[rabbitmq-discuss] Acked Message getting redelivered

Thu Jan 19 12:33:06 GMT 2012

On 19/01/12 08:45, Iain Hull wrote:
> From: Eugene Kirpichov [mailto:ekirpichov at gmail.com]
>> Acks are indeed asynchronous.
>
> So this means that the only reliable way to implement "at most once
> delivery" is to remember the messages uuids that you have processed,
> which won't work for clustered consumers.  Is this correct or is there
> another way to achieve this?

It sounds like you find losing messages slightly acceptable (certainly 
better than getting dups) and are thus only using acks because you are 
using qos. Is that right?

If so then the simple thing to do would be to just ignore (after acking) 
any messages which have the "redelivered" flag set.

Cheers, Simon

-- 
Simon MacMullen
RabbitMQ, VMware