NVIDIA/AMD temp monitoring script

As requested in another thread.

This script runs under Ubuntu 12.04. Should not be an issue on other distros.

The script name is: sys_temp.pl
Place this in your $HOME/scripts/temp directory. You can put it anywhere
you want but changes within the script will have to be made to reflect this
change.

I have modified this script for posting to exclude my actual node names, but
will be using the following for examples: lunar1_raspi, lunar2, lunar3, lunar4.
In the scripts change these to your node names/hostnames.

There are embedded debug statements in the form of "print". These are
bracketed ({}) by ($dbg) statements. setting my $dbg=1 will cause them to
print out. Make sure debug is not "on" when running as a cron job.

This script is written in perl and provides audible alarms and email when one
of the following events occur:
1. a NVIDIA GPU violates an arbitrary temperature
2. an AMD GPU violates an arbitrary temperature
3. a RaspberryPi CPU violates an arbitrary temperature

Support for CPU temps does not exist except for the Pi.

Some items to note:

1. These arbirtrary temperatures are defined in the script using these
variables:
my $violation_temp_pi = 38.0;
my $violation_temp_nvidia = 64.0;
my $violation_temp_amd = 60.0;

These values can be changed to whatever values suits your location.

2. You will have to find some "free" sound files on the web for your
audible alarms. Each machine should have a unique sound file.
The ones I use are definded in the script as follows:

if ($node_name eq 'lunar1'){
$alarm_sound = "$HOME/scripts/temp/sonar.mp3";
$send_to = 'your_user@mail.lunar1'; }
elsif ($node_name eq 'lunar2'){
$alarm_sound = "$HOME/scripts/temp/romulan_alarm.mp3";
$send_to = 'your_user@mail.lunar2';
}

I would sugges using the full path in the above rather then "$HOME"

NOTE: your /etc/hosts will have to have "mail.lunar2", etc.
in the alias
portion of your hosts entries. True for all other hosts.

3. Please read the comments at the top of the script for prerequisites.

4. In the script you will notice a reference to "/usr/sbin/sendmail". This
is because when you install "postfix" it will generate "/usr/sbin/sendmail"
as a softlink to postfix. This might be an issue if your distro handles
this differently.

5. Would be nice to also monitor CPU temps. This requires accessing the
bios and I have not yet been successful using the "sensor" command or
the "dmidecode" command. This might not be as "easy" as getting the
the GPU temps - bios' vary and so do distribution tools.

SCRIPT follows:

#!/usr/bin/perl -w

#
# sys_temp.pl - this perl script
#
# script function: a. send email when GPU temp exceeds a set value on a non-raspberry pi.
# In case of NVIDIA and AMD GPUs fan speed as a percentage will
# also be sent.
# b. send email when CPU temp exceeds a set value on a raspberry pi
#
# will be run as a cronjob every X minutes - you decide - see below for format
# initially the violation temps that trigger alarms/email will be an arbitrary value
#
# history:
# 080513 - initial write of script
# 080713 - made into a generic module which supports Linux Pi and non Pi Linux machines
# Temps on Pi are CPU temps
# Temps on non Pi machines are GPU temps. CPU temps are not provided.
# 081013 - Added alarm logic for Non Pi Linux. Pi Linux seems to have issues with sound
# 022214 - changed to postfix and away from ssmtp
# 070714 - rewrite to include AMD GPU
# supported GPUs: AMD and NVIDIA
# supported CPUs: PI only
# the sys_temp.pl script is expected to reside in $HOME/scripts/temp/
#
# prerequisites: sudo apt-get install perl $violation_temp){

#
# Pi mail starts here
#

if ($is_vcgencmd_present){ #must be Pi
#send email

open(MAIL, "|$email_program -t") || die "Can't open mail program $!\n";

# Mail Header

print MAIL "To: $send_to\n";
print MAIL "From: $mail_from\n";
print MAIL "Subject: $node_name $email_subject\n\n";

print MAIL <

current CPU temp is: $current_celsius_temp$celsius_symbol ($current_fahrenheit_temp$fahrenheit_symbol) - which violates the max operating temp of: $violation_temp$celsius_symbol.

EOF
close MAIL;

}else{ # must be GPU machine, i.e., all other machines employ either NVIDIA or AMD GPUs

#
# Non Pi mail starts here
#
open(MAIL, "|$email_program -t") || die "Can't open mail program $!\n";

# Mail Header

print MAIL "To: $send_to\n";
print MAIL "From: $mail_from\n";
print MAIL "Subject: $node_name $email_subject\n\n";

print MAIL <

current GPU temp is: $current_celsius_temp$celsius_symbol ($current_fahrenheit_temp$fahrenheit_symbol) - which violates the max operating temp of: $violation_temp$celsius_symbol.

current fan speed is: $fan_value$percent_symbol

EOF
close MAIL;

}

#
# sound alarm
# based upon hostname
# each host has a unique sound
#

system("/usr/bin/mpg123", "-q", "--loop", "5", $alarm_sound);
}
END SCRIPT

hope the cut and paste of the file is clean. I checked it and it seems fine but ....

You will need to install postfix NOT sendmail and postfix will need to be configured for email to work.

Once you make changes to the script for your hostnames you can use some mp3 file you can find on your distro for initial checkout.

Hope this helps.