延时监控

TotalTimeMs=queue + local + remote（线上环境这个占比99%以上） + response send time

produce: kafka.network:type=RequestMetrics,name=TotalTimeMs,request=Produce
consumer: kafka.network:type=RequestMetrics,name=TotalTimeMs,request=FetchConsumer
follower:kafka.network:type=RequestMetrics,name=TotalTimeMs,request=FetchFollower

以consumer为例：一个请求FetchFollower耗费的所有时间

ObjectName：kafka.network:type=RequestMetrics,name=TotalTimeMs,request=FetchConsumer
ClassName：com.yammer.metrics.reporting.JmxReporter$Histogram
说明：Information on the management interface of the MBean

属性值：

名称	值
50thPercentile	1.0
75thPercentile	1.0
95thPercentile	3.0
98thPercentile	6.0
999thPercentile	83.826
99thPercentile	10.0
Count	4820
Max	441.0
Mean-平均	1.7898340248962656
Min	0.0
StdDev-标准差	8.634806791760138

Demo：

JMXConnector conn = null;
        MBeanServerConnection mbsc = null;
        ObjectName objectName = null;
        double mean = 0;
        try {
            conn = KafkaJmxUtil.getJmxConnection("localhost", 9999);
            mbsc = conn.getMBeanServerConnection();
            objectName = new ObjectName("kafka.network:type=RequestMetrics,name=TotalTimeMs,request=FetchConsumer");
            mean = (Double) mbsc.getAttribute(objectName, "Mean");
        } catch (Exception e) {
            e.printStackTrace();
        }
        System.out.println("Mean：" + mean);

Mean：1.7898340248962656