On 26/04/12 23:42, Matthew Sackman wrote:
> However, the same persistent msg sent to many durable queues only gets
> written to disk once. Each queue has to do a small amount of on-disk
> writing to record that the msg is in the queue, but the msg itself only
> gets written once. Thus the extra cost of accidentally sending the msg
> to another durable queue should be very low.
Indeed. From the OP's description I rather suspect the bottleneck is not
rabbit but the consumer which is writing messages to mongodb.