[rabbitmq-discuss] Dead-Lettering Performance

[Emile Joubert]

On 20/03/13 15:11, Maciej Lopacinski wrote:
> My question:
>  - is it valid use of message expiration time?

It is valid to use TTL to implement delayed delivery. Whether it will
meet your expectations in this case is hard to say. If a message (to be
delayed by 110s) is published before another message (to be delayed by
10s) to the same queue then they will be expired almost simultaneously
after 110s.

>  - are there any caveats around dead-lettering performance?

Dead-lettering a large number of messages could cause a spike in memory use.

>  - are there any guarantees around dead-lettering performance?

I'm not sure what kind of guarantee you require. The possible memory
spike will be fixed in a future release.

>  - does dead-lettering has same "priority" as others "internal
> operations" in RabbitMQ broker?

The Erlang scheduler treats processes performing dead-lettering the same
as all the other broker processes.



-Emile