I have developed a chat application using Pusher. I want to disconnect all the inactive or unwanted connections so as to avoid crossing the peak limit. How can I do that?
You'll firstly need to identify which connections can be terminated. You could implement this with a timer to determine if a user is inactive or even allow the user to mark themselves as inactive somehow. Once you've done this, you've identified a connection to disconnect. With the Pusher Java client API, it's as simple as calling disconnect().
https://pusher.com/docs/client_api_guide/client_connect#disconnecting
Hope that helps!
Related
I kinda came across a strange problem.
In our application (based on React-native) we hosted 70 concurrent clients but the peak in the monitoring page showed 380 connections.
I assume maybe clients exit and come back or reload the app somehow so Ably connections regenerates again and therefore the peak increases.
Now the question: is there any way to force Ably disconnect all unused connections so the peak decreases? (Maybe from back-end)
Thanks.
By default, the connection will stay active until closed explicitly (using connection.close()), or two minutes after the connection is disconnected unexpectedly to allow for connection state recovery.
Recent versions of ably-js in a browser environment automatically close the connection on page reload (that is, the closeOnUnload client option defaults to true) -- this is just a connection.close() added to a beforeunload handler. The trouble is that isn't going to do anything in a React Native environment, which doesn't use that event.
So you probably just need to actively manage your Ably connection using React Native app lifecycle events. If you don't want it to stay active when the app is backgroundend, then in the handler for the app being in the background (per the React Native AppState event), close the Ably connection. Then re-open it (call connect()) when the app is active again.
For other possible reasons your peak connection count may be higher than expected, see Why are my peak connection counts higher than expected? and How does Ably count peak connections?.
Socket.IO timeout (disconnect) occurs if there are no activity present in the socket, but how is detected that are no activity? I dont find information about that. How works the detection process? For example:
If user closes the website tab, occurs disconnect?
There is any way user loses connection and disconnect will not be executed?
Is possible cache information informs that Im online even if I leave the website?
If you open your browser's developer tools in the network panel, you can filter your requests to ws requests (web socket requests), in there you can see your active web sockets connections. If you choose one connection, you can see the headers, the frames, the cookies and timing. If you choose the frames option you can see what's being sent and received, between your browser and your web socket server.
The next image will make it clear for you, it's chrome's developer tools:
Now in there you can see there are some numbers, basically your browser and your server are doing ping pong. You can read more about these numbers in this answer SocketIO Chrome Inspector Frames
This ping pong is what keeps the socket alive so we know that there are no timeouts. As for the disconnect and the connect I advise you to read more about the WebSockets API, in there you can see there are event listeners for onclose, onerror, onmessage and onopen.
So answering this question:
For example: If user closes the website tab, occurs disconnect? There
is any way user loses connection and disconnect will not be executed?
No, the onclose event will be fired, but even if any cosmic reason the onclose isn't fired you will eventually disconnect due to timeout.
As for your other question:
Is possible cache information informs that Im online even if I leave
the website?
Yes, that's not up to sockets, that's up to you and your implementation. You can keep a list of online users and only update that list from time to time, let's say 10 minutes. You can keep the online users lists and between the time you update your online users list, some of them are already disconnected.
I am new to node.js and trying to develop group chat using node.js and socket.io. I am able to do group chat also able to manage data in the system.
Now, the problem with me is with offline users i.e not connected to internet.
I am having connection stream of this user and as if I do receiverUserSocket.emit("sendMsg",data) there is no way to verify if this user received message or not.
Yes, there is an event .on("disconnect") but I am getting delay of approx 30 - 40 seconds.
Is there any way we can identify that the user to whom we want to send message is online or offline.
The best way is to use socket.io heartbeats - assuming you're using socket.io.
Heartbeats and the problem with detecting a terminated connection are well explained here:
Advantage/disadvantage of using socketio heartbeats
Keep in mind that you can control heartbeat timeout values (thus forcing the disconnect event to appear much faster), as explained here:
Controlling the heartbeat timeout from the client in socket.io
But that might put much more strain on your server.
I am using webrtc.io to create the socket connections for my audio, video chat application. I want to preserve all the socket connection so that I can send updates to all the end users when the node.js server is restarted.
I am using Mongodb as the database for this application. Is there any way to store in the database and retrieve it back when the server is restarted?
I'm going to give you a common life situation to explain this.
Suppose you have a mobile phone that you cannot make calls from and you can only receive calls.
Someone calls you and you can talk to them, messages pass backwards and forwards on a constant connection. This was better than SMS because you could only respond to an SMS that was sent to you as well but now you have this constant connection to talk freely on.
Now in those statements I just described what Websockets are and the difference between that an Http. Next I'll apply this to what you are asking.
Now suppose on this phone where you can only keep talking on calls you receive from someone else, your battery runs out. You find a power source to plug into and get your phone working again. So do you expect your phone to just suddenly re-establish the call that dropped when your battery ran out?
You do not initiate the connection you are talking about. So you cannot "make the call back" or "re-establish the call". This is a strictly "the customer calls you" scenario.
The best you can do is maintain the session state to the subsequent re-connection "picks up where you left off". But on a hang-up the client has to call you back.
For better availabilty you need to proxy the connection and share over multiple application server nodes, all with access to the same session state.
So I am developing (more playing around with) a realtime game in node.js, I am also using Redis and Sockets.io. I have players create a lobby and join it (kind of like a pre-game chat room, where you can talk to players and select game settings) . The client is written in HTML/CSS/JS, Anyway I want to be able to tell when players disconnect from the lobby, to update the number of players joined on the interface (and joined player names).
Two options I have thought about are:
Using redis' key value timeout feature, to remove a particular field if it is not updated in x amount of time. I would then have the host check the existance of this field to check for DC's. I do wonder if this is highly inefficient, as many users potentially will be playing, so will it be bad to have many timeout values in redis and also many other users polling these fields.
I could use the sockets.io on('disconnect', ..) to update the field. However I am not sure if this event will fire if for example a users pc freezes?
Anyway I am open to any other ideas also!
Socket.io have a 'heartbeat' to check connection still alive. Default heartbeat timeout is 15s. You can read more about configuring it in this wiki. If heartbeat fails (user pc freezes) then socket.io will emit 'disconnect' event.
Socket.io should suffice. You can configure it to use heartbeats to ping the socket and check its health. If a user's computer freezes it will, in effect, not be able to respond to these heartbeats, causing it to force a disconnect.
To test this you could set up your Socket.io to use heartbeats, then connect via a browser onn a different computer. While in the browser past into the console an infinite loop. Causing it to simulate a freeze.