rabbitmq之Fair dispatch

正常情況下，rabbitmq server（broker）會盲目地，按照message數目平均地分發message給consumer，而不考慮有的message處理起來比較耗時，有的處理起來很快，

場景問題是，假設有兩個consumers，consumerA和consumerB，4條messages，evenly， message1，message3會分發給consumerA去處理，message2，message4會分發給consumerB去處理，有的時候，message1，message3處理起來很簡單，耗時很少，幾秒鐘就處理完畢，message2，message4處理比較耗時，每條message都得處理幾分鐘，則由於rabbitmq默認只會按照數目平均的分發給consumers，這樣會導致consumerA 幾秒鐘就處理完message1和3，休息，而consumerB還在處理message2中，message4還沒開始處理，導致不公平。通常情況下，rabbitmq server只會just blindly dispatches every n-th messageto the n-th consumer.

解決方法：通過channel.basicQos(1);來保證RabbitMQ not to give more thanone message to a worker at a time. Or, in other words, don't dispatcha new message to a worker until it has processed and acknowledged theprevious one. Instead, it will dispatch it to the next worker that is not still busy. 如果不設置這個參數，則rabbitmq會數目均勻的一次性平均分發給各個consumers

int prefetchCount = 1;
channel.basicQos(prefetchCount);

由於queue的size有限，當queue裏面的messages滿了，你需要增加consumers個數來處理message或者其他的strategy

Using message acknowledgments and prefetchCount you can set up awork queue. The durability options let the tasks survive even ifRabbitMQ is restarted.

reference: http://www.rabbitmq.com/tutorials/tutorial-two-java.html