[rabbitmq-discuss] does rabbit eat memory?

Fri Nov 26 16:51:51 GMT 2010

On Fri, Nov 26, 2010 at 04:44:51PM +0000, Robert Fuller wrote:
> Thanks Mathew, this makes sense and also maybe sheds some light on the
> larger picture which is this: Over a number of hours during our
> producers were producing a few million messages at about the same rate
> the consumer was able to process them. However, because the consumer
> started after the producer, there remained at all times some thousands
> of messages in the queue. Memory consumption seemed to rise and rise.
> Is it only when the queue reaches a zero size that the memory is
> cleared?

No - provided the queue is staying at a constant length, and the
consumers are acking messages (the "messages unacknowledged" field
should stay constant and low), the memory usage should not grow
continuously.

> One I can think of would be to maintain a rotation of queues... after
> putting a million messages onto any queue move on to the next queue.
> The consumer, finding no messages on a queue will move to the next
> queue in the rotation. Is this reasonable?

It's reasonable, but should be utterly unnecessary.

Matthew