn't really a issue with Supermicro boards. I've had to wait enough times for the BMC to initialize after pulling the plug briefly that it should do the trick for him. Most boards have an LED for BMC status you can watch if you're being extra careful.


Sent from my iPhone using Tapatalk

 

[BLinux]

just to report back for the benefit of someone else who might find this thread.

the BMC reset was indeed this command:

# ipmitool raw 0x3c 0x40

takes a few seconds, but the BMC shuts off and boots up again, and all the data is cleared. even after it is up and you can read the sensors with "ipmitool sensor", many of the readings will be blank for a few seconds.

Unfortunately, this did NOT resolve my problem with the fans being at full speed. So, I ran:

# ipmitool raw 0x30 0x45 0x00

which revealed "1", meaning "full speed" mode. some how the BMC had switched to "full speed" mode, not sure how?

I then changed it to "optimal" mode 2:

# ipmitool raw 0x30 0x45 0x01 0x02

Then everything went back to how it was prior to all this.

 

[HorizonXP]

I'm running into this exact same problem, but the commands aren't resolving it. Going from "full speed" mode to "optimal" mode is having no impact.

Interestingly, optimal mode is 3 for my X9DRD-iF, and mode 2 doesn't exist.

 

[HorizonXP]

Did some more troubleshooting today, eventually moving the whole system to an 825 chassis. No matter what I do, the board just runs the case fans at full speed. 14k RPM in the 1U chassis, and 6.8k RPM in this 2U chassis. I've done unit resets, resets to factory defaults, raw commands, and none of them have any effect. Fan mode changes have no effect, and I'm not hitting the low RPM threshold.

This is very annoying. I can't believe that I've potentially hosed a board here.