[rabbitmq-discuss] Query about performance
mgg at giagnocavo.net
Mon Jan 20 20:28:07 GMT 2014
>> i) Removed basicQos() call from the consumer code, so that the full
> capacity is used
>Consider using a larger prefetch count instead of unlimited.
I'll just an an anecdote here: Shovel does unlimited prefetch by default. This lead to our RabbitMQ instances grinding to a halt under any sort of load due to exhausting memory. A 2K msg prefetch solved everything with no perf loss.
More information about the rabbitmq-discuss