1.簡介
系統(tǒng)越來越大，我們所需要關(guān)聯(lián)的組件也越來越多遥诉，為了更好的排查問題，我們需要對不同的指標(biāo)進(jìn)行監(jiān)控噪叙，從而發(fā)現(xiàn)問題矮锈，Metrics作為一款監(jiān)控指標(biāo)的度量類庫，提供了許多工具幫助開發(fā)者來完成自定義的監(jiān)控工作睁蕾。

2.環(huán)境
添加maven依賴

<dependency>
    <groupId>io.dropwizard.metrics</groupId>
    <artifactId>metrics-core</artifactId>
    <version>${metrics.version}</version>
</dependency>

3.使用
Metrics中MetricRegistry是中心容器苞笨，它是程序中所有度量的容器，所有新的度量工具都要注冊到一個MetricRegistry實例中才可以使用子眶，盡量在一個應(yīng)用中保持讓這個MetricRegistry實例保持單例瀑凝。

Metrics提供了多個指標(biāo) Gauges、Counters臭杰、Histograms粤咪、Meters、Timers

3.1Timers
Timer用于統(tǒng)計QPS和耗時硅卢，這里用于統(tǒng)記DB查詢所耗時間射窒，由于try-catch-resources屬性自動stop了，也可以顯示調(diào)用Context stop方法

@RestController
@RequestMapping("/metric")
public class MetricCtrl {

    private static final MetricRegistry metricRegistry = new MetricRegistry();

    private Timer timer;

    private static ConsoleReporter reporter = ConsoleReporter.forRegistry(metricRegistry).build();//使用控制臺打印

    @Autowired
    private DailyTradeMarkingDao dailyTradeMarkingDao;//獲取用戶天維度的標(biāo)記

    @PostConstruct
    public void setup() {
        timer = metricRegistry.timer("dbQuery");
        reporter.start(10, TimeUnit.SECONDS);//報告10秒輸出一次
    }

    @PostMapping("/timer")
    public JsonResult testTimer() {
        List<DailyTradeMarking> markingList = null;
        try (Timer.Context context = timer.time()) {
            markingList = dailyTradeMarkingDao.getByAccountIdAndStockId(401422L, 14L);
        }
        return JsonResult.ok(markingList);
    }
}

顯示結(jié)果：
-- Timers ----------------------------------------------------------------------
dbQuery
             count = 7
         mean rate = 0.12 calls/second
     1-minute rate = 0.08 calls/second
     5-minute rate = 0.02 calls/second
    15-minute rate = 0.01 calls/second
               min = 42.75 milliseconds
               max = 1228.59 milliseconds
              mean = 211.34 milliseconds
            stddev = 401.27 milliseconds
            median = 46.35 milliseconds
              75% <= 87.25 milliseconds
              95% <= 1228.59 milliseconds
              98% <= 1228.59 milliseconds
              99% <= 1228.59 milliseconds
            99.9% <= 1228.59 milliseconds

3.2Meters
Meter用于測試QPS 将塑，這里統(tǒng)計/meter接口的QPS是多少

{
    meter = metricRegistry.meter("qps");
    @PostMapping("/meter")
    public JsonResult testMeter() {
        meter.mark();//根據(jù)不同的需求設(shè)置不同的n ,默認(rèn)1
        return JsonResult.ok();
    }
}


顯示結(jié)果：


-- Meters ----------------------------------------------------------------------
qps
             count = 6
         mean rate = 0.02 events/second
     1-minute rate = 0.09 events/second
     5-minute rate = 0.02 events/second
    15-minute rate = 0.01 events/second

3.3Histograms
Histogram 是一個統(tǒng)計圖表，可以統(tǒng)計次數(shù)蝌麸、最大点寥、最小、平均值等, 這里記錄reques長度的大小

{
    
    histogram = metricRegistry.histogram("request.size");
    @PostMapping("/histogram")
    public JsonResult testHistogram(HttpServletRequest request) {
        histogram.update(request.toString().length());
        return JsonResult.ok();
    }
}
顯示結(jié)果：
-- Histograms ------------------------------------------------------------------
request.size
             count = 8
               min = 52
               max = 52
              mean = 52.00
            stddev = 0.00
            median = 52.00
              75% <= 52.00
              95% <= 52.00
              98% <= 52.00
              99% <= 52.00
            99.9% <= 52.00

3.4 Counters
counter用來統(tǒng)計次數(shù)来吩，這里用來統(tǒng)計大于5的個數(shù)有多少個

{
    
    counter = metricRegistry.counter("count");
    counterFive = metricRegistry.counter("counterFive");

    @PostMapping("/counter")
    public JsonResult testCounter() {
        Random random = new Random();
        counter.inc();
        if (random.nextInt(10) > 5) {
            counterFive.inc();
        }
        return JsonResult.ok();
    }

}


顯示結(jié)果：
-- Counters --------------------------------------------------------------------
counterFive
             count = 1
count
             count = 7

3.5Gauges
gauge可以實現(xiàn)自己的度量標(biāo)準(zhǔn)敢辩，這里用來檢測緩存的大小

{
    private LoadingCache<Integer, Integer> cache = CacheBuilder.newBuilder()
        .maximumSize(1000)
        .refreshAfterWrite(10, TimeUnit.SECONDS).build(
        new CacheLoader<Integer, Integer>() {
            @Override
            public Integer load(Integer key) throws Exception {
                return 1;
            }
        }
    );
    gauge = metricRegistry.gauge("cache.size", new MetricRegistry.MetricSupplier<Gauge>() {
        @Override
        public Gauge newMetric() {
            return new Gauge<Long>() {
                @Override
                public Long getValue() {
                    return cache.size();
                }
            };
        }
    });
    
    @PostMapping("/gauge")  
    public JsonResult testGauge() throws ExecutionException {
        Random random = new Random();
        cache.get(random.nextInt(100));
        return JsonResult.ok(gauge.getValue());
    }
}




顯示結(jié)果：
-- Gauges ----------------------------------------------------------------------
cache.size
             value = 10
-- Gauges ----------------------------------------------------------------------
cache.size
             value = 11

4.數(shù)據(jù)上報（顯示）

ConsoleReporter reporter = ConsoleReporter.forRegistry(metricRegistry).build();//輸出在控制臺

也可以繼承ScheduledReporter實現(xiàn)自己的聚合格式