MatrixPortal hangs and disconnects from USB after reading URL multiple times

[jldeon]

CircuitPython version

Adafruit CircuitPython 7.2.0 on 2022-02-24; Adafruit Matrix Portal M4 with samd51j19
Board ID:matrixportal_m4

Code/REPL

import time
import board
import traceback

from adafruit_matrixportal.matrixportal import MatrixPortal
from secrets import secrets

matrix_portal = MatrixPortal(status_neopixel=board.NEOPIXEL, debug=True)
network = matrix_portal.network
display = matrix_portal.graphics.display

url = "http://example.com"

while True:
    try:
        print(f"Reading url: {url}")
        matrix_portal.set_background(0x400000)
        resp = network.fetch(url)
        matrix_portal.set_background(0x000000)
        print(resp.content.decode('utf-8')[:64])
    except (ValueError, RuntimeError) as e:
        print("Likely data read error:")
        traceback.print_exception(None, e, e.__traceback__)
    except OSError as e:
        print("Likely wifi connection error:")
        traceback.print_exception(None, e, e.__traceback__)
    except Exception as e:
        print("Unknown error:")
        traceback.print_exception(None, e, e.__traceback__)
    time.sleep(1)

Behavior

After some time passes (random, but typically less than an hour), the device stops responding.

The last thing printed is "Retrieving data..." (which is a print that can't be turned off in the CircuitPython libraries). The LED matrix is stuck on red (so the call to fetch hasn't returned.)

After some time passes, USB disconnects. The board has to be hard-reset to recover.

There is no Python backtrace or any other error message printed via the serial port.

Description

No response

Additional information

I tested this initially with a text file hosted on an Apache server on the local LAN. I then tested with example.com just to rule out anything weird on my Apache instance. (So you ought to be able to repro this with various URLs, basically)

You can probably omit most of the exception handling from my example code. I've found that sometimes the MatrixPortal doesn't want to connect to wifi or the response gets screwy on occasion, so I handle those. However, I haven't hit those exceptions when performing this test, so it's likely not needed.

[jldeon]

Slightly more minimal example:

import time
import board

from adafruit_matrixportal.matrixportal import MatrixPortal

matrix_portal = MatrixPortal(status_neopixel=board.NEOPIXEL, debug=True)
network = matrix_portal.network

url = "http://example.com"

while True:
    print(f"Fetch start")
    resp = network.fetch(url)
    print(f"Fetch end")
    time.sleep(1)

[ajsirota]

Can confirm this.

[kroezem]

Same issue. This code will not run properly without access to my Spotify api.
MAR30.zip

[mpicker90]

same issue. I tried resetting the Network connection every couple minutes as well, but the issue still persists. Running on the latest 7.2.4 uf2

[jepler]

It's not a fix, but an option you have is to try the hardware watchdog, which would reset the whole microcontroller (the one where CircuitPython resides) if your program stops running normally.

Before doing this, be comfortable with entering safe mode on your board, because an errant watchdog reset can make it difficult to modify your code.py.

You'd initialize the watchdog with a timeout:

from microcontroller import watchdog as w
from watchdog import WatchDogMode
w.timeout=2.5 # timeout in seconds
w.mode = WatchDogMode.RESET
w.feed()

and within your top-level loop you'd keep the watchdog happy with w.feed(). If it's ever NOT called for 2.5 seconds, the whole microcontroller is reset, similar to pressing the reset button.

[Wintaru]

I battled this as well and worked with staff to figure out the issue. We tried replacing hardware too which helped failure rates but it still fails. Trying the watchdog thing now and it seems to be stable so far, I just started a rather long print so we'll see if it's still running in the morning.

[dhalbert]

For those of you who had this problem, what version of the NINA-FW firmware are you using on the ESP32 that's on the MtarixPortal? That is printed out by the simple Internet test given here: https://learn.adafruit.com/adafruit-matrixportal-m4/internet-connect. The key value is esp.firmware_version.

[Wintaru]

Firmware vers. bytearray(b'1.2.2\x00')

[mpicker90]

Firmware vers. bytearray(b'1.2.2\x00')

[jldeon]

Same here:

Firmware vers. bytearray(b'1.2.2\x00')

[mpicker90]

I tried using the watchdog time out as suggested above. Just had it crash and reset automatically. Thanks for the suggestion @jepler

[Wintaru]

I tried using the watchdog time out as suggested above. Just had it crash and reset automatically. Thanks for the suggestion @jepler

I had to increase the timeout to 15, and make sure you have the feed call both there and in your loop otherwise it's not fed soon enough to enter the loop.

Also, if your loop is longer than the timeout, you'll have to feed the watchdog somehow or shorten your loop sleep.

[anecdata]

It's worth testing whether NINA v1.7.4 improves things, though I couldn't definitively tie any NINA fixes to this problem.

It seems odd to me that this is only reported on MatrixPortal and not other M4+Airlift configs like PyPortal.