[rabbitmq-discuss] my gen_fsm's supervisor getting channel 'DOWN' messages

[Ben Hood]

Hi Jack,

On Mon, Jun 22, 2009 at 12:55 AM, Jack Moffitt<jack at collecta.com> wrote:
> I have a gen_fsm that monitors a rabbitmq queue.  Everything works
> fine unti lthere is an error.
>
> When I connect and open the channel, I put a monitor on it so that I
> get a 'DOWN' message when the channel dies.  Before I down the channel
> myself, I always demonitor the reference.
>
> Unfortunately, for some reason whenever I close the channel and
> connection, the supervisor starts trying to shut down my process.  I'm
> opening the connection with start() not start_link(), so I don't
> understand why my supervisor is being notified about AMQP connection
> process states.
>
> This is actually causing us serious problems because Erlang
> supervisors can't do backoff, so this usually results in the
> supervisor exceeding the restart limits and terminating the whole
> application.

Alexis has brought this to my attention, and I gather that there has
been some initial discourse, but unfortunately I don't have anything
with which I can diagnose the issue.

Is it possible for you to send a cut down version of some code that
provokes this symptom and also the log or error output that it
produces for you?

Thx,

Ben