Register for the iXsystems Community to get an ad-free experience and exclusive discounts in our eBay Store.

Script: Hybrid CPU & HD Fan Zone Controller

Western Digital Drives - The Preferred Drives of FreeNAS and TrueNAS CORE

Stux

Wizened Sage
Joined
Jun 2, 2016
Messages
4,163
Hello everybody.
I'm still very new to the FreeNAS world.
I'm setting up a FreeNAS box in my living room so I want it to be whisper quiet and thought a script such as Stux's or Kevin Horton's mod could help tremendously, but unfortunately neither of the two works as expected on my machine.
Could it be because I'm using a Supermicro X11 board (X11SSM-F)?
This is what I get when I run Stux's script at debug level 1:

Code:
root@FreeNAS:~ # ./hybrid_fan_controller.pl
2017-07-07 12:40:56: CPU Temp: 36.0 <= 36, CPU Fan going low.
2017-07-07 12:40:57: Maximum HD Temperature: 37
2017-07-07 12:40:57: Drives are warm, going to 80%
2017-07-07 12:41:09: CPU fan speed should be low, but 2000 > 1600.
2017-07-07 12:41:26: CPU fan speed should be low, but 2000 > 1600.
2017-07-07 12:41:31: Resetting BMC

If I understand correctly, the BMC is ignoring the ipmitools commands to lower fan speeds (CPU fan speed should be low, but 2000 > 1600) even after being reset.
Do you have any ideas how I could fix this? Thank you!
Which fan headers are you using for what fans?

First thing you should check is using your IPMI set the fanmode to standard/optimal and verify that your fans do actually slow down. You need PWM fans for them to be controllable.
 

glauco

Senior Member
Joined
Jan 30, 2017
Messages
338
Right, I'm using FAN1 to connect the CPU cooler and FANA with a splitter to connect two push-pull fans that cool the drive cages.
 

glauco

Senior Member
Joined
Jan 30, 2017
Messages
338
You're right, switching to other fan modes (standard, optimal, heavy I/O) doesn't change a thing, the fans keep going at full speed (2000 rpm the CPU fan, 1500 rpm the two HDD fans). So this isn't related to your script. My IPMI firmware revision is the latest (1.13). Where else can I bang my head? Perhaps there'a BIOS setting I should change?
 

Stux

Wizened Sage
Joined
Jun 2, 2016
Messages
4,163
You're right, switching to other fan modes (standard, optimal, heavy I/O) doesn't change a thing, the fans keep going at full speed (2000 rpm the CPU fan, 1500 rpm the two HDD fans). So this isn't related to your script. My IPMI firmware revision is the latest (1.13). Where else can I bang my head? Perhaps there'a BIOS setting I should change?
If its a Fractal case, my experience is they almost always use 3 pin non-PWM fans. A PWM fan will have a 4 pin fan header, with 4 pins, with 4 cables going to the fan.

I would replace any fans that you want to control with PWM fans. I'm quite happy with Noctua.
 

glauco

Senior Member
Joined
Jan 30, 2017
Messages
338
My fans are all 4-pin PWM, I should have mentioned that. I have included them in my signature now.
Perhaps the problem has to do with the upper threshold values I've assigned to FAN1?
Code:
root@FreeNAS:~ # ipmitool sensor list all | grep -e FAN1 -e FANA -e "CPU Temp"
CPU Temp  | 37.000  | degrees C  | ok  | 0.000  | 0.000  | 0.000  | 95.000  | 100.000  | 100.000
FAN1  | 400.000  | RPM  | cr  | 300.000  | 500.000  | 700.000  | 2500.000  | 3000.000  | 3500.000
FANA  | 1100.000  | RPM  | ok  | 300.000  | 500.000  | 700.000  | 25300.000 | 25400.000 | 25500.000

Is there a way to reset those values (can't remember what they were initially)?
 

rovan

Member
Joined
Sep 30, 2013
Messages
32
Hi Glauco, I had problems initially with my setup, but it was all config based ^^. The IPMI needed to be reset to factory defaults to restore the default settings, you can do this from the IPMI administration. Then you need to completely power off (Important). After re-configuring the fan setting to standard or heavy, and resetting the sensor thresholds everything worked properly. I did have to modify the fan control script with a different low speed for the fans, as the Noctua fans run too slow for the Supermicro mainboard to detect (300RPM) ~30%, I set this to 32% ($hd_fan_duty_low) fixed my constant fan spin-up spin-down issue...

I can see your fan models Noctua (<3) so these thresholds might work, I'm using similar values. The upper values are set to some rediculous amount initally (22,000 RPM i think or 25,000 LOL) so it's not really useful.

The HDD NF-F12 Min 300 - Max 1500
The HDD NF-A14 Min 300 - Max 1500 (Fine to set the same :) )
The CPU FAN Min 400 - Max 2000
Code:
ipmitool sensor thresh "FANA" upper 1600 1700 1800
ipmitool sensor thresh "FAN1" upper 2100 2200 2300
ipmitool sensor thresh "FANA" lower 100 200 300
ipmitool sensor thresh "FAN1" lower 200 300 400


I'd set the cpu fan duty up in the fan control script to match your CPU fan, no harm in trying the default values though, just keep an eye on it carefully when you're testing it out.
 

