heartbeat support for AMQP

[zstyblik]

Please,

either enable or make configurable heartbeat in mod AMQP. This should be configurable and/or enabled.

https://github.com/balabit/syslog-ng-3.5/blob/master/modules/afamqp/afamqp.c#L334
https://github.com/omniti-labs/Net--RabbitMQ/blob/master/amqp.h#L234

Thank you.

[Edited by @Kokan]

As rabbitmq already supports heartbeat, from syslog-ng side it only has to make it configurable. For details see amqp_login function.

[algernon]

As far as I understand, just setting that option to some configurable value would not be enough, syslog-ng would need to send heartbeat packets too, if there's no other traffic. Doing that is a bit more involved than adding a configurable parameter.

Am I seeing it correctly? Or is it enough to set that argument, and receive heartbeat frames from the server, without sending any ourselves?

[zstyblik]

I'm not familiar with C/C++ implementation of AMQP and higher-level languages do this for you. So it really is hard to tell. One thing, though. You must not block indefinitely and must keep polling. Otherwise, you might miss communication from MQ server which ultimately(?) leads to being kicked out and connection terminated.

[zstyblik]

Am I seeing it correctly? Or is it enough to set that argument, and receive heartbeat frames from the server, without sending any ourselves?

Actually, I believe server itself doesn't send any heartbeats and it's up to client to send those.

[zstyblik]

https://stackoverflow.com/questions/21020945/how-to-enable-server-side-heartbeat-for-rabbitmq <<< not exactly reliable source, but I guess it makes sense. If server doesn't hear from client in defined/agreed heartbeat, then it proclaims client to be dead and will close connection.

[algernon]

Mmmhm. That's a bit harder to implement, but I have an idea: using ivykis timers might just get the job done in a neat way.

[ihrwein]

@dnsjts could you look at this?

[furiel]

As a hint, here is how I created a working environment for development. With better understanding of rabbitmq, there is probably a simpler way. Please share if you know one!

1. Starting rabbitmq server from dockerhub.

$ docker run --name rabbitmq --net=host -ti rabbitmq

2. Adding logs exchange for vhost "/".
   For some reason the default "" exchange did not work with the example script from the rabbitmq tutorial, so I needed to create another one.

• Enable webui:

docker exec -ti rabbitmq rabbitmq-plugins enable rabbitmq_management

This will start a web service on http://localhost:15672

• Add exchange:
  In the webui with guest:guest. On the Exchanges tab, fill the form
  name: logs
  type: fanout
  durability: transient
  (rest with default value)
  Add exchange.

3. start syslog-ng with the configuration example below. Comment out example-msg-generator() if you do not want periodic messages.

@version: 3.20

log {
    source { stdin(flags(no-parse)); };
    source { example-msg-generator(); };
    destination { amqp(vhost("/")
                  exchange("logs")
                  body("hello world")
                  username(guest) password(guest)); };
};

To read logs, one can use the python snippet provided in the rabbitmq tutorial:

#!/usr/bin/env python
import pika

connection = pika.BlockingConnection(
    pika.ConnectionParameters(host='localhost'))
channel = connection.channel()

exchange="logs"

channel.exchange_declare(exchange=exchange, exchange_type='fanout')

result = channel.queue_declare('', exclusive=True)
queue_name = result.method.queue

channel.queue_bind(exchange=exchange, queue=queue_name)

print(' [*] Waiting for logs. To exit press CTRL+C')

def callback(ch, method, properties, body):
    print(" [x] %r" % body)

channel.basic_consume(
    queue=queue_name, on_message_callback=callback, auto_ack=True)

channel.start_consuming()