Node.js process dies if RabbitMQ node dies.

[aangelidis]

Hi,

I have the following.

const servicebus = require('servicebus');
const retry = require('servicebus-retry');

bus = servicebus.bus({
    url: 'amqp://localhost:6672'
});
bus.use(retry({
    store: new retry.MemoryStore()
}));

// This will set autoDelete to false, but we must manually ack messages.
var options = {
    ack: true
}

bus.subscribe('events.*', options, (msg) => {
        msg.handle.ack(() => {
            console.log('Acked message: ' + JSON.stringify(msg));
        });
});

console.log("events-management started.");

The moment the Rabbit node where servicebus is connected to dies (dokcer kill that 1 node), the Node.js process dies also with no exception caught.

We only see: Process finished with exit code 0

[mateodelnorte]

By adding a bus.on('connection.closed', fn), you can latch on to the connection closing:

const servicebus = require('servicebus');
const retry = require('servicebus-retry');

bus = servicebus.bus({
  url: 'amqp://localhost:5672'
});
bus.use(retry({
  store: new retry.MemoryStore()
}));

// This will set autoDelete to false, but we must manually ack messages.
var options = {
  ack: true
}

bus.subscribe('events.*', options, (msg) => {
  msg.handle.ack(() => {
    console.log('Acked message: ' + JSON.stringify(msg));
  });
});

console.log("events-management started.");

process.on('uncaughtException', (err) => {
  console.log(err);
});

bus.on('connection.close', console.log.bind(this, 'connection.close'))

This probably should be emitted as an error on bus, itself. I'll spin up a quick PR, reference you, and you can see if it fits what you'd like.

[aangelidis]

Thanks. :) I'm wondering if it's worth putting a reconnect option with retries. Especially if using HAProxy load balancing.

[aangelidis]

Btw, this doesn't fire...
process.on('uncaughtException', (err) => { console.log('Here: ' + err); });

This does...
bus.on('connection.close', console.log.bind(this, 'connection.close'))

[mateodelnorte]

I'm debugging the uncaughtException issue now. Looks like it's coming from amqplib. Just raised the following issue there: amqp-node/amqplib#334

For now, the best thing to do is listen for connection.close. I'll update here as soon as I hear back.

In general, I promote crashing your application and restarting when an error occurs (https://www.joyent.com/node-js/production/design/errors). If you'd like to write a module that wraps servicebus and provides retry-connection with incremental backoff, I'd be happy to link to it.

[aangelidis]

Ok. I'll track it here.

[javadevmtl]

Btw is there a way to contact you so we can discuss the docs a bit see if I can document some of it for you? :)

[mateodelnorte]

Yup. Shoot me a dm on nycnode.com/slack

[mateodelnorte]

Circling back here with a better way to communicate about helping out!

https://gitter.im/servicebus-node/Lobby

Cheers!