Stux

Wizened Sage
Joined
Jun 2, 2016
Messages
4,163
You can use 0,0,200 for the low thresholds for the Noctua 300rpm fans and then you should be able to use 30% no problem.
 

glauco

Senior Member
Joined
Jan 30, 2017
Messages
338
Hi Glauco, I had problems initially with my setup, but it was all config based ^^. The IPMI needed to be reset to factory defaults to restore the default settings, you can do this from the IPMI administration. Then you need to completely power off (Important). After re-configuring the fan setting to standard or heavy, and resetting the sensor thresholds everything worked properly. I did have to modify the fan control script with a different low speed for the fans, as the Noctua fans run too slow for the Supermicro mainboard to detect (300RPM) ~30%, I set this to 32% ($hd_fan_duty_low) fixed my constant fan spin-up spin-down issue...
Thank you rovan, you nailed it!
The missing piece was resetting IPMI to factory defaults and switching off the PSU. Just to play it safe, I waited a couple of minutes before turning it back on. After that, stux's script magically worked as expected!
To be honest, it took me a while to figure out how to reset IPMI to factory defaults because I couldn't find how to do it in IPMI View.
Then I thought perhaps the web UI would allow me to do that... and so eventually I found it (Maintenance -> Factory Default).
Even changing ADMIN password works in the web UI but not in IPMI View, unless I'm doing something wrong...
 

Stux

Wizened Sage
Joined
Jun 2, 2016
Messages
4,163
Thought this was an interesting screenshot.

I've been building a new ESXi/FreeNAS AIO box, based on a Xeon D 1541 X10SDV-TLN4F in a Fractal Node 304, with six 7200rpm drives.

I've finished all the commissioning, except for tuning the fan control script.

I've now finished tuning the parameters to what I regard as optimal noise/temperature control and am running an extended heat soak torture test. The basic goal is to ensure that the box can run without overheating in a worst case scenario.

The script is running at boot already, and is reading CPU temps and controlling fan speeds via remote IPMI rather than local IPMI.

Screen Shot 2017-08-11 at 8.34.07 PM.png


Left pane is the fan control log, right panes are (top to bottom), solnet_array_tester, mprime, and a little script to output the cpu core temps.

This is currently running at an unrealistically maximum load, with maximum CPU burn. CPU temps are higher than I would like, but the CPU is rated for higher still. And those HDs idle at 39C.

This is currently maximum cooling as currently configured, and if its not sufficient, I'll have to revisit the cooling.

I'll check this periodically via remote over the weekend, and if it starts over-heating I'l shut it down.

Next steps would be to look at the exhaust fan.
 
Last edited:

Stux

Wizened Sage
Joined
Jun 2, 2016
Messages
4,163
Opened the case, and adjusted exhaust fan to maximum... Core temps dropped 2C and have been holding at 76C for a while now. HD temps are holding steady at 42C.

Screen Shot 2017-08-11 at 9.08.54 PM.png


The exhaust fan is currently non PWM, so since I don't like the noise it makes at full speed, I'll probably replace it with a Noctua PWM so it can be controlled via the HD fan loop.
 

Stux

Wizened Sage
Joined
Jun 2, 2016
Messages
4,163
Opened the case, and adjusted exhaust fan to maximum... Core temps dropped 2C and have been holding at 76C for a while now. HD temps are holding steady at 42C.

The exhaust fan is currently non PWM, so since I don't like the noise it makes at full speed, I'll probably replace it with a Noctua PWM so it can be controlled via the HD fan loop.
Been running for nearly 24 hours now. Stabilized at 39C for HDs and 74C for CPUs. Perfectly acceptable (once I cranked up the exhaust fan).

IMG_1137.PNG
 

walkman

Newbie
Joined
Nov 2, 2017
Messages
1
Then you need to completely power off (Important).
Wish I was reading your comment yesterday :D I have been wasting hours with this and got halfway through the thread. Turning off then on completely solved the issue. Seems like you have to reset the BMC if you want to apply the new fan profile...
 

Grinchy

Member
Joined
Aug 5, 2017
Messages
64
First of all, thank you for this great script!

Will this script wake up sleeping Harddrives? Or will it reset the "sleep timer"?

(For Example: Harrddrives are supposed to Sleep after 20min, but this Scripts checks them every 5min, so they can't go to sleep).


I'm using this script for a week now, but my Harddrives won't go to sleep anymore. Changed a lot of things last week, so I'm not sure where the problem lies. :-(
 

Kevin Horton

Neophyte Sage
Joined
Dec 2, 2015
Messages
679
First of all, thank you for this great script!

Will this script wake up sleeping Harddrives? Or will it reset the "sleep timer"?

(For Example: Harrddrives are supposed to Sleep after 20min, but this Scripts checks them every 5min, so they can't go to sleep).


I'm using this script for a week now, but my Harddrives won't go to sleep anymore. Changed a lot of things last week, so I'm not sure where the problem lies. :-(
This may be a consequence of the script asking the hard drives for their temperature every few minutes. Maybe that wakes the drives up.
 

Stux

Wizened Sage
Joined
Jun 2, 2016
Messages
4,163
Bingo. My drives don't sleep, so I don't know the impact on slumbering drives.
 

Grinchy

Member
Joined
Aug 5, 2017
Messages
64
Using this Script for some time now, but I have one big Problem.

I've put "$hd_max_allowed_temp" to 45°, but even at 42° it's still on "$hd_fan_duty_low = 30". With 43° it changes to "$hd_fan_duty_med_low = 45".

It didn't went over 43°, so i couldn't test wether it goes to "$hd_fan_duty_med_high" with 44°.

Code:
2018-01-21 22:11:57: CPU Temp: 32.0 <= 50, CPU Fan going low.

2018-01-21 22:11:57: Maximum HD Temperature: 43

2018-01-21 22:11:57: Drives are warming, going to 50%

2018-01-21 22:14:58: Maximum HD Temperature: 43

2018-01-21 22:18:00: Maximum HD Temperature: 42

2018-01-21 22:18:00: Drives are cool enough, going to 35%

2018-01-21 22:21:00: Maximum HD Temperature: 42

2018-01-21 22:24:01: Maximum HD Temperature: 42

2018-01-21 22:27:03: Maximum HD Temperature: 42

2018-01-21 22:30:03: Maximum HD Temperature: 42

2018-01-21 22:33:05: Maximum HD Temperature: 42



Shouldn't this be a little bit more flexible? Or is this normal? Would be perfect if $hd_fan_duty_med_low would start at ~40°.

