[rabbitmq-discuss] correct configuration for large message volume?

[Matthias Radestock]

James,

On 26/12/11 05:31, James Carr wrote:
> I've been doing some experiments with rabbit over the holidays and one
> thing that I run into that always bugs me is if I get a queue filled up
> with >  1 million messages being published continuously (but not
> consumed) on my laptop I usually wind up with rabbitMQ getting into an
> "odd" state.

OS?

RabbitMQ version? Make sure it's 2.7.1.

Erlang version? Make sure it's R15B, particularly if you are on Windows.

> By odd I mean:
> * management console lists "?" for the number of messages in the queue

That is strange indeed.

> * calling rabbitmq-server stop doesn't stop the broker (I wind up having
> to do a hard kill -9).

The most likely reason is that the broker is "catching up", i.e. 
processing messages that where hitherto held in internal buffers.

Does 'rabbitmqctl status' still work?

> * CPU usage usually sticks around 100% (even long after the producer has
> been shut down)

Ditto.

> * Memory high watermark is not met (not even close

The HWM is set to 40% of available RAM by default since the broker can 
easily go over that by some margin. Does it actually exceed 100% in your 
case?


Matthias.