Watchdog possible on ESP32 versions?

[hagenbuch]

Is your feature request related to a problem? Please describe.

I didn't find anything in the documentation, so:

I'm using WLED 15.4.0 successfully. on an ESP32 Gleopto hardware with only 5 m of WS281x and a beefy power supply. I even added a beefy electrolytic capacitor right at the ESP32 side to improve stability.

The Wifi is mostly OK but due to the small antenna not 100% connected. The device may run several weeks in a row but then just stops working: The LEDs keep their state but the device is unaccessible. I run a different ESP32 device by Waveshare in the vicinity and that one remains accessible, so I rule Wifi problems out.

It would be great if we had a watchdog: Current ESP32 IDF seems to support that:

https://docs.espressif.com/projects/esp-idf/en/stable/esp32/api-reference/system/wdts.html

Describe the solution you'd like

It would be nice if we had a watchdog reboot the device if is does not seem to run anymore for, say, 5 minutes - or a configurable number of seconds or minutes.

Describe alternatives you've considered

The only other option I see is to reboot it by powercycling which I'd like to avoid.

[DedeHai]

in general a good idea. there already is a watchdog running so if the device really "hangs" it will reboot.
the question that immediately arises is: what should trigger such a reboot

• no wifi for X minutes?
• no user connected for X hours?
• NTP server unreachable for X hours?
• ...?

this could be something for a usermod though with settings and checkmarks on what triggers a reboot. I don't see this as a core feature, the goal should not be to have workarounds but a stable core and use stable hardware.

[softhack007]

WLED already has a watchdog feature, but it requires a custom build

If memory serves well, we added this around 0.14.0 - but then disabled the task watchdog later, because the watchdog did more harm than good.

WLED/wled00/wled.cpp

Lines 242 to 256 in 8e94cf5

	#if WLED_WATCHDOG_TIMEOUT > 0
	// we finished our mainloop, reset the watchdog timer
	static unsigned long lastWDTFeed = 0;
	if (!strip.isUpdating() || millis() - lastWDTFeed > (WLED_WATCHDOG_TIMEOUT*500)) {
	#ifdef ARDUINO_ARCH_ESP32
	esp_task_wdt_reset();
	#else
	ESP.wdtFeed();
	#endif
	lastWDTFeed = millis();
	}
	#endif
	if (doReboot && (!doInitBusses || !configNeedsWrite)) // if busses have to be inited & saved, wait until next iteration
	reset();

[willmmiles]

A network watchdog usermod is a good idea. It dovetails somewhat with the current usermods/Fix_unreachable_netservices_v2, which intentionally generates periodic network traffic.

[hagenbuch]

Interesting informations, thanks!

I would add that the WLED device would have at least the responsability to keep its main loop running, that is: shoving bits to the outputs, keeping the clock updated and the buffers (http) open. That would suffice for me. Maybe also make sure MQTT connections stay open (if applicable)

Besides, I see the responsability for its reachability more on the router's side, I would be happy if the ESP responded to pings, maybe. If that's too much hassle, responding to http GET requests would be nice, too.

I would not make the update time too short - it is better to see the device struggling (by restarting, say, after a minute of freeze), rather then rebooting undetected every some milliseconds (which might mean a lot of wear, too?)

So in conclusion, I keep my hand up and thatnk you for y'all's contributions and your fantastic work!