Also, right now $hd_polling_interval is at 180, but i thought about putting it to 1300, so that my Drives can go to sleep. For cooling the Drives it's to much to long, but there seems no other way :-(


Here's the Script:

Code:
#!/usr/local/bin/perl

# This script is designed to control both the CPU and HD fans in a Supermicro X10 based system according to both
# the CPU and HD temperatures in order to minimize noise while providing sufficient cooling to deal with scrubs
# and CPU torture tests. It may work in X9 based system, but this has not been tested.

# It relies on you having two fan zones.

# To use this correctly, you should connect all your PWM HD fans, by splitters if necessary to the FANA header.
# CPU, case and exhaust fans should then be connected to the numbered (ie CPU based) headers.  This script will then control the
# HD fans in response to the HD temp, and the other fans in response to CPU temperature. When CPU temperature is high the HD fans.
# will be used to provide additional cooling, if you specify cpu/hd shared cooling.

# If the fans should be high, and they are stuck low, or vice-versa, the BMC will be rebooted, thus it is critical to set the
# cpu/hd_max_fan_speed variables correctly.

# NOTE: It is highly likely the "get_hd_temp" function will not work as-is with your HDs. Until a better solution is provided
# you will need to modify this function to properly acquire the temperature. Setting debug=2 will help.

# Tested with a SuperMicro X10-SRi-F, Xeon E5-1650v4, Noctua 120, 90 and 80mm fans in a Norco RPC-4224 4U chassis, with Seagate NAS drives.

# This script can be downloaded from : https://forums.freenas.org/index.php?threads/script-hybrid-cpu-hd-fan-zone-controller.46159/

# The script was originally based on a script by Kevin Horton that can be found at:
# https://forums.freenas.org/index.php?threads/script-to-control-fan-speed-in-response-to-hard-drive-temperatures.41294/page-3#post-282683

# More information on CPU/Peripheral Zone can be found in this post:
# https://forums.freenas.org/index.php?threads/thermal-and-accoustical-design-validation.28364/

# stux

# VERSION HISTORY
#####################
# 2016-09-19 Initial Version
# 2016-09-19 Added cpu_hd_override_temp, to prevent HD fans cycling when CPU fans are sufficient for cooling CPU
# 2016-09-26 hd_list is now refreshed before checking HD temps so that we start/stop monitoring devices that
#			have been hot inserted/removed.
#			"Drives are warm, going to 75%" log message was missing an unless clause causing it to print
#			every time
# 2016-10-07 Replaced get_cpu_temp() function with get_cpu_temp_sysctl() which queries the kernel, instead of
#			IPMI. This is faster, more accurate and more compatible, hopefully allowing this to work on X9
#			systems. The original function is still present and is now called get_cpu_temp_ipmi().
#			Because this is a much faster method of reading the temps,  and because its actually the max core
#			temp, I found that the previous cpu_hd_override_temp of 60 was too sensitive and caused the override
#			too often. I've bumped it up to 62, which on my system seems good. This means that if a core gets to
#			62C the HD fans will kick in, and this will generally bring temps back down to around 60C... depending
#			on the actual load. Your results will vary, and for best results you should tune controller with
#			mprime testing at various thread levels. Updated the cpu threasholds to 35/45/55 because of the improved
#			responsiveness of the get_cpu_temp function
###############################################################################################
## CONFIGURATION
################

## DEBUG LEVEL
## 0 means no debugging. 1,2,3,4 provide more verbosity
## You should run this script in at least level 1 to verify its working correctly on your system
$debug = 1;

## CPU THRESHOLD TEMPS
## A modern CPU can heat up from 35C to 60C in a second or two. The fan duty cycle is set based on this
$high_cpu_temp = 80;		# will go HIGH when we hit
$med_cpu_temp = 60;		 # will go MEDIUM when we hit, or drop below again
$low_cpu_temp = 40;		# will go LOW when we fall below 35 again

## HD THRESHOLD TEMPS
## HD change temperature slowly.
## This is the temperature that we regard as being uncomfortable. The higher this is the
## more silent your system.
## Note, it is possible for your HDs to go above this... but if your cooling is good, they shouldn't.
$hd_max_allowed_temp = 45;	# celsius. you will hit 100% duty cycle when you HDs hit this temp.

## CPU TEMP TO OVERRIDE HD FANS
## when the CPU climbs above this temperature, the HD fans will be overridden
## this prevents the HD fans from spinning up when the CPU fans are capable of providing
## sufficient cooling.
$cpu_hd_override_temp = 85;

## CPU/HD SHARED COOLING
## If your HD fans contribute to the cooling of your CPU you should set this value.
## It will mean when you CPU heats up your HD fans will be turned up to help cool the
## case/cpu. This would only not apply if your HDs and fans are in a separate thermal compartment.
$hd_fans_cool_cpu = 0;		# 1 if the hd fans should spin up to cool the cpu, 0 otherwise




#######################
## FAN CONFIGURATION
####################

## FAN SPEEDS
## You need to determine the actual max fan speeds that are achieved by the fans
## Connected to the cpu_fan_header and the hd_fan_header.
## These values are used to verify high/low fan speeds and trigger a BMC reset if necessary.
$cpu_max_fan_speed	 = 1500;
$hd_max_fan_speed	 = 1500;


## CPU FAN DUTY LEVELS
## These levels are used to control the CPU fans
$fan_duty_high	= 80;		# percentage on, ie 100% is full speed.
$fan_duty_med	 = 50;
$fan_duty_low	 = 30;

## HD FAN DUTY LEVELS
## These levels are used to control the HD fans
$hd_fan_duty_high	 = 80;	# percentage on, ie 100% is full speed.
$hd_fan_duty_med_high	 = 65;
$hd_fan_duty_med_low	= 45;
$hd_fan_duty_low	 = 30;	# some 120mm fans stall below 30.


