[rabbitmq-discuss] fanout exchange that does not keep messages

[Brian Whitman]

Hmm... this is not working the way I expect.
If I do either Holger or Matthias's solution the write-only "logger" queue
continues to grow but the listener queues do stay at 0:

~/rabbitmq-server/scripts# ./rabbitmqctl list_queues name mode messages
messages_ready messages_unacknowledged durable
Listing queues ...
logger_2b02ca30-754b-11de-99c0-0023dfde57d8 mixed 0 0 0 false
logger mixed 110 110 0 false

What I would like to see here is that logger is always 0 messages. If there
are listeners or not, I don't want the queue called "logger" to ever retain
any messages.

For both "logger" (write only) and the random queue names (read only) I am
doing

    ch.exchange_declare(my_exchange, 'fanout', durable=False,
auto_delete=False)
    qname, n_msgs, n_consumers = ch.queue_declare(queue_name, durable=False,
exclusive=False, auto_delete=True)







On Mon, Jul 20, 2009 at 12:20 PM, Matthias Radestock <matthias at lshift.net>wrote:

> Brian, Holger,
>
> Holger Hoffstätte wrote:
> > Brian Whitman wrote:
> >> I am concerned about this because over the weekend the broker crashed
> >> (not sure why yet, but I assume it had something to do with probably
> >> ~50m messages sitting in the exchange.) How can I have messages be
> >> completely transient?
> >
> > This is done with the "immediate" flag for a message; if there is no
> > recipient, the message will be dropped.
>
> Actually, generally a better solution to the problem is to ensure that
> the queues do not outlive the consumers - by either marking them
> exclusive (if you only have one consumer per queue) or auto-delete.
>
> That way messages do really get dropped silently when there are no
> consumers. By contrast, when setting 'immediate' the server will send a
> basic.return to the producer.
>
> And buffering can still take place to deal with temporarily
> backlogged/slow consumers. By contrast, with 'immediate' the messages
> will get dropped when the consumers are busy (not just when they are not
> present).
>
>
> Regards,
>
> Matthias.
>
> _______________________________________________
> rabbitmq-discuss mailing list
> rabbitmq-discuss at lists.rabbitmq.com
> http://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20090720/79eeae2f/attachment.htm