Machine status tracking

I've got a little 5.5cm 3-colour epaper display. And I've got a Raspberry Pi that's on all the time, since it's a Pi-Hole. And there are a couple of other usually-on machines hanging around, too. So, I figured I should put the display to use and show some status for these machines.

But then the question came up: How do I get the other machines to tell the Pi-Hole about themselves once in a while? I could write a shell script that runs some commands over SSH on each machine, but an API sounded like more fun, so here we are.

Installation

To get this running for yourself, you'll need a machine to run the server, and one or more machines to run the client.

To install the server (I'm assuming Debian/Ubuntu here, but I imagine this will work on most distros that use systemd):

1. On the server machine, clone or copy this repo.
2. Compile the binary with go build.
3. As root, move the machine-status executable to /usr/local/bin/, renaming as required.
4. As root, copy the system service file to /etc/systemd/system/.
5. Run a few commands to get things started:

sudo systemctl daemon-reload
sudo systemctl enable machine-status.service
sudo systemctl start machine-status.service

1. Verify that the service is running:

sudo systemctl status machine-status.service

To install the client (I'm assuming you have a recent python3 installed):

1. On the client machine, copy machine-status-update.py to /usr/local/bin and make it executable:

sudo cp client/machine-status-update.py /usr/local/bin/machine-status-update
sudo chmod +x /usr/local/bin/machine-status-update

1. Make a directory to hold config.toml, and copy the file into it:

sudo mkdir /etc/machine-status-update
sudo cp config-files/config.toml /etc/machine-status-update

1. Adjust the settings in config.toml as required. At a minimum, you'll want to change machine_status_ip to the IP address or hostname of the machine running the server.

2. Copy the machine-status-cron file to the system's cron-job directory:

sudo cp config-files/machine-status-cron /etc/cron.d/machine-status-cron

1. Run the client, and fix any Python dependency issues that come up...

/usr/local/bin/machine-status-update

Finally, wait a bit (is your cron job running at the default 5-minute interval, or did you make it longer?), then query the server to see if everything is working...

curl http://servername:8080/machine-status/machines

(You should get a JSON object back, showing the status of any configured client machines.)

Go server

To generate the sample Go server from the API spec:

java -jar ../swagger-codegen/swagger-codegen-cli.jar generate -i api-specs/machine-status.yaml -l go-server -o go-server

And every time you regenerate it, you need to fix the import statement in main.go to update the sw line to read as follows:

  sw "polarweasel/machine-status/go"

(This is a reference to the module entry on line 1 of go.mod. If you use this code, change both to suit your situation.)

Generate Go models only

Or... just generate the models, and leave the implementation code alone. (This is using the selective generation options.)

# Run this from the `server` directory
java -Dmodels -DmodelDocs=false -DmodelTests=false \
     -jar ../swagger-codegen/swagger-codegen-cli.jar \
     generate -i api-specs/machine-status.yaml \
     -l go-server \
     -o go-server

Python client

The client and configuration file are in /client. This should be ready to go once the server exists. It does send valid output according to the mocking server, at least!

Mocking server

This is why Redocly is included in this repo...

Note that the API is slightly out of sync with the implementation, because I wimped out on implementing the problem+json error format.

# Pick up the current changes to the API spec
npm run bundle-internal

# Run the mocking server
npm run mock-internal