## FAN ZONES
# Your CPU/case fans should probably be connected to the main fan sockets, which are in fan zone zero
# Your HD fans should be connected to FANA which is in Zone 1
# You could switch the CPU/HD fans around, as long as you change the zones and fan header configurations.
#
# 0 = FAN1..5
# 1 = FANA
$cpu_fan_zone = 0;
$hd_fan_zone = 1;


## FAN HEADERS
## these are the fan headers which are used to verify the fan zone is high. FAN1+ are all in Zone 0, FANA is Zone 1.
## cpu_fan_header should be in the cpu_fan_zone
## hd_fan_header should be in the hd_fan_zone
$cpu_fan_header = "FAN4";	
$hd_fan_header = "FANA";



################
## MISC
#######

## IPMITOOL PATH
## The script needs to know where ipmitool is
$ipmitool = "/usr/local/bin/ipmitool";

## HD POLLING INTERVAL
## The controller will only poll the harddrives periodically. Since hard drives change temperature slowly
## this is a good thing. 180 seconds is a good value.
$hd_polling_interval = 180;	# seconds

## FAN SPEED CHANGE DELAY TIME
## It takes the fans a few seconds to change speeds, we allow a grace before verifying. If we fail the verify
## we'll reset the BMC
$fan_speed_change_delay = 10; # seconds

## BMC REBOOT TIME
## It takes the BMC a number of seconds to reset and start providing sensible output. We'll only
## Reset the BMC if its still providing rubbish after this time.
$bmc_reboot_grace_time = 180; # seconds

## BMC RETRIES BEFORE REBOOTING
## We verify high/low of fans, and if they're not where they should be we reboot the BMC after so many failures
$bmc_fail_threshold	= 2;	 # will retry n times before rebooting

# edit nothing below this line
########################################################################################################################



# GLOBALS
@hd_list = ();

# massage fan speeds
$cpu_max_fan_speed *= 0.8;
$hd_max_fan_speed *= 0.8;


#fan/bmc verification globals/timers
$last_fan_level_change_time = 0;		# the time when we changed a fan level last
$fan_unreadable_time = 0;			# the time when a fan read failure started, 0 if there is none.
$bmc_fail_count = 0;				# how many times the fans failed verification in the last period.

#this is the last cpu temp that was read		
$last_cpu_temp = 0;

use POSIX qw(strftime);

# start the controller
main();

################################################ MAIN

sub main
{
	# need to go to Full mode so we have unfettered control of Fans
	set_fan_mode("full");
	
	my $cpu_fan_level = "";
	my $old_cpu_fan_level = "";
	my $override_hd_fan_level = 0;
	my $last_hd_check_time = 0;
	my $hd_fan_duty = 0;

	
	while()
	{
		$old_cpu_fan_level = $cpu_fan_level;
		$cpu_fan_level = control_cpu_fan( $old_cpu_fan_level );
		
		if( $old_cpu_fan_level ne $cpu_fan_level )
		{
			$last_fan_level_change_time = time;
		}

		if( $cpu_fan_level eq "high" )
		{
			
			if( $hd_fans_cool_cpu && !$override_hd_fan_level && ($last_cpu_temp >= $cpu_hd_override_temp || $last_cpu_temp == 0) )
			{
				#override hd fan zone level, once we override we won't backoff until the cpu drops to below "high"
				$override_hd_fan_level = 1;
				dprint( 0, "CPU Temp: $last_cpu_temp >= $cpu_hd_override_temp, Overiding HD fan zone to $hd_fan_duty_high%, \n" );
				set_fan_zone_duty_cycle( $hd_fan_zone, $hd_fan_duty_high );
				
				$last_fan_level_change_time = time;
			}
		}
		elsif( $override_hd_fan_level )
		{
			#restore hd fan zone level;
			$override_hd_fan_level = 0;
			dprint( 0, "Restoring HD fan zone to $hd_fan_duty%\n" );
			set_fan_zone_duty_cycle( $hd_fan_zone, $hd_fan_duty );	
			
			$last_fan_level_change_time = time;
		}

		# periodically determine hd fan zone level
		
		my $check_time = time;
		if( $check_time - $last_hd_check_time > $hd_polling_interval )
		{
			$last_hd_check_time = $check_time;
	
			# we refresh the hd_list from camcontrol devlist
			# everytime because if you're adding/removing HDs we want
			# starting checking their temps too!
			@hd_list = get_hd_list();
			
			my $hd_temp = get_hd_temp();
			$hd_fan_duty = calculate_hd_fan_duty_cycle( $hd_temp, $hd_fan_duty );
			
			if( !$override_hd_fan_level )
			{
				set_fan_zone_duty_cycle( $hd_fan_zone, $hd_fan_duty );

				$last_fan_level_change_time = time; # this resets every time, but it shouldn't matter since hd_polling_interval is large.
			}
		}
		
		# verify_fan_speed_levels function is fairly complicated		
		verify_fan_speed_levels(  $cpu_fan_level, $override_hd_fan_level ? $hd_fan_duty_high : $hd_fan_duty );
		
					
		# CPU temps can go from cool to hot in 2 seconds! so we only ever sleep for 1 second.
		sleep 1;

	} # inf loop
}

sub get_hd_list
{
	my $disk_list = `camcontrol devlist | sed 's:.*(::;s:).*::;s:,pass[0-9]*::;s:pass[0-9]*,::' | egrep '^[a]*da[0-9]+\$' | tr '\012' ' '`;
	dprint(3,"$disk_list\n");

	my @vals = split(" ", $disk_list);
	
	foreach my $item (@vals)
	{
		dprint(2,"$item\n");
	}

	return @vals;
}

sub get_hd_temp
{
	my $max_temp = 0;
	
	foreach my $item (@hd_list)
	{
		my $disk_dev = "/dev/$item";
		my $command = "/usr/local/sbin/smartctl -A $disk_dev | grep Temperature_Celsius";
		
		dprint( 3, "$command\n" );
		
		my $output = `$command`;

		dprint( 2, "$output");

		my @vals = split(" ", $output);

