[rabbitmq-discuss] Fwd: Queues and auto_delete

[Ben Hood]

---------- Forwarded message ----------
From: Garrett Smith <g at rre.tt>
Date: Thu, Aug 6, 2009 at 8:54 PM
Subject: Re: [rabbitmq-discuss] Queues and auto_delete
To: Ben Hood <0x6e6562 at gmail.com>


On Thu, Aug 6, 2009 at 2:10 PM, Ben Hood<0x6e6562 at gmail.com> wrote:
> Garrett,
>> Also, queues where auto_delete = true seems to be deleted
>> *immediately* after the last channel is closed. IIRC, the broker
>> should wait for a period of time to allow clients to disconnect and
>> reconnect without losing messages.
>
> This is what the spec says about autodelete:
>
> "If set, the queue is deleted when all consumers have finished using
> it. Last consumer can be cancelled either explicitly or because its
> channel is closed. If there was no consumer ever on the queue, it
> won't be deleted."

This is what I'm referring to (from amqp-xml-doc0-8.pdf):

"amq_queue_02 The server SHOULD allow for a reasonable delay between
the point when it determines that a queue is not being used (or no
longer used), and the point when it deletes the queue. At the least it
must allow a client to create a queue and then create a consumer to
read from it, with a small but non­ zero delay between these two
actions. The server should equally allow   for clients that may be
disconnected prematurely, and wish to re­consume from the same queue
without losing messages. We would recommend a configurable timeout,
with a suitable default value being one minute."

I understand this is a "should", but it seems to me a useful recommendation.

If a client gets (unpredictably) disconnected (e.g. a brief network
problem), the current implementation would immediately delete the
queue, yeah?

I'm switching over from qpid (which does provide a delay before
auto-deleting) so am trying to sort through any redesigns I'll need.
Sounds like I'll want to set auto-delete to false and explicitly
handle queue deletion from the clients.