🛠 Logger Python température + qualité Wi-Fi avec gestion de pannes réseau

Objectives

Build a supervision service tailored to the TWO DOCUMENTED FLAWS of the Zero 2 W: throttling at 85 °C and unstable radio (errors like “brcmfmac ... failed: -5”, disconnections following the Bookworm upgrade — raspberrypi/firmware#1962). The program must therefore survive network issues: short timeouts, error capture, a loop that never dies, and local logging.

Hardware

• 1 Raspberry Pi Zero 2 W on the network (2.4 GHz, WPA2)
• 1 local “server” machine listening for HTTP POST requests (a “python3 -m http.server” is sufficient for testing, or Flask)
• Packages: “sudo apt install python3-requests iw” (iw provides signal strength measurement)

Complete Code

#!/usr/bin/env python3
"""Température SoC + RSSI Wi-Fi vers un tableau de bord local.
Pensé pour la Zero 2 W : throttling à 85 °C et radio 2,4 GHz capricieux."""
import subprocess, time
import requests

WLAN = "wlan0"
URL = "http://192.168.1.10:8080/ingest"

def soc_temp_c:
    # vcgencmd parle au GPU : lecture fiable (les zones sysfs traînent).
    out = subprocess.check_output(["vcgencmd", "measure_temp"], text=True)
    return float(out.split("=")[1].split("'")[0])   # "temp=48.3'C"

def wifi_rssi:
    out = subprocess.run(["iw", "dev", WLAN, "station", "dump"],
                         capture_output=True, text=True).stdout
    for line in out.splitlines:
        if "signal" in line:
            return int(line.split[1])   # dBm : > -70 bien, < -85 marginal
    return None

while True:
    payload = {"host": "zero2w", "temp_c": soc_temp_c, "rssi_dbm": wifi_rssi}
    try:
        requests.post(URL, json=payload, timeout=5)
    except requests.RequestException as e:
        # Les micro-coupures brcmfmac arrivent : on journalise, on continue.
        print("upload rate:", e, flush=True)
    time.sleep(60)

systemd service (automatic restart if the process truly dies) with “Restart=on-failure” and “RestartSec=10”.

Expected Verification

The server receives one POST per minute with temperature and RSSI. Robustness test: cut the Wi-Fi at the router level (“sudo ip link set wlan0 down”) — the script displays “upload rate” and resumes as soon as the network returns, without restarting. Under CPU load, observe the temperature↔frequency correlation in the data.

Common Errors

• No timeout on requests: the POST hangs during a brcmfmac disconnection and the service stops responding — timeout=5 is non-negotiable.
• Letting exceptions propagate out of the loop: a single uncaught failure kills the service; the try/except must wrap the upload, not the entire program.
• Parsing temperature via /sys/class/thermal: possible, but vcgencmd is the reference measurement on this SoC.
• Neglecting the 512 MB of RAM: a logger that queues everything in memory will eventually hit OOM; bound the queue (the last 20 data points are sufficient).
• Pointing the URL to an external address: in case of DNS failure, the DNS timeout may exceed the HTTP timeout — prefer a local IP.