The correct way of handling connection problems

[greatvovan]

What is the correct approach of keeping application running in case of Rabbit connection problems? I am imitating connection issues by disconnecting network cable (or enabling/disabling network adapter in Windows).

I tried:

• handling AMQP exceptions from protocol functions with subsequent reconnect,
• setting heartbeat=30,
• on_error callback,
• approach from https://github.com/Polyconseil/aioamqp/issues/65#issuecomment-301737344

Neither of this makes library reconnect in my test case. I am getting the following results:

• nothing throws any exceptions and on_error callback does not fire,
• in rare cases library functions throw exceptions and on_error fires, but any further call to protocol function (including connect()) hangs forever, and eventually:
• message callback basically stops being called.

In the first case it is even impossible to identify that a problem exists, as everything looks like just there is no messages. Does anybody have any idea?

My environment:

• CPython 3.6 (natively on Windows 8.1 or in Docker Toolbox),
• aioamqp 0.10.0.

[clodo]

Hi @greatvovan, I have the same problem and I am trying this approach from pika library using futures and add_done_callback. If it works, I'll let you know.

[RemiCardona]

Honestly, without reading your code, it's impossible to say where the issue is. The code in #65 is what I have had in production for months now and it's pretty robust.

Like I said in #65, the on_error callback is busted, don't use it.

If you can't provide a sample code that breaks, there's nothing I can do to help you out. Sorry.