[Discuss] Thermal monitoring

[Rich Braun]

Jerry Feldman asked:
> But, I don't have any ACPI support: /proc/acip/thermal_zone is empty.
> I'm looking for a quick and dirty way to check to see if the unit temps are
> ok since this system is our Oracle server.

My experience on this topic has been woeful. Support for sensors is not
standardized and remains very much dependent on which CPU and which
motherboard you have.  Most distros support a package "sensors" (sometimes
named "lm-sensors") to provide the command-line tool, but once you have that
then you need something else (in my case, a cobbled-together home-brew script)
to feed it into anything from Nagios to Cacti to munin to a simple
email-when-it-gets-hot script.

Example on one of my own servers (a Core I5-760).  Absolutely no luck getting
lm-sensors working:

 # ls /proc/acpi/thermal_zone/
 # sensors
 No sensors found!
 Make sure you loaded all the kernel drivers you need.
 Try sensors-detect to find out which these are.
 # sensors-detect
 # sensors-detect revision 5818 (2010-01-18 17:22:07 +0100)
  ...
 Now follows a summary of the probes I have just done.
 Just press ENTER to continue:

 Driver `it87':
  * ISA bus, address 0x290
    Chip `ITE IT8720F Super IO Sensors' (confidence: 9)

 Do you want to overwrite /etc/sysconfig/lm_sensors? (YES/no):
 # modprobe i2c-dev
 # sensors
 No sensors found!
 ---

But on a newer Core I5-2500K, it all works fine:

 # sensors
 coretemp-isa-0000
 Adapter: ISA adapter
 Core 0:      +37.0°C  (high = +80.0°C, crit = +98.0°C)
  ...
 Core 3:      +35.0°C  (high = +80.0°C, crit = +98.0°C)

So, alas, I find myself just tossing out motherboards that don't cooperate (if
the machine is mission-critical).  Not sure how to determine prior to buying
one whether it will.

-rich