Pi Pico shutdowns and won't restart despite use of watchdog

[ivancornut]

Hello,

I have a Pi Pico (RP2040) that is running continually as a datalogging device. It writes the data to an SD card and uses ligthsleep between two timesteps. The code runs on many devices over months with no issues. I use the watchdog to reset the device in case it hangs.

Unfortunately I setup a new device in the field and it stops working after two days. Instead of the watchdog resetting the device as it should I have to turn it off and on again to get it working. Then it works fine, but only for two days again. If the time after which it sopped working were variable I would suspect hardware but I am suspecting something linked to micropython.

I am no specialist but maybe the RAM is saturated (would a gc.collect() be useful in that case ?) ? Is there anything I can do other than programming a machine.reset() every day ? I am running micropython release 26.0

Thank you for your answers

[davefes]

I run two ESP32 dataloggers. Both do a machine.reset() every day. One does a gc.collect() every hour when I save RAM data to flash. The other does a gc.collect() every processing loop (once a second). AI suggests that doing a gc.collect() every loop is problematical. I have a software watchdog running and still get the odd watchdog timeout, but I have managed to keep things running for long periods.

[robert-hh]

I setup a new device in the field and it stops working after two days. Instead of the watchdog resetting the device as it should I have to turn it off and on again to get it working.

If both boards have the same firmware & software set-up, then a difference in behavior should only depend on the hardware itself or the data the process.

AI suggests that doing a gc.collect() every loop is problematical.

Per se that should not be problematic. Only gc.collect() may take some time, which could matter if fast response is required.

[ivancornut]

Ok yes honestly this is really a mystery for me. Maybe comes from some of the dependencies. I guess maybe using gc.collect() once in a while and reset once a day will do the trick. Thank you both for your answers