Monitoring CPU temperatures on a Raspberry Pi with MQTT

by Posted on

Recently I’ve had problems with my Raspberry Pi 3 overheating though I use a heat sink for the processor and have a very modest load on the machine. When the RPi is in this state, it shows a thermometer warning icon and it is not possible to login. As I can not access the machine, it is hard to investigate the cause of the heat problem (if a process has gone totally wild e.g.). The only way to resolve this is to do a hard reboot (then the temperature goes down again).

To investigate this further, I want to monitor the CPU temperature without accessing the RPi via ssh or a direct login. My idea is to let the RPi gather board temperature values regularly and then publish these via MQTT. An MQTT subscriber will see to that the measurements are propagated to a cloud service so that I can monitor the values in an external application. With IFTTT I can add alerts on the measurements (e.g. when the CPU temperature goes over 60 C, an e-mail should be sent to me). This way I will be notified before the temperature/load gets too high and I will hopefully be able to intercept the problem by logging in to the machine before it locks up.

Solution overview

I will add a Python script that is scheduled with cron to start every ten minutes. The script checks the CPU temperature, formats the data and publishes the value via a local MQTT broker (mosquitto). I use an existing Python service (developed in A self-hosted MQTT environment for Internet of Things – Part 3) as an MQTT subscriber that forwards the data to a cloud service (Adafruit IO).

The data can then be monitored via the Adafruit IO web dashboard or via a mobile app (I use DataFeeds on iOS which works fine for my purposes).

I have setup an rule in IFTTT (IF This Then That), that sends me an iOS notification when the measurement value exceeds a certain limit.

Getting the CPU temperature from a Raspberry Pi

The vcgencmd command can be used for getting Raspberry Pi-specific properties. To get the broadcom chip temperature, the measure_temp argument is used:

vcgencmd measure_temp

This will return a string like Temp=42.4’C.

Setting up the measurement Python script

As I want to publish the temperature, I use a Python script to execute the vcgencmd and then format it. This tutorial shows how it can be done:

https://www.raspberrypi.org/learning/temperature-log/worksheet/

As in the tutorial, I use subprocess.check_output() to execute the command and then a regular expression to extract the actual value from the returned string.


#!/usr/bin/env python3
from subprocess import check_output
from re import findall
def get_temp():
temp = check_output(["vcgencmd","measure_temp"]).decode("UTF-8")
return(findall("\d+\.\d+",temp)[0])
print(get_temp())

view raw

printcputemp.py

hosted with ❤ by GitHub

After the temperature value is fetched, I connect to my local MQTT broker and publish a message on the Home/RPI3/Temp topic.


#!/usr/bin/env python3
import paho.mqtt.publish as publish
from subprocess import check_output
from re import findall
def get_temp():
temp = check_output(["vcgencmd","measure_temp"]).decode("UTF-8")
return(findall("\d+\.\d+",temp)[0])
def publish_message(topic, message):
print("Publishing to MQTT topic: " + topic)
print("Message: " + message)
publish.single(topic, message, hostname="192.168.1.16")
temp = get_temp()
publish_message("Home/RPI3/Temp", temp)

view raw

measurecputemp.py

hosted with ❤ by GitHub

The script is setup to run once every 10 minutes with cron:

crontab -e

(add a row like this one):

*/10 * * * * /home/myuser/mypath/measurecputemp.py

Note that the py-file needs to start with a Python shebang (#!/usr/bin/env python3) and the file needs to have execution rights set (use chmod ugo+x measurecputemp.py).

Routing the data to Adafruit IO

The temperature is published as a message to my local MQTT broker. I have a subscriber that acts as proxy for forwarding the data to Adafruit IO. I use this subscriber for most of the messages that I publish from my MQTT devices at home. I have described this subscriber in more detail in a previous post:

A self-hosted MQTT environment for Internet of Things – Part 3

This Python script is setup to run as a service as described in:

Daemonize that Python script!


