Random ConnectionError('Error UNKNOWN while writing to socket. Connection lost.')

[andrijan]

Hey there!

I've been using channels in production for a while and everything works smoothly. Except when looking at the logs (or Sentry) I see some random ConnectionError in my consumers.

import json

from channels.generic.websocket import AsyncJsonWebsocketConsumer


class Consumer(AsyncJsonWebsocketConsumer):
    async def connect(self):
        self.id = self.scope["url_route"]["kwargs"]["id"]
        self.group_name = f"group_{self.id}"
        if not self.channel_layer:
            return

        await self.channel_layer.group_add(self.group_name, self.channel_name)  # Here the error is triggered
        await self.accept()

    async def disconnect(self, close_code):
        if not self.channel_layer:
            return
        await self.channel_layer.group_discard(self.group_name, self.channel_name)

    async def update(self, event):
        await self.send(text_data=json.dumps(event))

Screenshot from sentry where the error is triggered in the code above

Screenshot from sentry below the other error

So my question is basically, are these ConnectionErrors normal, due to network connectivity issues from the users?

And if so, should I just ignore them (With a try/except block)?
If not, is there anything I should be doing differently in the consumer or maybe some configuration issue?

I'm using daphne and it is running on Heroku

[carltongibson]

Hi @andrijan.

Yes... these things can happen. Often there's a timeout somewhere, or something similar, that causes the connection to be closed (or, yes too, there can just be an error.)

Folks have had success sending a heartbeat every so often, to make sure a connection isn't shut for being idle.

All this depends on your provider… — I don't know if Heroku have any guidance on long-lived connections.

It's not really something I can explicitly help you with though, I'm afraid.

HTH