Why was there a disconnections when increasing the client from 20 to 30 or more

[arun6894]

I've been testing the library features from the following to understand more.
https://github.com/miguelgrinberg/quick-socketio-tutorial

Asynclient.py

import asyncio
import socketio


class Check:
    def __init__(self):
        self.sio = socketio.AsyncClient(handle_sigint=False, logger=True, engineio_logger=True)
        self.sio.event('connect', self.connect, namespace='/')
        self.sio.on('mult', self.mult, namespace='/')
        self.sio.event('connect_error', self.connect_error, namespace='/')
        self.sio.event('disconnect', self.disconnect, namespace='/')
        self.sio.on('client_count', self.client_count, namespace='/')
        self.sio.event('room_count', self.room_count, namespace='/')
        self.sio.event('user_joined', self.user_joined, namespace='/')
        self.sio.event('user_left', self.user_left, namespace='/')

    async def main_connect(self):
        await self.sio.connect('http://localhost:8000', headers={'X-Username': 'miguel'})


    async def connect(self):
        print('connected')
        result = await self.sio.call('sum', {'numbers': [1, 2]})
        print(result)



    async def connect_error(self,e):
        print(e)


    async def disconnect(self):
        print('disconnected')


    async def mult(self,data):
        return data['numbers'][0] * data['numbers'][1]


    async def client_count(self,count):
        print('There are', count, 'connected clients.')


    async def room_count(self,count):
        print('There are', count, 'clients in my room.')


    async def user_joined(self,username):
        print('User', username, 'has joined.')


    
    async def user_left(self,username):
        print('User', username, 'has left.')

    async def run(self):
        await self.main_connect()


async def main():
    async def loop(i):
        obj1 = Check()
        await obj1.run()
        await obj1.sio.wait()
    x = [loop(i) for i in range(30)]
    await asyncio.gather(*x)



# async def main():
#     await sio.connect('http://localhost:8000',
#                       headers={'X-Username': 'miguel'})
#     await sio.wait()


asyncio.run(main())

Server.py

import random
import socketio
import uvicorn

sio = socketio.AsyncServer(async_mode='asgi')
app = socketio.ASGIApp(sio, static_files={
    '/': './public/'
})
client_count = 0
a_count = 0
b_count = 0


async def task(sid):
    await sio.sleep(5)
    result = await sio.call('mult', {'numbers': [3, 4]}, to=sid)
    print(result)


@sio.event
async def connect(sid, environ):
    global client_count
    global a_count
    global b_count

    username = environ.get('HTTP_X_USERNAME')
    print('username:', username)
    if not username:
        return False

    async with sio.session(sid) as session:
        session['username'] = username
    await sio.emit('user_joined', username)

    client_count += 1
    print(sid, 'connected')
    sio.start_background_task(task, sid)
    await sio.emit('client_count', client_count)
    if random.random() > 0.5:
        sio.enter_room(sid, 'a')
        a_count += 1
        await sio.emit('room_count', a_count, to='a')
    else:
        sio.enter_room(sid, 'b')
        b_count += 1
        await sio.emit('room_count', b_count, to='b')


@sio.event
async def disconnect(sid):
    global client_count
    global a_count
    global b_count
    client_count -= 1
    print(sid, 'disconnected')
    await sio.emit('client_count', client_count)
    if 'a' in sio.rooms(sid):
        a_count -= 1
        await sio.emit('room_count', a_count, to='a')
    else:
        b_count -= 1
        await sio.emit('room_count', b_count, to='b')

    async with sio.session(sid) as session:
        await sio.emit('user_left', session['username'])


@sio.event
async def sum(sid, data):
    result = data['numbers'][0] + data['numbers'][1]
    return {'result': result}


if __name__ == '__main__':
    uvicorn.run(app, host='localhost', port=8000)

What could be the reason behind this connection, disconnect and closed. It would be great if I got some info behind the scenes. I can go from there to modify things. Thank you.

[miguelgrinberg]

Running your code I don't see any disconnections.