		# grab 10th item from the output, which is the hard drive temperature (on Seagate NAS HDs)
		  my $temp = "$vals[9]";
		chomp $temp;
		
		if( $temp )
		{
			dprint( 1, "$disk_dev: $temp\n");
			
			$max_temp = $temp if $temp > $max_temp;
		}
	}

	dprint(0, "Maximum HD Temperature: $max_temp\n");

	return $max_temp;
}

###########################
# verify_fan_speed_levels()
# this function verifies a fan zone is high, when it should be high, and low when it should be low.
# you pass in the cpu_fan_level and the hd_fan_duty (note: level vs duty!). If the hd fan duty is
# overridden, then you need to pass in the overridden duty.
#
# The tricks are that
#	1) we need to wait at least 10 seconds after changing a fan level before checking if its
#		made the change.
#	2) if we do read the change, and its not right, we should try again, after redoing the change
#	3) when the BMC has been reset, we can read rubbish... we shouldn't just re-set the BMC in this case
#	   as it should become good, but it also might not become good... in which case we should reset it!
#	4) if we do reset the BMC, we need to reverify
#	5) we don't want to re-verify continuously, so if its all good we wait an extra 60 seconds, unless
#	   the fans change in the meantime.
#
# to accomplish that, we use a few globals:
#
#	last_fan_level_change_time	this is the time when the last fan change was made, and should be updated,
#						 whenever a fan change is made. We also updated it each time through the
#					verify function so that we will not re-verify until our delay has expired
#
#	fan_unreadable_time		this is the time that the fan read failures started, or 0 if there are none
#					once the failure has exceeded the bmc_reboot_grace_time threshold, we will reboot.
#
#	bmc_fail_count			this is how many times in a row the fan speeds have not been what they should've
#					been. If we exceed bmc_fail_threshold, then we reboot the bmc.
#
# Configuration globals used:
#
#	bmc_fail_threshold		how many times the bmc can have the wrong fan speeds in a row before we reboot.
#	fan_speed_change_delay		how many seconds we wait until after fan change before verifying, and thence how
#					often we verify.
#	hd/cpu_max_fan_speed		depending on if we want high or low, if the fan speed is over or under, we'll regard
#					it as a failure.
#
sub verify_fan_speed_levels
{
	my( $cpu_fan_level, $hd_fan_duty ) = @_;
	dprint( 4, "verify_fan_speed_levels: cpu_fan_level: $cpu_fan_level, hd_fan_duty: $hd_fan_duty\n");

	my $extra_delay_before_next_check = 0;
	
	my $temp_time = time - $last_fan_level_change_time;
	dprint( 4, "Time since last verify : $temp_time, last change: $last_fan_level_change_time, delay: $fan_speed_change_delay\n");
	if( $temp_time > $fan_speed_change_delay )
	{
		# we've waited for the speed change to take effect.
		
		my $cpu_fan_speed = get_fan_speed("CPU");
		if( $cpu_fan_speed < 0 )
		{
			dprint(1,"CPU Fan speed unavailable\n" );
			$fan_unreadable_time = time if $fan_unreadable_time == 0;
		}
		
		my $hd_fan_speed = get_fan_speed("HD");
		if( $hd_fan_speed < 0 )
		{
			dprint(1,"HD Fan speed unavailable\n" );
			$fan_unreadable_time = time if $fan_unreadable_time == 0;
		}
		
		if( $hd_fan_speed < 0 || $cpu_fan_speed < 0 )
		{
			# one of the fans couldn't be reliably read

			my $temp_time = time - $fan_unreadable_time;
			if( $temp_time > $bmc_reboot_grace_time )
			{
				#we've waited, and we still can't read fan speed.
				dprint(0, "Fan speeds are unreadable after $bmc_reboot_grace_time seconds, rebooting BMC\n");
				reset_bmc();
				$fan_unreadable_time = 0;
			}
			else
			{
				dprint(2, "Fan speeds are unreadable after $temp_time seconds, will try again\n");	
			}		
		}
		else
		{
			# we have no been able to read the fan speeds

			my $cpu_fan_is_wrong = 0;
			my $hd_fan_is_wrong = 0;	
			
			#verify cpu fans
			if( $cpu_fan_level eq "high" && $cpu_fan_speed < $cpu_max_fan_speed )
			{
				dprint(0, "CPU fan speed should be high, but $cpu_fan_speed < $cpu_max_fan_speed.\n");
				$cpu_fan_is_wrong=1;
			}
			elsif( $cpu_fan_level eq "low" && $cpu_fan_speed > $cpu_max_fan_speed )
			{
				dprint(0, "CPU fan speed should be low, but $cpu_fan_speed > $cpu_max_fan_speed.\n");
				$cpu_fan_is_wrong=1;
			}
			
			#verify hd fans
			if( $hd_fan_duty >= $hd_fan_duty_high && $hd_fan_speed < $hd_max_fan_speed )
			{
				dprint(0, "HD fan speed should be high, but $hd_fan_speed < $hd_max_fan_speed.\n");
				$hd_fan_is_wrong=1;
			}
			elsif( $hd_fan_duty <= $hd_fan_duty_low && $hd_fan_speed > $hd_max_fan_speed )
			{
				dprint(0, "HD fan speed should be low, but $hd_fan_speed > $hd_max_fan_speed.\n");
				$hd_fan_is_wrong=1;
			}
			
