Skip to content


Subversion checkout URL

You can clone with
Download ZIP


reconnect_failed gets never fired #652

Pita opened this Issue · 27 comments

If I stop the node server so every reconnect of the client fails, the reconnect_failed event gets never fired

Here is the position in the code where reconnect_failed should get fired

for some reason it never enters the else part


I am having the same issue. It is reproducible with a simple test case:


var io = require('').listen(4000);

io.sockets.on('connection', function (socket) {
  socket.emit('news', { hello: 'world' });
  socket.on('my other event', function (data) {


<script src="http://localhost:4000/"></script>
  var socket = io.connect('http://localhost:4000',
    {'reconnection delay': 100,
     'max reconnection attempts': 4});
  socket.on('news', function (data) {
    socket.emit('my other event', { my: 'data' });

  socket.on('connect', function () {
  socket.on('disconnect', function () {
  socket.on('connecting', function (x) {
    console.log("connecting", x);
  socket.on('connect_failed', function () {
  socket.on('close', function () {
  socket.on('reconnect', function (a, b) {
    console.log("reconnect", a, b);
  socket.on('reconnecting', function (a, b) {
    console.log("reconnecting", a, b);
  socket.on('reconnect_failed', function () {

hello world

Run node on the app.js file. Open the HTML file in a browser. Open the debug console in the browser. Note that you got a 'connected' event. Control-c the node server. Note that you get a bunch of reconnecting messages, but never a reconnection_failed.

The reconnection options don't seem to make a difference. I included them here so you don't have to wait a long time to see the failure.




the same


+1, here's how I'm dealing with it:

Client side code (Coffeescript):

# Set max reconnects here
max_reconnects = 5

socket = io.connect("http://localhost:3000",
  "reconnection delay": 100
  "max reconnection attempts": max_reconnects

socket.on "reconnecting", (delay, attempt) ->
    console.log "attempting reconnect - " + attempt
    if attempt is max_reconnects
        console.log "all reconnect attempts failed"
        # Show the user an error etc.

This will return the following in the console.log after you Ctrl+C the server:

  1. attempting reconnect - 1
  2. attempting reconnect - 2
  3. attempting reconnect - 3
  4. attempting reconnect - 4
  5. attempting reconnect - 5
  6. all reconnect attempts failed

Same, the issue is still here.


Same here, testing on 0.9.5. The problem happens in the following block:

Plus, redoTransports is always undefined on the first time. Seems this is not considering when try multiple transports has been disabled, which should make the the reconnect_failed to always fire in those cases as well.


Same issue here.


I ran into this problem as well. I've definitely found the cause, hopefully a fix. redoTransports is undefined at first because only tries multiple transports on the last reconnection attempt (this is a side note in the config doc). I think the intention here is to do one last connection attempt with all transports, and finally give up.

The problem is, the handshake is always done with XHR or jsonp. If the server is not reachable, the handshake will always fail. In the handshake code, there is a line:
if (xhr.status == 200) {
} else {
!self.reconnecting && self.onError(xhr.responseText);
this bizzare, hard to read line !self.reconnecting && self.onError(xhr.responseText); means that self.onError will only be called if self.reconnecting is false. (I don't like this kind of shorthand - use an else if). Normally, there is a reconnect timer that will fire, triggering another call to reconnect. However, the timer is not set up on the last reconnection attempt. This code path never runs, meaning reconnect_failed is never called.

The solution I think is to also start the timer on the final reconnection attempt. Once I figure out how, I'll submit a pull request for a patch.


+1 I can't get connect_failed or reconnect_failed to emit


I managed to fix the problem by changing this section:

if (!self.redoTransports) {
  self.on('connect_failed', maybeReconnect);
  self.options['try multiple transports'] = true;
  self.transport = self.getTransport();
  self.redoTransports = true;
} else {

After the self.connect() line I added:

self.publish('reconnecting', self.reconnectionDelay, self.reconnectionAttempts);
self.reconnectionTimer = setTimeout(maybeReconnect, self.reconnectionDelay);

This causes the event to get fired properly after the last failed attempt.


Nice, the connection failure event is trigging for me with that mod.


@robbrit i just found that the changes what you mentioned are already there in my but still the reconnect_failed event is not getting fired. And also one more thing what i found it, even if i specify 10 reconnect attempts it tries only 5 or 6 times and stops. Please any one help to solve this.


@veeru-as The first code block is what is already there, the second code block is what I added. Also keep in mind that the reconnect delays are exponential, so after a while it seems like it has stopped but it is just waiting a long time.


Is there any partical reason why this reconnect_failed is in the docs but it never triggers?


Is there any reason why this bug has not been fixed??


I went through it in chrome. The variable causing the issue is "self.redoTransports" because it is always undefined. This is most likely a breakdown in the config section of the code.




The fix from robbrit still appears to work for me in 0.9.11.


Can we please get this pushed in? ffs.


@robbrit Thanks for the fix. Helped me fix the issue I was having.


Can anybody tell what redoTransports does? I cannot find it anywhere throughout the codebase and it is never set, so it always enters the if(!self.redoTransports)... which sucks. Removing the negation for the condition works just fine then: if(self.redoTransports)...


@robbrit Have you ever had any problems with your fix ?

I implemented your fix, it fixed the issue with reconnect_failed BUT something seems to be going wrong with re-connecting to the server, the client is creating more 'new connections' than it should.

Maybe it's just how i'm starting from the client, these are my options.

socket = io.connect(url, {
        'connect timeout': 5000, // 5 seconds should be enough
        'try multiple transports': true,
        'reconnect': true,
        'reconnection delay': 500,
        'reconnection limit': 5000,
        'max reconnection attempts': 3,
        'sync disconnect on unload': false,
        'auto connect': true,
        'flash policy port': 10843,
        'force new connection': true //IMP: to be able to re-connect using our own logic

I'm posting this just for the sake of other people who might come across the same issue. I'm not using re-connect failed anymore because i switched to: 'max reconnection attempts': Infinity


@kurteknikk: we had a problem where we would get connections, but they would never fire the disconnect event and so the server would never know when the clients would disconnect. I didn't bother fixing it, we ended up switching over to SockJS, which solved a lot of the problems we were having.


Can I suggest this at least goes in for the 1.0 release?

@mokesmokes mokesmokes referenced this issue in socketio/

reconnection fixes #564


Would be nice, still need it 2 years later.


Agreed ... just ran into it myself this morning :(

@jbg-zuba jbg-zuba referenced this issue from a commit in JackboxGames/
@jbg-zuba jbg-zuba Added fix from socketio/ 80b3daa
@agilethomas agilethomas referenced this issue from a commit in agilethomas/
@agilethomas agilethomas Fixed issue with reconnect_failed event not getting triggered
See robbrit's comment in this thread:
This issue was closed.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Something went wrong with that request. Please try again.