Fixed exponential reconnect not waiting twice as much each time

[bakkerthehacker]

Fixed exponential reconnect not wait twice as much. Moved possibleAuthErrorHandler to addAllListeners. Fixed reconnect removing all externally added events.

[bakkerthehacker]

Any update on this?

[bcjwilson]

Post merge, still seeing memory leak problem due to reconnect event handlers.

[bcjwilson]

Belay that, working locally. How do we get this out to npm?

[bakkerthehacker]

I'm assuming that postwait or someone else will decide that the code is stable enough and that enough new features have been merged in, rev the code, and push to npm.

[bcjwilson]

Thanks for fixing that. I'm now seeing a new problem, not sure whether this is expected behavior.

Example code:

  // not running RabbitMQ locally
  var connection = amqp.createConnection();

  connection.on('error', function(err){
    console.log("ERROR: " + err.message);
  });

Output (one second between each line):

ERROR: connect ECONNREFUSED
ERROR: connect EADDRNOTAVAIL
ERROR: connect EADDRNOTAVAIL
...
ERROR: connect EADDRNOTAVAIL

The first connection attempt is refused, after that it's address not available. Shouldn't they all be connection refused?

[bakkerthehacker]

i could look into it

[bcjwilson]

Thanks, I'm not yet familiar enough with the amqp source to make a suggestion. The little research I've done suggests that EADDRNOTAVAIL means you're trying to bind a socket that's already bound. But connecting from a client-side app shouldn't have to do that, right?

Maybe it has to do with keeping the old connection around?

Side question: Is there a way to turn off the reconnect logic from client code? When RabbitMQ dies, I have to rebuild the world on the client side anyway (reestablish transient exchanges, make a new queue etc.) So I'd kinda like to just do the reconnecting myself.

[bakkerthehacker]

https://github.com/postwait/node-amqp#connection-options-and-url

[bcjwilson]

Thanks, that's just what I needed - also, looks like the patch made it to npm :)