			#verify both fans are good
			if( $cpu_fan_is_wrong || $hd_fan_is_wrong )
			{
				$bmc_fail_count++;
				
				dprint( 3, "bmc_fail_count:  $bmc_fail_count, bmc_fail_threshold: $bmc_fail_threshold\n");
				if( $bmc_fail_count <= $bmc_fail_threshold )
				{
					#we'll try setting the fan speeds, and giving it another attempt
					dprint(1, "Fan speeds are not where they should be, will try again.\n");

					set_fan_mode("full");

					set_fan_zone_level( $cpu_fan_zone, $cpu_fan_level );
					set_fan_zone_duty_cycle( $hd_fan_zone, $hd_fan_duty );
				}
				else
				{
					#time to reset the bmc
					dprint(1, "Fan speeds are still not where they should be after $bmc_fail_count attempts, will reboot BMC.\n");
					set_fan_mode("full");
					reset_bmc();
					$bmc_fail_count = 0;
				}
			}
			else
			{
				#everything is good. We'll sit back for another minute.

				dprint( 2, "Verified fan levels, CPU: $cpu_fan_speed, HD: $hd_fan_speed. All good.\n" );
				$bmc_fail_count = 0; # we succeeded

				$extra_delay_before_next_check = 60 - $fan_speed_change_delay; # lets give it a minute since it was good.
			}	

				
			#reset our unreadable timer, since we read the fan speeds.
			$fan_unreadable_time = 0;
									
		}
			
		#reset our timer, so that we'll wait before checking again.
		$last_fan_level_change_time = time + $extra_delay_before_next_check; #another delay before checking please.
	}
	
	return;
}

################################################# SUBS

# need to pass in last $cpu_fan
sub control_cpu_fan
{
	my ($old_cpu_fan_level) = @_;

#	my $cpu_temp = get_cpu_temp_ipmi();	# no longer used, because sysctl is better, and more compatible.
	my $cpu_temp = get_cpu_temp_sysctl();

	my $cpu_fan_level = decide_cpu_fan_level( $cpu_temp, $old_cpu_fan_level );

	if( $old_cpu_fan_level ne $cpu_fan_level )
	{
		dprint( 1, "CPU Fan changing... ($cpu_fan_level)\n");
		set_fan_zone_level( $cpu_fan_zone, $cpu_fan_level );	
	}

	return $cpu_fan_level;
}

sub calculate_hd_fan_duty_cycle
{
	my ($hd_temp, $old_hd_duty) = @_;
	my $hd_duty;


	if ($hd_temp >= $hd_max_allowed_temp  )
	{
		$hd_duty = $hd_fan_duty_high;
		dprint(0, "Drives are too hot, going to $hd_fan_duty_high%\n") unless $old_hd_duty == $hd_duty;
	}
	elsif ($hd_temp >= $hd_max_allowed_temp - 1 )
	{
		$hd_duty = $hd_fan_duty_med_high;
			dprint(0, "Drives are warm, going to $hd_fan_duty_med_high%\n") unless $old_hd_duty == $hd_duty;
		
	}
	elsif ($hd_temp >= $hd_max_allowed_temp - 2 )
	{
		$hd_duty = $hd_fan_duty_med_low;
		dprint(0, "Drives are warming, going to $hd_fan_duty_med_low%\n") unless $old_hd_duty == $hd_duty;
	}
	elsif( $hd_temp > 0 )
	{
		$hd_duty = $hd_fan_duty_low;
		  dprint(0, "Drives are cool enough, going to $hd_fan_duty_low%\n") unless $old_hd_duty == $hd_duty;
	}
	else
	{
		$hd_duty = 100;
		dprint( 0, "Drive temperature ($hd_temp) invalid. going to 100%\n");
	}
	
	return $hd_duty;
}

sub build_date_string
{
	my $datestring = strftime "%F %H:%M:%S", localtime;
	
	return $datestring;
}

sub dprint
{
	my ( $level,$output) = @_;
	
#	print( "dprintf: debug = $debug, level = $level, output = \"$output\"\n" );
	
	if( $debug > $level )
	{
		my $datestring = build_date_string();
		print "$datestring: $output";
	}

	return;
}

sub dprint_list
{
	my ( $level,$name,@output) = @_;
		
	if( $debug > $level )
	{
		dprint($level,"$name:\n");

		foreach my $item (@output)
		{
			dprint( $level, " $item\n");
		}
	}

	return;
}

sub bail_with_fans_full
{
	dprint( 0, "Setting fans full before bailing!\n");
	set_fan_mode("full");
	die @_;
}


sub get_fan_mode_code
{
	my ( $fan_mode )  = @_;
	my $m;

	if(	 $fan_mode eq	'standard' )	{ $m = 0; }
	elsif(	$fan_mode eq	'full' )	 { $m = 1; }
	elsif(	$fan_mode eq	'optimal' )	 { $m = 2; }
	elsif(	$fan_mode eq	'heavyio' )	{ $m = 4; }
	else					 { die "illegal fan mode: $fan_mode\n" }

	dprint( 3, "fanmode: $fan_mode = $m\n");

	return $m;
}

sub set_fan_mode
{
	my ($fan_mode) = @_;
	my $mode = get_fan_mode_code( $fan_mode );

	dprint( 1, "Setting fan mode to $mode ($fan_mode)\n");
	`$ipmitool raw 0x30 0x45 0x01 $mode`;

	sleep 5;	#need to give the BMC some breathing room

	return;
}	

# returns the maximum core temperature from the kernel to determine CPU temperature.
# in my testing I found that the max core temperature was pretty much the same as the IPMI 'CPU Temp'
# value, but its much quicker to read, and doesn't require X10 IPMI. And works when the IPMI is rebooting too.
sub get_cpu_temp_sysctl
{
	# significantly more efficient to filter to dev.cpu than to just grep the whole lot!
	my $core_temps = `sysctl -a dev.cpu | egrep -E \"dev.cpu\.[0-9]+\.temperature\" | awk '{print \$2}' | sed 's/.\$//'`;
	chomp($core_temps);

	dprint(3,"core_temps:\n$core_temps\n");

	my @core_temps_list = split(" ", $core_temps);
	
	dprint_list( 4, "core_temps_list", @core_temps_list );

	my $max_core_temp = 0;
	
	foreach my $core_temp (@core_temps_list)
	{
		if( $core_temp )
		{
			dprint( 2, "core_temp = $core_temp C\n");
			
