- Integrated Systems
- About Us
- Integrated Systems
- About Us
06-25-2006 02:03 AM
I have a ML150 G2 SATA server with Debian Linux installed. I want to monitor hw sensors data - temperatures & fan speed (and also voltages).
As far as I know, there is a Winbond w83792d chip in this server. After installing kernel module for this sensors, the "sensors" program reports correct data. But compared to values in BIOS setup, I have found that one fan and one temperature is missing in linux.
It will output only 2 fans (I have 3 fans installed - cpu, front and back - and BIOS displays all of them). Also it will output only 3 temperatures - two of them are for CPU0 and CPU1 (not present on my system, so the values is bogus) and one "system" temperature. But again BIOS has one more temperature.
Is there any other sensors chip in this server? Or any other way to monitor the temperatures?
And do you know where are the temp sensors actually located on the motherboard? One temp is around 58degress, the other is around 36degress.
Solved! Go to Solution.
06-25-2006 08:35 AMSolution
There may be several sensor chips in that server - the w83792d chip has only three temperature sensor inputs, so the fourth temperature reading must come from elsewhere.
At a hunch, I would try adm10?? sensor drivers - the "hplog" tool (included in the hpasm driver package for RedHat ES4) reports that DL385 and DL585 servers have a handful of sensor chips of those types. Basic logistics would suggest that the same chip might be in use in other models too.
The names of the temperature outputs are almost arbitrary: they may be correct according to how the sensors are wired on w83792d driver developer's motherboard, but on other motherboards the chip may be wired completely different.
Check out the /etc/sensors.conf file: the file specifies the names for the sensors and allows to "hide" those inputs that are not used. The default file may need some changes for ML150 G2.
Looks like the w83792d chip has input channels for seven fans, so your missing third fan might be connected to a fan channel that is masked out by the default configuration. Or the missing fan might be connected to the same chip that provides the fourth temperature reading, whatever that is.
The temperature of 58 degrees (Celsius, I presume) sounds about typical for CPU temperature. Many new CPU models do have an integrated temperature sensor inside the actual CPU chip. It's just a single diode with documented thermal characteristics. It is connected to the motherboard's sensor chip through the CPU socket.
06-25-2006 08:52 PM
Re: ML150 G2 sensors data
Yes, I have tried sensors-detect long time ago when I was installing the server. It found 2 sensors, but for one of them (w83792d) there was no driver yet and the second (w83627thf) is giving bogus results.
I tried it again and with the same result. The readings of w83627thf are bogus. Readings of w83792d seem to be similar to what BIOS says, except that some values are missing.
Adapter: ISA adapter
VCore: +0.73 V (min = +0.70 V, max = +1.87 V)
+12V: +12.10 V (min = +3.16 V, max = +7.78 V)
+3.3V: +3.18 V (min = +0.53 V, max = +0.10 V)
+5V: +5.04 V (min = +0.00 V, max = +4.64 V)
-12V: -14.91 V (min = -14.91 V, max = -14.91 V)
V5SB: +5.11 V (min = +0.00 V, max = +0.00 V)
VBat: +0.61 V (min = +0.00 V, max = +1.15 V)
fan1: 3054 RPM (min = 84375 RPM, div = 2)
CPU Fan: 0 RPM (min = -1 RPM, div = 2)
fan3: 0 RPM (min = 10546 RPM, div = 4)
M/B Temp: +41 C (high = +0 C, hyst = +0 C) sensor = thermistor
CPU Temp: +32.0 C (high = +80 C, hyst = +75 C) sensor = thermistor
temp3: +50.0 C (high = +80 C, hyst = +75 C) sensor = thermistor
vid: +0.275 V (VRM Version 9.0)
Sound alarm disabled
Adapter: SMBus I801 adapter at 1100
VCoreA: +1.31 V (min = +1.10 V, max = +1.45 V)
VCoreB: +0.53 V (min = +1.10 V, max = +1.45 V) ALARM
VIN0: +1.21 V (min = +1.41 V, max = +1.60 V) ALARM
VIN1: +2.59 V (min = +1.41 V, max = +1.60 V) ALARM
VIN2: +1.52 V (min = +1.41 V, max = +1.60 V)
VIN3: +3.42 V (min = +1.41 V, max = +1.60 V) ALARM
5VCC: +4.60 V (min = +4.34 V, max = +5.06 V)
5VSB: +4.58 V (min = +4.34 V, max = +4.99 V)
VBAT: +3.12 V (min = +3.01 V, max = +3.41 V)
Fan1: 2766 RPM (min = 753 RPM, div = 8)
Fan2: 0 RPM (min = 753 RPM, div = 8)
Fan3: 1607 RPM (min = 672 RPM, div = 8)
Fan4: 0 RPM (min = 0 RPM, div = 1)
Fan5: 0 RPM (min = 0 RPM, div = 2)
ERROR: Can't get FAN6 data!
ERROR: Can't get FAN7 data!
Temp1: +54.0 C (high = +60.0 C, hyst = +1.0 C)
Temp2: +2.0 C (high = +80.0 C, hyst = +75.0 C) ALARM
Temp3: +56.0 C (high = +80.0 C, hyst = +75.0 C) ALARM
chassis: Chassis intrusion detection ALARM
I have identified that Temp1 (on w83792d) is CPU temp - because of fast changes when CPU is load is full. I just wanted to know where the other temp sensor is physically located when it is also quite hot.
I just tried to run sensors-detect on our ML110 G2 and it found no chips. When I was installing it, I was googling all chip IDs found on the motherboard to determine which is sensors chip :) and then I found kernel module for it (it is smsc47b397). But both servers are in datacenter now, so it's not possible to look closely at all chips inside... So maybe there is one more chip, but this is undetectable by sensors-detect. Another option is to try to load all available sensors modules, but I don't know if this is safe enough on a production machine.