Health and webhook

Health and webhook

The AP keeps an eye on its labels and can tell you when something needs attention.

What counts as a problem

Flag	Means
silent	No poll for 3 × the label's interval
unseen	Known from before, but not heard from since the AP started
low_battery	Below 2500 mV (a label that reports 0 cannot measure it)
draw_failed	The AP could not draw its screen
content_error	It has content it cannot show, such as old firmware that does not take images

Lost radio

On an ESP32-S3 AP the labels' radio is an ESP32-C6, on a cable or over WiFi. When the S3 has had no contact with a C6 for 30 seconds (over WiFi up to 2 minutes, when the C6 just loses power), the report lists it under radio_problems as radio_lost, with how long it has been gone, and ok is false. Until it answers again the labels it serves hear nothing from the AP. A webhook gets the report both when the radio is lost and when it is back. The dashboard shows it in red at the top, and Home Assistant as a repair. Firmware from before this sends no radio_problems.

"radio_problems": [{"radio": 0, "id": "000098880E7177C8", "via": "uart", "problems": ["radio_lost"], "lost_s": 300}],
"radio_lost": 1

Asking for it

python3 esl_client.py health                    # exits 1 when something needs attention
curl -H "Authorization: Bearer $TOKEN" "http://openesl.local/api/health?silent_s=600&battery_mv=2400"

The thresholds can be changed per call. The dashboard shows the same report.

Webhook

With a webhook set, the AP posts the same report to your URL whenever it changes, including when a radio is lost or back (checked every 30 seconds):

python3 esl_client.py config --webhook http://homeserver:8123/api/webhook/esl
python3 esl_client.py config --webhook ""       # turn it off

Plain HTTP, no retries: a failed post is logged and the change goes out with the next one. A Home Assistant webhook works well as the receiver (Home Assistant).

Long-term test

esl_client.py log week.csv writes a row per label per minute until you stop it, and esl_client.py report week.csv sums it up: outages, AP restarts, the longest time each label went without a poll, and the lowest battery.