why on_close only was called once, so it causes the client to reconnect automatically only once, even I set reconnect parameter.

[GooVincent]

Hi, everyone, I am trying to implement reconnecting ws server when close by the ws server. What I am using websocket-client version is 1.7.0. And I found on_close was just triggerd once when the ws server closed the socket.

Following is my demo code.

client.demo

def on_close():
       print('on_close is triggered')
        ws.run_forever(dispatcher=rel, reconnect=5)
        rel.signal(2, rel.abort)  # Keyboard Interrupt
        rel.dispatch()

def main()
      ws = websocket.WebSocketApp(
                    f"ws://{host}/ws",
                    on_open=on_open,
                    on_message=on_message,
                    on_close=on_close,
                    on_error=on_error,
                    ping_timeout=3,
                    ping_interval=5)
      
       ws.run_forever(dispatcher=rel, reconnect=5)
       rel.signal(2, rel.abort)  # Keyboard Interrupt
       rel.dispatch()

server.demo

async def test(ws):
    print(f'running test..')
    while True:
        current_time = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
        await asyncio.sleep(3)
        print(f"{current_time}")
        break


async def main():
    ip = "localhost"
    port = 8765
    async with websockets.serve(test, ip, port):
        print(f'ws server running on {ip}:{port}')
        await asyncio.Future()  

What I wanna is the client will keep reconnecting to the server, as I call run_forever() again in on_close, but it only happen once, why?

Please anyone could help me on this?

[bubbleboy14]

Hey @GooVincent, a few notes:

1. only call rel.dispatch() once (in your example, main()) looks like the correct place)
2. no need to call rel.signal() - websocket-client takes care of that here:

   websocket-client/websocket/_app.py

   Line 142 in a394b46

   dispatcher.signal(2, dispatcher.abort) # keyboard interrupt

3. ping_timeout and ping_interval kwargs should be passed to run_forever(), not the WebSocketApp constructor

If you provide complete (runnable via "python client.py") client and server examples, I can take a crack at your issue.

[GooVincent]

Here is my complete client.py

import websocket
import rel
from functools import partial


class WsClient(object):
    def __init__(self) -> None:
        # websocket.enableTrace(True)
        host = "localhost:8765"

        self.ws = websocket.WebSocketApp(
            f"ws://{host}/ws",
            on_open=partial(self.on_open),
            on_message=partial(self.on_message),
            on_error=partial(self.on_error),
            on_close=partial(self.on_close),
            on_ping=partial(self.on_ping),
            on_pong=partial(self.on_pong))

    def on_message(self, ws, message):
        print(f"received message: {message}")

    def on_error(self, ws, error):
        print(f"{ws} error: {error}")

    def on_close(self, ws, close_status_code, close_msg):
        print(f"### {ws} closed with code={close_status_code}, msg={close_msg}###")

        # Always to reconnet to server
        self.set_run_option()

    def on_open(self, ws):
        print(f"{ws} connected to server")

    def on_ping(self, ws, data):
        """
            handle msg when receive ping from the other side
        """
        print(f'<<<<<<<<received a pinging, data:[{data}]')

    def on_pong(self, ws, data):
        """
            handle msg when receive pong from the other side
        """
        print(f'<<<<<<<<received a pongong, data:[{data}')

    def set_run_option(self):
        self.ws.run_forever(dispatcher=rel,
                            reconnect=5)

                            # websocket-client will exit when ping timeout
                            # ping_interval=int(main_conf['Client']['ping_interval']),
                            # ping_timeout=int(main_conf['Client']['ping_timeout']),
                            # ping_payload=self.protocol_handle_util.ping_payload()
    def run(self):
        self.set_run_option()
        rel.dispatch()


def main():
    ws_client = WsClient()
    ws_client.run()


if __name__ == '__main__':
    main()

and the client will print like following. then stucked, watch the link status, there are two CLOSE_WAIT links.