#!/usr/bin/env python3
import paho.mqtt.client as mqtt
import datetime
import time
from Adafruit_IO import MQTTClient
from adafruit_credentials import ADAFRUIT_IO_USERNAME, ADAFRUIT_IO_KEY
from mylogger import logger
#
# Subscribes to messages via a local MQTT broker
# and forwards them to a cloud service (Adafruit IO)
#
def on_connected(client, userdata, flags, rc):
print("Connected to local MQTT broker with result code "+str(rc))
client.subscribe("Home/#")
def on_disconnected(client,userdata,rc):
print("Disconnected from local MQTT broker")
try_connect_to_local_broker(client)
def adafruit_connected(client):
print("Connected to Adafruit IO")
def on_message(client, userdata, msg):
print(str(datetime.datetime.now()) + ": " + msg.topic + " " + str(msg.payload))
#
# Forward the data to Adafruit IO. Replace topic with a valid feed name
#
feedname=msg.topic.replace("/","_")
print("Publish to Adafruit feedname: " + feedname)
# Initialize the client that should connect to io.adafruit.com
adafruitClient = MQTTClient(ADAFRUIT_IO_USERNAME, ADAFRUIT_IO_KEY,service_port=1883)
adafruitClient.on_connect = adafruit_connected
adafruitClient.connect()
adafruitClient.loop()
adafruitClient.publish(feedname,msg.payload)
adafruitClient.disconnect()
def try_connect_to_local_broker(client):
print("trying to connect to local broker")
connOK=False
while(connOK == False):
try:
client.connect("192.168.1.16", 1883, 60)
connOK = True
except:
connOK = False
time.sleep(2)
while(True):
#
# Initialize the client that should connect to the local MQTT broker
#
try:
client = mqtt.Client()
client.on_connect = on_connected
client.on_disconnect = on_disconnected
client.on_message = on_message
try_connect_to_local_broker(client)
# Blocking loop to the local Mosquitto broker
client.loop_forever()
except:
print("Failed connection to local MQTT broker");
time.sleep(10)

view raw

subscriber_adafruit_proxy.py

hosted with ❤ by GitHub

The data can then be monitored on the Adafruit IO dashboard:

Rpitemp_adafruitio.jpg

Monitoring the data with a mobile app

I use the free and excellent DataFeeds app by MonoHelix Labs for viewing my Adafruit IO feeds on iOS. You can get it from here:

https://itunes.apple.com/us/app/datafeeds-feed-monitor-for/id1163614771?mt=8

datafeedsrpitemp

Setting up an applet/rule with IFTTT

With an account on IFTTT, it is possible to connect to your Adafruit IO account and act on changes in the feed data. I have setup an applet that gives me notifications on my mobile phone when the RPi-temperature is greater or equal to 60 C.

IFTTTapplet.png

Adding additional measurements

While at it, I add measurements for disk-, memory and CPU-usage as well. These measurements can be fetched with the psutil library. I have made a new Python script that gets these values (and the CPU temp) and transforms them to MQTT messages. The script takes one argument that can be specified as the name of the computer that is measured, e.g. RPI3, RPI2. This way I can easily use the script on all my Raspberry Pi:s.


#!/usr/bin/env python3
import paho.mqtt.publish as publish
from subprocess import check_output
from re import findall
import psutil
import sys
def get_temp():
temp = check_output(["vcgencmd","measure_temp"]).decode("UTF-8")
return(findall("\d+\.\d+",temp)[0])
def get_disk_usage():
return str(psutil.disk_usage('/').percent)
def get_memory_usage():
return str(psutil.virtual_memory().percent)
def get_cpu_usage():
return str(psutil.cpu_percent(interval=None))
def publish_message(topic, message):
print("Publishing to MQTT topic: " + topic)
print("Message: " + message)
publish.single(topic, message, hostname="192.168.1.16")
if __name__ == '__main__':
if len(sys.argv) != 2:
print("Specify a computer name as argument to " + __file__)
sys.exit(2)
computer_name = sys.argv[1]
print("Doing measurements for: " + computer_name)
publish_message("Home/" + computer_name + "/Temp", get_temp())
publish_message("Home/" + computer_name + "/DiskUsagePercent", get_disk_usage())
publish_message("Home/" + computer_name + "/MemoryUsagePercent", get_memory_usage())
publish_message("Home/" + computer_name + "/CpuUsagePercent", get_cpu_usage())

view raw

systemmonitor.py

hosted with ❤ by GitHub

To use this script with cron, set execution rights on the file and specify the schedule with crontab:

