Temperature reader for Raspberry Pi

Temperature reader periodically reads the RPi CPU temperature and publish it as the Prometheus readable metric on http interface.

Prerequisities

• It is expected that there is Raspberry Pi OS Lite 64 bit is running on the target machine
• It is possible to burn it onto SD card via rpi-imager
• There must be SSH access enabled on the target

Find the target IP

If the target RPi is connected to the same network as your host, you can find the RPi's IP with the following command.

nmap -p 22 --open -oG - $(hostname -I | cut -f1 -d ' ')/24

• -p 22 Only scans specified port 22 (SSH)
• --open Only shows open ports
• -oG Output scan in grepable format
• hostname -I Shows all addreses for the host
• cut -f1 -d ' ' Cuts the first IP address

Add key based authentication

Generate new key pair for RPi target.

ssh-keygen -t rsa -f ~/.ssh/rpi_rsa

Add your public key to the allowed keys on host.

ssh-copy-id -i ~/.ssh/rpi_rsa.pub pi@PI_IP_ADDRESS

Add key to ssh agent

ssh-add ~/.ssh/rpi_rsa

Now you can log to the RPi without password.

Cross compilation

As the Raspberry Pi runs on the different environment than your host machine, it is needed to cross compile the binary. One way to do it is to use docker image with the correct environmanet installed and build the binary inside the container. First you need to install docker and then, inside the project folder tun the following command.

docker build . -t rpi-cross-compile-image

It builds the docker image from the Dockerfile supplied in the project root. Then there is script cross.sh, which runs the docker image and executes commands supplied to it. So you can build the binary inside container.

./cross.sh cargo build

And then run the binary on the remote target.

./cross.sh cargo run pi@PI_IP_ADDRESS

Tere is another script run.sh used as a custom cargo runner. It copies and run the binary on the raspberry over SSH.

Test

To test the temperature measuring you can run stress test on the target machine. First you need to install stress package.

sudo apt install stress

And the run the command. It spawns 4 processes for 600 seconds. It calculate square root of some garbage, so it loads CPU to 100%.

stress -c 4 -t 600

You can see the temperature rising on the web page.

http://PI_IP_ADDRESS:8081/metrics

You can also run prometheus and grafana on your host machine and create a beautiful temperature chart :)