Health Monitoring with lm_sensors

A fairly simple tool that can catch hardware problems before they become disasters.

A few years ago, a friend of mine had problems with his computer. It became unreliable; he would get odd errors, or the system would hang. He re-installed the operating system and his key applications, but the problems persisted. This cost him time, and his time is valuable.

The problem turned out to be simple: the cooling fan on his CPU's heat sink was dead. The overheating CPU, in turn, caused random problems. He wound up replacing the CPU as well as the heat sink, because the CPU had been damaged.

He could have avoided all of this if he had been running some sort of computer health monitoring system, such as lm_sensors.

lm_sensors is a set of Linux kernel modules for monitoring the vital signs of a computer: the voltages from the power supply, the temperature of the system and the CPU and fan speeds. lm_sensors includes a command-line utility for checking the current readings. A variety of graphical tools also are available for putting a pretty face on lm_sensors.

Installing lm_sensors

Before you attempt to install the lm_sensors kernel module, you should have the I2C driver modules installed. I2C, which stands for Inter-IC bus, is a simple serial data system for connecting chips so they can talk to one another. Most motherboards with health monitoring features use an I2C bus to access those features.

Make sure the I2C drivers are installed on your system. (If you build your own kernel, the I2C options are located under Character Devices.)

Install the packages that provide lm_sensors, or build it from source code and install. Then enter sensors-detect, a script that figures out how to install lm_sensors on your system. It will try various I2C modules and then try various lm_sensors modules, until it finds a combination that works on your system. When it is done, it provides instructions for how to set up configuration files in /etc, which load the correct modules when your system boots.

Once the modules for lm_sensors are installed and working, you can run the sensors command from a shell and receive some useful output. But you probably aren't done yet.

Customizing lm_sensors

Your next step is to edit the /etc/sensors.conf file. This file sets some custom parameters that make lm_sensors work with your computer system. For example, you can add a label that changes Temp1 to CPU Temp; you can disable Temp3 completely if you don't have a sensor and it is reporting nonsense; and you can customize the math functions used to calculate the displayed values.

Ideally, before you edit /etc/sensors.conf, you should reboot your computer and enter the BIOS setup screens. (For most computers, you hit the Del key or the F1 key during bootup to enter the BIOS. Check the owner's manual for your system, or watch the screen during bootup for a message like Hit <Del> to enter Setup.) The BIOS setup should have a sub-menu showing the same numbers you would like lm_sensors to report. Make a note of the readings you are seeing. For example, if the CPU temperature is about 60 degrees Celsius, write that down.

Now, booted back into Linux, run the sensors command. If the numbers are all there and look correct, you are done. If not, you need to customize /etc/sensors.conf.

Sample sensors Output

Take a look at the top of the output of the sensors command; it shows you with which chip lm_sensors is communicating. Find that chip in /etc/sensors.conf and look at the settings you can customize. /etc/sensors.conf is liberally commented, making it easier to figure out what you need to do. You also can read the man page, man sensors.conf.

Once you have customized /etc/sensors.conf, you must run the command sensors -s to put your changes into effect. Then run the sensors command once again to inspect the values it reports.

Additionally, you should make sure sensors -s runs each time your system boots.


You have the option to install sensord, the lm_sensors monitoring dæmon. You can configure this to log either to standard log files or to a round-robin database (a constant-size database set up to hold, for example, a week's worth of readings; new readings overwrite the oldest). By editing /etc/syslog.conf, you can arrange to receive e-mail when sensor readings go out of bounds.



Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

Do we need to have health monitoring sensors installed.

Anonymous's picture

Hi all,

Apart from our normal PC, do we need additional hardware, to use this.
to get status of PC health.


Lm_sensors; Adam Curtis

Anonymous's picture

Sensors in most cases would have been useful when my computer malfunctioned. What I would like to know is are there sensors for video cards. GPU's like to overheat somtimes.

Re: Lm_sensors; Adam Curtis

Anonymous's picture

the S3 savage4 has one. There's a module for Matrox g400, so I think it does it too.

Moaning Goat Meter

steveha's picture

I really, really wanted to mention the Moaning Goat Meter in this article. I didn't because it was too much of a stretch (since MGM doesn't yet support lm_sensors). But check it out:

Do not miss the FAQ!

If you are a Perl hacker, maybe you can add lm_sensors support to MGM. MGM is crying out for it (or moaning for it, perhaps).

If you use KDE, you can monitor lm_sensors with KSysGuard

Anonymous's picture

You may need to add sensors to the list of inputs in the /etd/ksysguard.conf file, but it it will allow the hardware sensor status to be displayed in the KDE 'panel'.

It's not as pretty as GKrellM, though. It also won't do logarithmic scaling (useful for things like network data traffic). I've still to investigate if GKrellM can do that either.

warning to Thinkpad owners

Anonymous's picture

Be very careful with LMsensors if you have a IBM ThinkPad laptop. There is a risk of damaging it with these programs.

for more information see: