On 14 August 2015 at 15:29:18, Micheal Pike (
mpik...@gmail.com) wrote:
> I have a Subscriber that is installed as a windows service on
> the same server as RabbitMq is installed. Before starting the
> windows service, the erl process is taking up about 6% cpu. However,
> after I start the windows service, the erl process jumps to and
> stays at 65% and the windows service stays at around 14% cpu utilization
> which means that between the two processes the cpu utilization
> stays around approx 70% - and this is only one consumer! If I restart
> RabbitMq the cpu utilization proceeds back to the approximate
> 65% cpu utilization.
>
> If I move the client windows service to another server and start
> the service the RabbitMq on the original server never moves above
> 6% cpu utilization and the windows service never moves above
> .1% utilization at idle.
What kind of workload do you run? Is the consumer completely idle? (I doubt it is)
RabbitMQ's runtime on Windows does not support kernel polling (I/O completion ports in Windows
speak), so it's not a platform to run on if you are looking for efficiency.
My guess is that your consumer maxes out a single queue on a 2 core machine,
so both processes consume about 60% of the CPU capacity total.
--
MK
Staff Software Engineer, Pivotal/RabbitMQ