I'm sorry if this is the wrong stack for this question.
I have a Node.js server running on Heroku. Whenever I commit something, the server restarts. I want to warn the users that there's going to be a restart so that I don't completely ruin their experience. Just a simple notification with the text "Server restart in X minutes".
Let's say I have the client side all set up and a Socket.IO emit is all that's needed for the notification to be shown. How would I do that? I thought of having some sneaky function on the client side that would make the server emit the notification, but I'm afraid it can be easily exploited.
The answer is to detect the SIGTERM signal that Heroku sends to your app to shut it down, and once that signal is detected, to emit the notification to every connected client:
process.on('SIGTERM', () => {
// send your signal
})
https://devcenter.heroku.com/articles/dynos#shutdown
However, a better user experience is just to ensure that users never even know your server restarted. With node this shouldn't be difficult, you only need:
To ensure your process ends and starts quickly (quickly being a few seconds)
To ensure your clients all have reconnection logic (this is built into socket.io so you shouldn't have to do anything)
Optionally, to turn on preboot, which will make the delta between one server going down and another coming up be close to zero (heroku features:enable -a myapp preboot)
https://devcenter.heroku.com/articles/preboot#enabling-and-disabling-preboot
Related
Is there any way (except polling) to watch other processes start event in NodeJS?
I want to pop a notification when a user runs the black-listed apps.
I think you can use a websocket for this. A websocket is a protocol which remains the connection with the user. After you then detect that they run a forbidden application you can emit a signal which then can create this notification.
Socket.io is a convenient NPM package for websockets.
Hopefully this helps!
I'm building MEAN application with socket.io. When page is just loaded, socket connection is established and kept live while user navigates to various pages, thanks to single-page nature of the app.
The user information is available in my socket connection thanks to passport.socketio.
However, when user logs in or out, I want the connection to be re-initialized, since otherwise socket will contain obsolete data about the user. Currently, I tried to implement it in this way: when user logs in / out, server disconnects this particular client's socket by calling socket.disconnect();.
On the client side, I listen for disconnect event, and try to re-establish the connection, like this:
_socket.on('disconnect', function(reason) {
_socket.connect();
});
Ok, now, when user logs out or in, server disconnects the client, this client connects back, and user information in the socket is up-to date. So far, so good.
But, consider different case when connection is broken: server is restarted. Previously, it "just worked": when I stop my server, connection is broken, but when I start server again, connection is automatically re-established. But after I've added my _socket.connect(); call, it doesn't work anymore: connection is still down until I refresh the page in the browser.
I've checked that when server calls disconnect();, the reason given to disconnect handler is: io server disconnect. And when server stops, the reason is: transport close.
Ok, then, I've implemented my disconnect handler as follows:
_socket.on('disconnect', function(reason) {
if (reason === 'transport close'){
// don't do anything special
} else {
_socket.connect();
}
});
Now it works. But, all of it seems as absolute dirty hack. At the very least, the reasons given (io server disconnect and transport close) seem to be just human-readable strings, so they might change in the future, and this will cause my code to stop working. And, well, there should be better way to do this; I must miss something essential, but unfortunately I can't find any good documentation on socket.io.
So, the question is: what is the correct way for the server to reconnect some particular client?
Additionally, if you have any recommendations on resources to learn about socket.io, I'd appreciate it very much as well.
I am going to design a system where there is a two-way communication between clients and a web application. The web application can receive data from the client so it can persist it to a DB and so forth, while it can also send instructions to the client. For this reason, I am going to use Node.JS and Socket.IO.
I also need to use RabbitMQ since I want that if the web application sends an instruction to a client, and the client is down (hence the socket has dropped), I want it to be queued so it can be sent whenever the client connects again and creates a new socket.
From the client to the web application it should be pretty straightforward, since the client uses the socket to send the data to the Node.JS app, which in turn sends it to the queue so it can ultimately be forwarded to the web application. From this direction, if the socket is down, there is no internet connection, and hence the data is not sent in the first place, or is cached on the client.
My concern lies with the other direction, and I would like an answer before I design it this way and actually implement it, so I can avoid hitting any brick walls. Let's say that the web application tries to send an instruction to the client. If the socket is available, the web app forwards the instruction to the queue, which in turn forwards it to the Node.JS app, which in turn uses the socket to forward it to the client. So far so good. If on the other hand, the internet connection from the client has dropped, and hence the socket is currently down, the web app will still send the instruction to the queue. My question is, when the queue forwards the instruction to Node.JS, and Node.JS figures out that the socket does not exist, and hence cannot send the instruction, will the queue receive a reply from Node.JS that it could not forward the data, and hence that it should remain in the queue? If that is the case, it would be perfect. When the client manages to connect to the internet, it will perform a handshake once again, the queue will once again try to send to Node.JS, only this time Node.JS manages to send the instruction to the client.
Is this the correct reasoning of how those components would interact together?
this won't work the way you want it to.
when the node process receives the message from rabbitmq and sees the socket is gone, you can easily nack the message back to the queue.
however, that message will be processed again immediately. it won't sit there doing nothing. the node process will just pick it up again. you'll end up with your node / rabbitmq thrashing as it just nacks a message over and over and over and over, waiting for the socket to come back online.
if you have dozens or hundreds of messages for a client that isn't connected, you'll have dozens or hundreds of messages thrashing round in circles like this. it will destroy the performance of both your node process and rabbitmq.
my recommendation:
when the node app receives the message from rabbitmq, and the socket is not available to the client, put the message in a database table and mark it as waiting for that client.
when the client re-connects, check the database for any pending messages and forward them all at that point.
I have a web based multiplayer game. It happens from time to time that someone is kicked out because server did not get expected message from client. It seems from my logs that client did not disconnect, just did not send message or server did not receive it. My question here is "Does this things happen normally from time to time?" Should i use some kind of callback mechanism to ensure message is delivered and if not send it again or is there some issue that i am not aware?
socket.io already provides ACKs and message ID tracking, on top of TCP.
Also, socket.io uses pings to check the connection. So, if you say that the client is not disconnected, and the server tells that the client is not disconnected, then the connection is still there.
The problem must be situated elsewhere.
Are you sure there is not a bug in either part of the implementation? Showing some code snippets could help, as well as the environment you are using.
I have a socket.io client connected to a node.js server. If I kill node.js at the command line, the client immediately freezes (i.e., communication stops), but there is a ~20 second delay before the "disconnect" event is fired. Is this behavior by design? Is there a configuration option to reduce the delay in firing the disconnect event?
It appears that this behavior changed in a relatively recent (last 6 months) update of socket.io. Before the reconnect functionality was built in to socket.io itself, I implemented my own reconnect logic using a "disconnect" event handler and at that time the "disconnect" event fired almost instantly when server communication halted.
I think this is likely a design pattern. The client may be presuming the server is 'temporarily' unreachable (network trafic etc) and essentially will keep trying to reach it... until the client timeout kicks in.
I send a disconnect (socket.disconnect()) to the server directly from the client, and I don't get this issue.