[rabbitmq-discuss] RabbitMQ QUEUE MESSAGES

[Ceri Storey]

(02/07/13 17:17), rabbit15 wrote:
> Hi thanks for the answer, but it is not what I'm looking for.
> I don't want to drop any messages, because my producer is faster than
> consumer, but not all the time and all the data is necessary.There is time
> when it doesn't produce any messages. I don't want to have any other buffer
> of course, but RabbitMQ doesn't let my producer to sending messages even if
> there is a lot of free memory. I'm asking how can I setup RabbitMQ to remove
> that buffer. My server hardware is 32 CPUs and 64 GB RAM, I don't think I
> need better hardware, because even if I store about 8 millions messages
> which takes about 600 bytes (everyone) than it turn on horrible flow control
> which prevents sending as fast as I need. Please tell me how can I turn off
> flow control. My application wants to queued messages on RabbitMQ server in
> its memory not disk and I don't want to be prevented by RabbitMQ sending by
> producer or consuming messages by the consumer.
If you need to store an unbounded number of messages in the queue, then
as i see it, you've got two options:

 1. Use the rabbitmq-toke plugin, which allows messages to be stored in
    tokyo cabinet. That said, I don't know how well supported and/or
    documented that is.
 2. Use another queue server, such as bitly's nsq, beanstalkd, or even
    Kafka.

Assuming that you've got very basic requirements in terms of your
messaging protocols (ie: one message to one consumer), then option 2)
might well work out better for you, to be honest. Rabbit makes a lot of
trade-offs that focus on flexible messaging topologies and reliable
delivery, all of which can work against raw throughput, and especially
the ability to act as an infinite buffer.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20130702/b693ec67/attachment.htm>