*/10 * * * * /home/myuser/mypath/systemmonitor.py RPI3

 

Published by Lars

10 Thoughts

  1. Good morning!

    There is something that I didn’t understand: The script that sends the data do Adafruit is called periodically as well as the script that collect the data? Or it stays running in background always?

    Luiz

    Like

    Reply

    1. Hi Luiz,
      The script that collects data is scheduled with cron to be executed every 10 minutes. It publishes the data to a local MQTT broker. The MQTT broker is always running (as a service). I have several MQTT subscribers that are also always running (as services) and they listen to messages sent to the broker. One of the services filters out specific MQTT topics and sends the message data to Adafruit IO.
      I have a separate post that describes how to make the Python scripts run as services on Raspberry Pi: https://larsbergqvist.wordpress.com/2016/06/18/daemonize-that-python-script/

      Like

      Reply

  2. Thanks for the script.
    Worked great for me, except that the cpu load was every time almost 100% on my raspberry 2B.
    This because python was using at the moment the measuring was done a lot of cpu himself.
    Did some sleeping so the cpu can settle down again and now the load is most of the time 4%, that will match with TOP.

    time.sleep(5)
    cu = get_cpu_usage()

    publish_message(“Home/” + computer_name + “/Temp”, get_temp())
    publish_message(“Home/” + computer_name + “/DiskUsagePercent”, get_disk_usage())
    publish_message(“Home/” + computer_name + “/MemoryUsagePercent”, get_memory_usage())
    publish_message(“Home/” + computer_name + “/CpuUsagePercent”, cu)

    Like

    Reply

  3. Hello! I want to read temp at one RPi and send it to the another Rpi – where I have HomeAssistant. Can you please tell me how to send the data? I have done the part printcputemo + measurecputemp. BIG THANKs!

    Like

    Reply

    1. Hi Filip! You can use the paho mqtt library in Python as described in the last listing in this blog post (systemmonitor.py). Change the hostname in the publish.single-call to an IP-address of an MQTT broker that HomeAssistant uses. You can read about how to setup an MQTT broker for HomeAssistant here: https://www.home-assistant.io/docs/mqtt/broker
      So your script will publish data to an MQTT broker. HomeAssistant can then use an MQTT sensor that listens to to the topic of that data (https://www.home-assistant.io/integrations/sensor.mqtt/)

      Like

      Reply

  4. Hi, thank you for your tutorial,

    Some Things, where i got issues:

    You need to install the required Libs over pip / pip3 depend on python version.
    First you need pip 🙂

    ===
    Python 2:
    sudo apt install python-pip

    Python 3:
    sudo apt install python3-pip

    After pip installations you can install the required libs (required)

    pip install paho-mqtt
    pip install psutil

    ===

    For the crontab on raspbian buster you need to change:

    /usr/bin/python /home/pi/scriptname.py mydevicename

    if you wish to log, you can also add ” >> and log file”
    /usr/bin/python /home/pi/scriptname.py mydevicename >> /home/pi/myscript.log

    Not set up the crontab to sudo user the script, you will get some permissions issued by executing.

    thank you stoffl

    Like

    Reply

    1. The first argument to the script is used as a computer_name variable. You can hard code a name when calling the script like “systemmonitory.py mycomputername” or, if on a linux-like OS, use the $HOSTNAME environment variable like “systemmonitor.py $HOSTNAME”

      Like

      Reply

Leave a comment