[rabbitmq-discuss] Checking consumer message delivery

[Chris Duncan]

On 19/12/12 10:34, Simon MacMullen wrote:
> On 19/12/12 09:08, Chris Duncan wrote:
>> I'm doing some testing around consumer behaviour and I'd like to confirm
>> that what I'm seeing on the wire is expected RabbitMQ behaviour. My
>> environments are running Ubuntu 12.04.1 LTS (32bit and 64bit), Erlang
>> R14B04 and RabbitMQ 3.0.1.
>>
>> In my simple test I publish one message to a queue. Then I register a
>> consumer on that queue. What I see coming back from the server (via
>> Wireshark) is a frame that contains a consume-ok and a delivery message
>> set as follows:
>
> <snip>
>
>> Is this what you would expect to see or should the consume-ok and
>> delivery set be returned as separate frames? Any help would be much
>> appreciated.
>
> AFAICS that's one *ethernet* frame containing two AMQP methods 
> composed of four *AMQP* frames. (The deliver method consists of three 
> AMQP frames).
>
> We make no attempt to make AMQP frame boundaries match underlying IP 
> packets / ethernet frames - it's transparent to clients, and it would 
> be less efficient for small messages and impossible for large ones.
>
> Cheers, Simon
>
Thanks Simon, that clarifies my understanding of ethernet frames as 
opposed to AMQP frames.