			$max_core_temp = $core_temp if $core_temp > $max_core_temp;
		}
	}

	dprint(1, "CPU Temp: $max_core_temp\n");

	$last_cpu_temp = $max_core_temp; #possible that this is 0 if there was a fault reading the core temps

	return $max_core_temp;

}

# reads the IPMI 'CPU Temp' field to determine overall CPU temperature
sub get_cpu_temp_ipmi
{
	my $cpu_temp = `$ipmitool sensor get \"CPU Temp\" | awk '/Sensor Reading/{print \$4}'`;
	chomp $cpu_temp;

	dprint( 1, "CPU Temp: $cpu_temp\n");
	
	$last_cpu_temp = $cpu_temp; # note, this hasn't been cleaned.
	return $cpu_temp;
}

sub decide_cpu_fan_level
{
	my ($cpu_temp, $cpu_fan) = @_;

	#if cpu_temp evaluates as "0", its most likely the reading returned rubbish.
	if ($cpu_temp <= 0)
	{
		if( $cpu_temp eq "No")	# "No reading"
		{
			dprint( 0, "CPU Temp has no reading.\n");
		}
		elsif( $cpu_temp eq "Disabled" )
		{
			dprint( 0, "CPU Temp reading disabled.\n");
		}
		else
		{
			dprint( 0, "Unexpected CPU Temp ($cpu_temp).\n");
		}
		dprint( 0, "Assuming worst-case and going high.\n");
		$cpu_fan = "high";
	}
	else
	{
		if( $cpu_temp >= $high_cpu_temp )
		{
			if( $cpu_fan ne "high" )
			{
				dprint( 0, "CPU Temp: $cpu_temp >= $high_cpu_temp, CPU Fan going high.\n");
			}
			$cpu_fan = "high";
		}
		elsif( $cpu_temp >= $med_cpu_temp )
		{
			if( $cpu_fan ne "med" )
			{
				dprint( 0, "CPU Temp: $cpu_temp >= $med_cpu_temp, CPU Fan going med.\n");
			}
			$cpu_fan = "med";
		}
		elsif( $cpu_temp > $low_cpu_temp && ($cpu_fan eq "high" || $cpu_fan eq "" ) )
		{
			dprint( 0, "CPU Temp: $cpu_temp dropped below $med_cpu_temp, CPU Fan going med.\n");
			
			$cpu_fan = "med";
		}
		elsif( $cpu_temp <= $low_cpu_temp )
		{
			if( $cpu_fan ne "low" )
			{
				dprint( 0, "CPU Temp: $cpu_temp <= $low_cpu_temp, CPU Fan going low.\n");
			}
			$cpu_fan = "low";
		}
	}
		
	dprint( 1, "CPU Fan: $cpu_fan\n");

	return $cpu_fan;
}

# zone,dutycycle%
sub set_fan_zone_duty_cycle
{
	my ( $zone, $duty ) = @_;
	
	if( $zone < 0 || $zone > 1 )
	{
		bail_with_fans_full( "Illegal Fan Zone" );
	}

	if( $duty < 0 || $duty > 100 )
	{
		dprint( 0, "illegal duty cycle, assuming 100%\n");
		$duty = 100;
	}
		
	dprint( 1, "Setting Zone $zone duty cycle to $duty%\n");

	`$ipmitool raw 0x30 0x70 0x66 0x01 $zone $duty`;
	
	return;
}


sub set_fan_zone_level
{
	my ( $fan_zone, $level) = @_;
	my $duty = 0;
	
	#assumes high if not low or med, for safety.
	if( $level eq "low" )
	{
		$duty = $fan_duty_low;
	}
	elsif( $level eq "med" )
	{
		$duty = $fan_duty_med;
	}
	else
	{
		$duty = $fan_duty_high;
	}

	set_fan_zone_duty_cycle( $fan_zone, $duty );
}

sub get_fan_header_by_name
{
	my ($fan_name) = @_;
	
	if( $fan_name eq "CPU" )
	{
		return $cpu_fan_header;
	}
	elsif( $fan_name eq "HD" )
	{
		return $hd_fan_header;
	}
	else
	{
		bail_with_full_fans( "No such fan : $fan_name\n" );
	}
}

sub get_fan_speed
{
	my ($fan_name) = @_;
	
	my $fan = get_fan_header_by_name( $fan_name );

	my $command = "$ipmitool sdr | grep $fan";
	dprint( 4, "get fan speed command = $command\n");

	my $output = `$command`;
	  my @vals = split(" ", $output);
	  my $fan_speed = "$vals[2]";

	dprint( 3, "fan_speed = $fan_speed\n");


	if( $fan_speed eq "no" )
	{
		dprint( 0, "$fan_name Fan speed: No reading\n");
		$fan_speed = -1;
	}
	elsif( $fan_speed eq "disabled" )
	{
		dprint( 0, "$fan_name Fan speed: Disabled\n");
		$fan_speed = -1;

	}
	elsif( $fan_speed > 10000 || $fan_speed < 0 )
	{
		dprint( 0, "$fan_name Fan speed: $fan_speed RPM, is nonsensical\n");
		$fan_speed = -1;
	}
	else	
	{
		dprint( 1, "$fan_name Fan speed: $fan_speed RPM\n");
	}
	
	return $fan_speed;
}

sub reset_bmc
{
	#when the BMC reboots, it comes back up in its last fan mode... which should be FULL.

	dprint( 0, "Resetting BMC\n");
	`$ipmitool bmc reset cold`;
	
	return;
}




Different Question:
Would it be possible to get a Schedule into this Script? So that it only check the drives between 7:00 and 24:00.
 