INFO:     Started server process [74679]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://localhost:8000 (Press CTRL+C to quit)
INFO:     127.0.0.1:64615 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.5071979 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64616 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.507901 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64617 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.507536 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64618 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.508235 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64619 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.506181 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64620 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.5067918 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64621 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.508907 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64622 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.509312 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64623 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.5085702 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64624 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.510751 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64625 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.5111601 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64626 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.5103319 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64627 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.5114949 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64628 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.512502 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64629 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.509999 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64630 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.512172 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64631 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.511831 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64632 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.5129359 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64633 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.5096488 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64634 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.514334 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64635 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.513658 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64636 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.5133078 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64637 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.514742 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64638 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.516087 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64639 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.5157492 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64640 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.515081 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64641 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.5139909 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64642 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.515418 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64643 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.5164201 HTTP/1.1" 200 OK
INFO:     127.0.0.1:64644 - "GET /socket.io/?transport=polling&EIO=4&t=1658223444.5168152 HTTP/1.1" 200 OK
username: miguel
username: miguel
iVcxW7wsugjY93cKAAAe connected
username: miguel
username: miguel
username: miguel
username: miguel
username: miguel
username: miguel
username: miguel
ltFtHWRPty-cEBzZAAAf connected
username: miguel
LK6u_SdSEMKygnKOAAAg connected
YPo10iG1VwBGC9riAAAh connected
0eznoNVgUIUM9-MIAAAi connected
username: miguel
8n9J5dHp310hMSFHAAAj connected
9tXy7OTquSqA5ajGAAAk connected
Sd0hTuIjT_mP3FoQAAAl connected
yjCeBjHAil_qoTCzAAAm connected
username: miguel
username: miguel
FJzPhYoIFeeTccvGAAAn connected
username: miguel
username: miguel
username: miguel
username: miguel
username: miguel
username: miguel
username: miguel
username: miguel
username: miguel
username: miguel
username: miguel
username: miguel
vtrVP0byPTXcgqBkAAAo connected
username: miguel
username: miguel
username: miguel
username: miguel
username: miguel
CMRCOKIwKqIaqppzAAAp connected
YkzpAekyNJAjXFwzAAAq connected
i68naCEqKRm3WYbYAAAr connected
EXvM-KAYIgTIJPDMAAAs connected
2ECgnC3CxJX6tysfAAAt connected
yLNmkmWHiXgK-KbPAAAu connected
_wDOfIPGoL7oO8x8AAAv connected
7wGiM4aCh2C95cgCAAAw connected
Tqanh_6S3QjPC5Y-AAAx connected
kyPG3JkHmGdkk200AAAy connected
hY79oclb_7JSS5omAAAz connected
t-UdOkV6_qI2ccBhAAA0 connected
iEs_cOXUYxuj_3fWAAA1 connected
ltwmPb37Y9SC-EOGAAA2 connected
OUfYQebeMEhWi68lAAA3 connected
zV-qnkJAKwu0UJFCAAA4 connected
hLh82bINk1HyJvfoAAA5 connected
lGMskFMt5fFeEAanAAA6 connected
e6VCD3eLKXhlryM4AAA7 connected
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12

[miguelgrinberg]

The problem is that you are connecting your 40-50 clients all at the same time, and each connection triggers 2 broadcasts to all connected users. This generates an immense amount of load that prevents the async loop from functioning, which in turn causes the disconnections.

Try the following:

async def main():
    async def loop(i):
        await asyncio.sleep(i/10)   # <--- allow a small delay between connections here
        obj1 = Check()
        await obj1.run()
        await obj1.sio.wait()
    x = [loop(i) for i in range(100)]
    await asyncio.gather(*x)

[arun6894]

@miguelgrinberg Thanks. With this small delay between the connections. It is performing pretty well. I can go up to 150 connections as well. 👍

Is it like the more the delay the less the load, that will not prevent the Async loop from functioning ?

When we broadcast 2 messages with just 150 clients take immense load and makes async loop disconnect and again reconnect. Is it something related to hardware ?. I ran this code in i7 8 core 16GB Ram. when I run the code it took just 4sec to make CPU up to 100% from 14%. I assume here the code looks perfect but when CPU can't handle it is behaving like this. Please correct me if I'm wrong.

For eg., if I need 1,00,000 connections 150* no_of_processors which takes up to lots of hardware resources.

I'm trying to understand things here, I hope you understand. In this case is it good to go with synchronous behavior considering the load from Async loop ?

Just need your valuable suggestion to make minimum 1000 connections per process with stability. So a single machine with 6 process can serve at least 5000 connections.

[miguelgrinberg]

Is it like the more the delay the less the load

No. The delay ensures that the load is spread over a period of time, instead of being requested all at the same time. The load is the same.

When we broadcast 2 messages with just 150 clients take immense load

That seems strange. Emitting to 150 clients is not that much, I wouldn't think it would disrupt the application. I tested 100 with your script and didn't seem to have any issues at all.

I ran this code in i7 8 core 16GB Ram

Async applications are single threaded, you are using one core, not 8.

if I need 1,00,000 connections

If you have an application with 1,000,000 connected clients and you need to broadcast a message to every single one of them that is going to require some serious infrastructure and experience. A standard broadcast is probably not what you want in that case, you'll definitely want to make sure you don't send unreasonable amounts of events out to clients, like you are doing now.

is it good to go with synchronous behavior considering the load from Async loop ?

Not sure what you mean here. Sync or async the stuff that your application does is the same, right? So how would things change if you use the same hardware?

[arun6894]

@miguelgrinberg Thanks for testing my script. Could you please confirm me the maximum clients you can connect without any interruption from your end?

[miguelgrinberg]

You continue to look at this from the wrong side. How can I give you a number, since the most important factor that determines how many clients you can take is what you do in your application? This package does not enforce any limits, it is up to you, the application developer, to use the resources that you have at your disposal in a reasonable way.

As I said above, broadcasting to 150 clients is a small task. If you are getting a different output, then my guess is that you are measuring this wrong. Like for example, you say that you are broadcasting to 150 users, but in reality you are broadcasting to 150 users on their connect handler, which means that you are sending 150 broadcasts, which according to my math end up sending over 11K messages and not just 150.