How do you keep an ESP32 AP running 24/7?

[davefes]

I am trying to running a simple server on an ESP32 AP that listens for a few clients. I think I have ways of telling if the server part is running, however the system locks-up every day or so.

Is there a way for the ESP32 itself to tell if the AP is running and then if not ... do a machine.reset()??

Would running asyncio with a server task and another task blinking an LED with a watchdog timer be able to tell me if the server parts locks-up?

This what I am running as the server:

while True:
    try:
     #  allow for immediate re-use
        server = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
        server.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
        server.bind((config.host, config.port))
    except:
        socket_count += 1

        try:
            with open('errors.txt', 'a') as outfile:
                outfile.write('something failed at the socket stage' + '\n')
                outfile.write('socket fail count' + str(socket_count) + '\n')
        except OSError:
            pass

        machine.reset()

    while True:
        try:
            server.listen()
            client, address = server.accept()
            break
        except:
            listen_count += 1

            try:
                with open('errors.txt', 'a') as outfile:
                    outfile.write('something failed at the listen stage' + '\n')
                    outfile.write('listen fail count' + str(listen_count) + '\n')
            except OSError:
                pass

  # try starting again if listen fails
    break

[glenn20]

One approach is to have a client which is expected to send heartbeat signals to your server. In this way, it doesnt matter what state your software/hardware gets into, but it does rely on at least one active heartbeat device on your network.

You could use the heartbeat to feed a watchdog or have your watchdog task check the time since the last heartbeat and stop feeding the watchdog if the heartbeat has timed out.

[davefes]

Thanks @glenn20. I recalled your ping examples in ESPNow but could not work-out how that was going re-boot the server. Now I understand.

I would prefer that the clients (solar-powered remotes) were not pinging all the time, so maybe another local client that can run off mains could be a solution. I will review your ping example.

[heiqiaokeli1947]

use microdot as webserver.
https://microdot.readthedocs.io/en/latest/index.html

[glenn20]

Hmmm. In my own tests, I have not been able to replicate the issues you were seeing (using mpy v1.22.1). I was speculating that possibly the wait_closed() was being called before the task was fully initialised and may have been returning immediately. And eventually the server object would be garbage collected causing the server task to terminate (or some other unexpected behaviour).

If you were to delete/comment the print('now do server.wait_closed'), then you should see the issues arise again.

[davefes]

An hour ago I verified that removing the print statement caused the error.

Doing it again the server now runs continuously ?!?

[davefes]

MicroPython v1.22.2 on 2024-02-22; Generic ESP32 module with SPIRAM with ESP32

I could try a Generic without SPIRAM.

[davefes]

MicroPython v1.22.2 on 2024-02-22; Generic ESP32 module with ESP32

And it just keeps running! The reason I wanted to use the SPIRAM variant is that when pictures and datalogs don't find their way to Gmail I save them locally. Wanted more room.

I'll do some stress testing with:

async def main():
    print('starting as an AP...')
    connect_ap()

  # start the server
    print('Setting up server')
    server = await asyncio.start_server(handle_client, '192.168.4.1', 10000)
    asyncio.create_task(blink_led())
    await server.wait_closed()

asyncio.run(main())

[davefes]

Maybe connect_ap() should only be done once before running main.