[rabbitmq-discuss] Which RabbitMQ approach/pattern to use?

[Emile Joubert]

Hi,

On 23/04/12 20:47, theBuckWheat wrote:
> My application sends a work file out to one of several available remote

If the application does not need to designate a slave then work files
can go into a shared queue that all slaves subscribe to.

If your application must designate a slave then you could create a
dedicated queue for each slave.

> slaves.    When the slave is done with the work, it posts a callback notice
> and stands by for the master to fetch the results file.   When that file has
> been downloaded, after a little housekeeping, the slave can be put on the
> list of those available to send work.  

That sounds like an RPC pattern. The 6th tutorial discusses this:
http://www.rabbitmq.com/tutorials/tutorial-six-python.html

> What is the best que structure to use with RabbitMQ to dispatch work to the
> pool of slaves?  I have to manage the list of slaves because some might be
> close to their bandwidth limits, so it does not appear I can use the
> load-balance pattern.   

It is not clear who is responsible for observing this constraint. If it
is the slave then it could take itself out of service. If it is the
application then you will need a separate queue per slave.


-Emile