Example code:

lora.join(activation=LoRa.OTAA, auth=(app_eui, app_key),timeout=0)
# wait for a connection to be established
while not lora.has_joined():
    pass

Which will hang forever if the first join fails (which happens to us often enough during educational workshops with LoPy4).

The docs say Internally the stack will automatically retry every 15 seconds until a Join Accept message is received. - which implies no wizardry is needed for the 15s intervals to join.

If I am overlooking something. any tips are welcomed!

sysname='LoPy4', nodename='LoPy4', release='1.18.1.r1', version='v1.8.6-849-b0520f1 on 2018-08-29', machine='LoPy4 with ESP32', lorawan='1.0.2', sigfox='1.0.1'

The only semi-reliable method we have for educational workshops (where 10+ devices are being powered on/off at the same time) is this:

print("Device EUI: " + binascii.hexlify(lora.mac()).upper().decode('utf-8'))

while not lora.has_joined():
    # if no connection in a few seconds, then reboot
    if utime.time() > 15:
        print("possible timeout")
        machine.reset()
    pass

Although it looks and feels like a dity workaround for something that is breaking somewhere at the OS-level.

Anyone with 5+ LoPy4's to test could replicate the issue easily and reliably. This failure state triggers all-the-time, it doesn't seem to be dependant on the Gateway model.

If this happened to a device in the field, without our workaround, it would remain in a hung state indefinitely. That, erm, concerns me a little.

    lora.add_channel(int(upstream.get('chan')), frequency=int(upstream.get('fq')), dr_min=0, dr_max=int(data_rate))  
    for index in range(0, 71):
        if index != upstream.get('chan'):
            lora.remove_channel(index)


        # create an OTA authentication params
        dev_eui = binascii.unhexlify(self.config["lora"]["otaa"]['dev_eui'])
        app_key = binascii.unhexlify(self.config["lora"]["otaa"]['app_key'])
        nwk_key = binascii.unhexlify(self.config["lora"]["otaa"]['nwk_key'])

        start_join_time = time.time()

        # lora.nvram_restore()

        if not lora.has_joined():
            if self.config["lora"]["region"] == "US915":
                lora.join(activation=LoRa.OTAA, auth=(dev_eui, app_key, nwk_key), timeout=0, dr=0)
            else:
                lora.join(activation=LoRa.OTAA, auth=(dev_eui, app_key, nwk_key), timeout=0, dr=int(self.config["lora"]["data_rate"]))


        # wait until the module has joined the network
        while not lora.has_joined():
            time.sleep(1)
            pycom.rgbled(0x000000)
            print('.', end='')
            time.sleep(1)
            pycom.rgbled(0x0000ff)

        lora.nvram_save()



    for i in range(0, 8):
        fq = 902300000 + (i * 200000)
        self.lora.add_channel(i, frequency=fq, dr_min=0, dr_max=self.config["lora"]["data_rate"])

@emmanuel-florent I've omitted the region settings to just show the minimal code. We're heavy-users of LoPy hardware; everything upstream of the join event is fine.

@robert-hh I agree there is a collision, though it's not attempting every 15 seconds. A failed connection never recovers, even if other failed devices are turned off (leaving 1 failed device "on" to attempt its own recovery).

It does appear more-and-more that there's an issue in the stack for this scenario. With REPL connected, sending a CTRL+C during a failure condition

Left as is, this bug could be viewed as a DDoS layer for Pycom LoPy4. Devices that attempt to connect on power-up could find themselves in this situation and never "connect" using any of the recommended examples in Pycom Docs. I doubt many people have experienced the issue as they are not holding educational workshops that invoke more-than-normal LoRaWAN events over a short time.

About Ctrl-C: The Lora stack/device is not fully reset on Ctrl-C. You have to push hard reset (button) or call machine.reset().

Gateway logs show 1 join request from the failed device(s), I watched the logs for two minutes each time awaiting any signs of life (there were none).

from network import LoRa
import pycom
import binascii
import socket

lora = LoRa(mode=LoRa.LORAWAN, region=LoRa.AS923)
app_eui = binascii.unhexlify('XXXXXXXXXXXXXX')
app_key = binascii.unhexlify('XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX')

lora.join(activation=LoRa.OTAA, auth=(app_eui, app_key), timeout=0)

print('Waiting for LoRaWAN network connection...')
while not lora.has_joined():
    pass

print('Network joined!')

while True:
    pass

while not lora.has_joined():
pass

I would add something like sleep(1) into the loop, giving the background tasks more time.