Recovering from "Websocket error: Connection to remote host was lost"

[ryanjflynn]

I'm trying to use this package for coding a bot app, but I'm consistently finding that after some time (varies, usually less than a couple hours) I'm seeing the logging produce ERROR:groupme-push:Websocket error: Connection to remote host was lost. ERROR:websocket:Connection to remote host was lost. - goodbye

and then the script terminates. I'm not sure if there is a way to pin down what is causing the connection to be lost, but more importantly I'm wondering what can be down to recover form this and restart the connection. In my script calling client.start() it seems like are not any exceptions being thrown/caught when it happens, so I'm not sure how to detect that the PushClient has stopped running once it gets the connection error.

[SZRabinowitz]

Hello. In my fork biden2020prez/groupme-push, it reconnects whenever that error happens. Install it using pip install git+https://github.com/biden2020prez/groupme-push --force.

Also, there is a good chance that those fixes will be added to this repository in the future.

[SZRabinowitz]

to give more details, you should add an argument called reconnect when initializing the pushclient, with an integer of how many seconds it should try to reconnect after an error.

for example, this is how I do it in my userbot

client = PushClient(access_token=api_key, on_message=handle_messages.on_message, disregard_self=ignore_self, reconnect=20)

In that example, if it has an error, it will reconnect after 20 seconds. It would probably also work if you used 1 second, I don't know why I didn't.

[cuuush]

Will work on a fix for this, thanks Biden for suggesting your fork :)

[SZRabinowitz]

Hi @cuuush. I simply pass the reconnect option that the user specifies over to websocket-client, which has a reconnect option. Should be simple :)

[SZRabinowitz]

Once you are at it...

I added two lines here which fixed errors when internet gets disconnected:

           else:
                print(
                    "Groupme got unhandled message: {}".format(
                        json.dumps(message, indent=4)
                    )
                )
                self.ws.close()
                self.start()

It was a patch, more than a fix. Basically groupme was sending an error about needing to reconnect. Unfortunately, I dont have the log anymore to tell you exactly what the error was

[cuuush]

Ha, that’s smart. I think GroupMe’s faye web socket sever has a max alive time, in my experience connections over 24h get killed. So we can just reconnect if it fails. Maybe we should also only try reconnecting a couple times in case there’s some other issue. Like the server is down or broken, so we don’t flood it.

[SZRabinowitz]

Hmm... I guess that's an option. Or if it fails to reconnect more than once, it can wait one hour before trying again

[SZRabinowitz]

I recall having another issue that the server closes the connection without an error, so reconnect and on_error doesn't work. But I have a bot running for a week now without issues, so I don't remember much about it.

[SZRabinowitz]

I am now looking at the changes I made in my fork.

I did that so users can catch errors starting the bot

[ryanjflynn]

Thank you guys for taking a look and thanks @biden2020prez for the info and example from your fork!