<websocket._app.WebSocketApp object at 0x7f967c2cffd0> connected to server
### <websocket._app.WebSocketApp object at 0x7f967c2cffd0> closed with code=1000, msg=###
<websocket._app.WebSocketApp object at 0x7f967c2cffd0> connected to server

[GooVincent]

Here is my complete client.py

import websocket
import rel
from functools import partial


class WsClient(object):
    def __init__(self) -> None:
        # websocket.enableTrace(True)
        host = "localhost:8765"

        self.ws = websocket.WebSocketApp(
            f"ws://{host}/ws",
            on_open=partial(self.on_open),
            on_message=partial(self.on_message),
            on_error=partial(self.on_error),
            on_close=partial(self.on_close),
            on_ping=partial(self.on_ping),
            on_pong=partial(self.on_pong))

    def on_message(self, ws, message):
        print(f"received message: {message}")

    def on_error(self, ws, error):
        print(f"{ws} error: {error}")

    def on_close(self, ws, close_status_code, close_msg):
        print(f"### {ws} closed with code={close_status_code}, msg={close_msg}###")

        # Always to reconnet to server
        self.set_run_option()

    def on_open(self, ws):
        print(f"{ws} connected to server")

    def on_ping(self, ws, data):
        """
            handle msg when receive ping from the other side
        """
        print(f'<<<<<<<<received a pinging, data:[{data}]')

    def on_pong(self, ws, data):
        """
            handle msg when receive pong from the other side
        """
        print(f'<<<<<<<<received a pongong, data:[{data}')

    def set_run_option(self):
        self.ws.run_forever(dispatcher=rel,
                            reconnect=5)

                            # websocket-client will exit when ping timeout
                            # ping_interval=int(main_conf['Client']['ping_interval']),
                            # ping_timeout=int(main_conf['Client']['ping_timeout']),
                            # ping_payload=self.protocol_handle_util.ping_payload()
    def run(self):
        self.set_run_option()
        rel.dispatch()


def main():
    ws_client = WsClient()
    ws_client.run()


if __name__ == '__main__':
    main()

and the client will print like following. then stucked, watch the link status, there are two CLOSE_WAIT links.

<websocket._app.WebSocketApp object at 0x7f967c2cffd0> connected to server
### <websocket._app.WebSocketApp object at 0x7f967c2cffd0> closed with code=1000, msg=###
<websocket._app.WebSocketApp object at 0x7f967c2cffd0> connected to server

I found to modify the member has_done_teardown, it will work. like:

def on_close(self, ws, close_status_code, close_msg):
      ...
      self.ws.has_done_teardown = False
     ...

But I am not sure if it is reasonable.

[bubbleboy14]

I tested your client code against my usual websocket server, @GooVincent, and didn't notice any problem with reconnects - every time I turn the server off and on, the client reconnects without issue.

In any case, it sounds like you found a workaround - glad to hear you have a solution, happy coding!

[GooVincent]

I tested your client code against my usual websocket server, @GooVincent, and didn't notice any problem with reconnects - every time I turn the server off and on, the client reconnects without issue.

In any case, it sounds like you found a workaround - glad to hear you have a solution, happy coding!

Perhaps to have a try with my server.py if possible.

Yes, setting self.ws.has_done_teardown = False will work. Anyway thanks for all your advice above.

[bubbleboy14]

@GooVincent I tried your server example, and finally see what you mean - this is about reuse of the WebSocketApp instance after the server closes the connection. Hah, I should have read your question more carefully, sorry.

Great work determining that this is as simple as flipping that bool - I added your change to this pull request (which addresses an ambiguity related to reconnects): #972.

@engn33r I stuck has_done_teardown = False near the top of run_forever(). That pull request also includes a fix related to #971 - LMK what you think.

[engn33r]

This should be fixed in v1.8.0 which includes PR #972. If this reoccurs, we can reopen the issue.

[bubbleboy14]

I think I encountered this as well! This was happening because WrappedDispatcher (via reconnect() and timeout()) wasn't passing reconnecting=True to setSock(). I think this was actually probably my oversight, oops, sorry.

Anyway, #983 addresses this issue, and also some related stuff.

Anyone having reconnect (or async write) issues, please LMK if this branch improves things on your end!