[rabbitmq-discuss] Publisher Confirms Multiple Acks and Multiple Nacks behavior

[Simon MacMullen]

On 13/11/12 22:00, Eric Swann wrote:
> General question about publisher confirms ack/nack expected behavior,
> more along the lines of multi-ack/multi-nack
>
> Does a multi-ack/nack occur and any time and just encapsulate a batch
> since the last ack or nack that happened?

Yes.

> Example:
> Lets say RMQ has sent me several individual confirm acks, no issues...
>
> I then get a multi-ack.  Does that just indicate that everything in the
> range since the last ack is now acked?

Yes.

> Now if I had some catastrophic issue, could I expect to get a multi-nack
> basically grouping several failed messages together (basically meaning
> that everything in the range since the last ack is now nacked)?

Yes.

> Assuming this multi-nack occurs, but things come back online, could I
> then receive a multi-ack for all of the messages that have succeeded
> after the multi-nack?

Yes.

> If that is the case, I would have to keep a buffer queue of all the
> ack/nacks and process them in order (since they are processed out of
> band).  Would that be the recommended approach?

Yes.

Hope this helps :-)

Cheers, Simon

-- 
Simon MacMullen
RabbitMQ, VMware