Last edited:

Stux

Wizened Sage
Joined
Jun 2, 2016
Messages
4,163
Iirc max hd temp is the temp when the fans spin to max. Set your max temp lower and it will work to keep the drives below that temp

It polls every 3 minutes by default so the drives don’t heat up too fast.

You could put a time of day check in if you wanted.
 

Kevin Horton

Neophyte Sage
Joined
Dec 2, 2015
Messages
679
Using this Script for some time now, but I have one big Problem.

I've put "$hd_max_allowed_temp" to 45°, but even at 42° it's still on "$hd_fan_duty_low = 30". With 43° it changes to "$hd_fan_duty_med_low = 45".

...

Shouldn't this be a little bit more flexible? Or is this normal? Would be perfect if $hd_fan_duty_med_low would start at ~40°.
This script "simply" sets the fans at several discrete speeds as a function of warmest drive temperature. You may need to adjust the fan speed that is commanded at various temperatures to fine tune it for your system.

If you want a more flexible fan speed control, there are a couple of PID control loop scripts posted on the forums. These scripts have a target average disk temperature, and adjust the fan speed up or down as required to attempt to keep the average disk temperature on target.

Two PID control scripts:

No matter which type of controller you use, it will work better if you tweak the variables that define the controller to optimize it for your system.
 

Johnnie Black

Neophyte Sage
Joined
May 10, 2017
Messages
809
Many thanks for your script, been using it for a awhile.

Is it possible to ignore some disks, i.e., I have some SSDs and those can get hot quickly during heavy writes, they also cool down fast, but since they easily reach 45/50C the script makes all fans go to 100% for a few seconds.

P.S. yesterday updated the BMC firmware and completely forgot to shutdown the script, so during the update it lost communication with the BMC and reset it, this is not good during the update and the update failed leaving the IPMI completely nonfunctional, I though I might have bricked it, but to my relief re-flashing it with a DOS flash drive brought it back to life, I'm sharing this so you might consider adding a warning on the first post, this was totally my fault but it might happen to others and really brick the BMC.

Edit to add the log during the update.

Code:
2018-02-21 14:09:23: Maximum HD Temperature: 48
2018-02-21 14:09:23: Drives are too hot, going to 100%
2018-02-21 14:09:40: CPU Fan speed: Disabled
2018-02-21 14:09:47: HD Fan speed: Disabled
2018-02-21 14:10:05: CPU Fan speed: Disabled
2018-02-21 14:10:11: HD Fan speed: Disabled
2018-02-21 14:10:28: CPU Fan speed: Disabled
2018-02-21 14:10:35: HD Fan speed: Disabled
2018-02-21 14:10:52: CPU Fan speed: Disabled
2018-02-21 14:10:58: HD Fan speed: Disabled
2018-02-21 14:10:58: Fan speeds are unreadable after 120 seconds, rebooting BMC
2018-02-21 14:10:58: Resetting BMC


After the failed update:

Code:
Unable to send RAW command (channel=0x0 netfn=0x30 lun=0x0 cmd=0x70 rsp=0xc1): Invalid command
Get Device ID command failed: 0x2 Unknown (0x02)
Get Device ID command failed: 0x2 Unknown (0x02)
Unable to open SDR for reading
Get Device ID command failed: 0x2 Unknown (0x02)
Get Device ID command failed: 0x2 Unknown (0x02)
Unable to open SDR for reading
Could not open device at /dev/ipmi0 or /dev/ipmi/0 or /dev/ipmidev/0: No such file or directory
2018-02-21 14:40:24: CPU Temp: 31.0 <= 40, CPU Fan going low.
Could not open device at /dev/ipmi0 or /dev/ipmi/0 or /dev/ipmidev/0: No such file or directory
2018-02-21 14:40:25: Maximum HD Temperature: 41
2018-02-21 14:40:25: Drives are cool enough, going to 20%
Could not open device at /dev/ipmi0 or /dev/ipmi/0 or /dev/ipmidev/0: No such file or directory
Could not open device at /dev/ipmi0 or /dev/ipmi/0 or /dev/ipmidev/0: No such file or directory
Could not open device at /dev/ipmi0 or /dev/ipmi/0 or /dev/ipmidev/0: No such file or directory
Could not open device at /dev/ipmi0 or /dev/ipmi/0 or /dev/ipmidev/0: No such file or directory
Could not open device at /dev/ipmi0 or /dev/ipmi/0 or /dev/ipmidev/0: No such file or directory
Could not open device at /dev/ipmi0 or /dev/ipmi/0 or /dev/ipmidev/0: No such file or directory
Could not open device at /dev/ipmi0 or /dev/ipmi/0 or /dev/ipmidev/0: No such file or directory
 
Last edited:

Kevin Horton

Neophyte Sage
Joined
Dec 2, 2015
Messages
679
Many thanks for your script, been using it for a awhile.

Is it possible to ignore some disks, i.e., I have some SSDs and those can get hot quickly during heavy writes, they also cool down fast, but since they easily reach 45/50C the script makes all fans go to 100% for a few seconds.
Yes, it is possible to modify the script to ignore SSDs. I did that exact mod with my adaptation of @Stux's script. On my version of the script, I modified the code in sub get_hd_list to remove SSDs. On my system, the following line works in that subroutine (I added the 'grep -v "SSD" ' to remove lines with "SSD" from the disk list):

Code:
	my $disk_list = `camcontrol devlist | grep -v "SSD" | sed 's:.*(::;s:).*::;s:,pass[0-9]*::;s:pass[0-9]*,::' | egrep '^[a]*da[0-9]+\$' | tr '\012' ' '`;



We need to find some output of camcontrol devlist that differs between your SSDs and the spinning disks. What is the result of camcontrol devlist?
 
Top