環(huán)境：
micrometer 1.8.2
prometheus 0.14.1
spring-boot-actuator 2.6.6

使用案例

<!-- Springboot啟動(dòng)actuator，默認(rèn)會引入依賴：micrometer-core -->
<dependency>
  <groupId>org.springframework.boot</groupId>
  <artifactId>spring-boot-starter-actuator</artifactId>
  <version>2.6.6</version>
</dependency>
<!-- micrometer橋接prometheus包，默認(rèn)會引入相關(guān)依賴：io.prometheus:simpleclient -->
<dependency>
  <groupId>io.micrometer</groupId>
  <artifactId>micrometer-registry-prometheus</artifactId>
  <version>1.8.2</version>
</dependency>

Timer

打點(diǎn)記錄任務(wù)的每次執(zhí)行時(shí)間。兜底默認(rèn)的時(shí)間窗口是1分鐘。如果想要修改可以配置：io.micrometer.core.instrument.distribution.DistributionStatisticConfig.Builder#expiry

Metrics.timer("my_name", "my_tag_1", "my_tag_2").record(() -> {
    doMyJob();
});

LongTaskTimer

與Timer類似，記錄任務(wù)執(zhí)行時(shí)間，官方注釋中也說了LongTask是一個(gè)主觀判斷，比如：1分鐘以上的任務(wù)
一個(gè)比較大區(qū)別在于多了一個(gè)接口方法：io.micrometer.core.instrument.LongTaskTimer#activeTasks
獲取當(dāng)前正在執(zhí)行中的任務(wù)數(shù)量???????

Metrics.more().longTaskTimer("my_name", "my_tag").record(doMyJob());

Gague

在服務(wù)器拉取指標(biāo)時(shí)，或者客戶端上報(bào)指標(biāo)時(shí)，調(diào)用提供的對象與方法獲取當(dāng)前指標(biāo)。即：記錄的是當(dāng)前狀態(tài)

RingBuffer<MatchingOutput> rb = disruptor.getRingBuffer();
Metrics.gauge("ringbuffer_remaining", Tags.of("my_tag_1", "my_tag_2"), rb, RingBuffer::remainingCapacity);

Counter

計(jì)數(shù)器打點(diǎn)

Metrics.counter("my_request", "my_tag_1", "my_tag_2").increment();

DistributionSummary

跟蹤事件的樣本分布。 一個(gè)例子是訪問 http 服務(wù)器的請求的響應(yīng)大小。???????

DistributionSummary ds =  DistributionSummary.builder("my.data.size")
    .tag("type", "my_type_1")
    .publishPercentileHistogram()
    .register(Metrics.globalRegistry);
ds.record(myValue);

配置actuator

配置指標(biāo)拉取端口，以及需要曝光的web接口???????

management:
  server:
    port: 9999
  endpoints:
    web:
      exposure:
        include: '*'
  metrics:
    tags:
      application: myAppName

Springboot整合啟動(dòng)流程

拉取指標(biāo)：http://localhost:9999/actuator/prometheus

servlet配置

接口自動(dòng)配置有很多入口，例如下面兩個(gè)

1. 普通web服務(wù)：org.springframework.boot.actuate.autoconfigure.endpoint.web.servlet.WebMvcEndpointManagementContextConfiguration#webEndpointServletHandlerMapping
2. 云服務(wù)商：org.springframework.boot.actuate.autoconfigure.cloudfoundry.servlet.CloudFoundryActuatorAutoConfiguration#cloudFoundryWebEndpointServletHandlerMapping

servlet邏輯

org.springframework.boot.actuate.metrics.export.prometheus.PrometheusScrapeEndpoint

@ReadOperation(producesFrom = TextOutputFormat.class)
public WebEndpointResponse<String> scrape(TextOutputFormat format, @Nullable Set<String> includedNames) {
    try {
        Writer writer = new StringWriter(this.nextMetricsScrapeSize);
        Enumeration<MetricFamilySamples> samples = (includedNames != null)
        ? this.collectorRegistry.filteredMetricFamilySamples(includedNames)
        : this.collectorRegistry.metricFamilySamples();
        format.write(writer, samples);
        String scrapePage = writer.toString();
        this.nextMetricsScrapeSize = scrapePage.length() + METRICS_SCRAPE_CHARS_EXTRA;
        return new WebEndpointResponse<>(scrapePage, format);
    }
    catch (IOException ex) {
        // This actually never happens since StringWriter doesn't throw an IOException
        throw new IllegalStateException("Writing metrics failed", ex);
    }
}

沒有配置過濾器，獲取枚舉對象
io.prometheus.client.CollectorRegistry#metricFamilySamples -》 io.prometheus.client.CollectorRegistry.MetricFamilySamplesEnumeration#MetricFamilySamplesEnumeration()

io.prometheus.client.CollectorRegistry.MetricFamilySamplesEnumeration

1. sampleNameFilter
2. collectorIter：對應(yīng)io.prometheus.client.CollectorRegistry#namesToCollectors屬性的value集合
3. 構(gòu)造器中查詢一次next：findNextElement

findNextElement

遍歷collectorIter迭代器一次，并收集一次指標(biāo)

1. io.prometheus.client.Collector#collect(io.prometheus.client.Predicate<java.lang.String>)
2. io.micrometer.prometheus.MicrometerCollector#collect
3. 遍歷io.micrometer.prometheus.MicrometerCollector#children集合中所有io.micrometer.prometheus.MicrometerCollector.Child對象
   1. 例如Gauge類型中的lambda匿名實(shí)現(xiàn)：io.micrometer.prometheus.PrometheusMeterRegistry#newGauge
4. 將遍歷的child中所有樣本，按照conventionName（例如：ringbuffer_remaining）分組，每個(gè)組對應(yīng)一個(gè)樣本家庭：io.prometheus.client.Collector.MetricFamilySamples
5. 返回List，將其迭代器賦給next屬性：io.prometheus.client.CollectorRegistry.MetricFamilySamplesEnumeration#next
6. 遍歷samples：io.prometheus.client.Collector.MetricFamilySamples#samples
7. 將sample（io.prometheus.client.Collector.MetricFamilySamples.Sample）數(shù)據(jù)寫入response響應(yīng)結(jié)果：org.springframework.boot.actuate.metrics.export.prometheus.TextOutputFormat#CONTENT_TYPE_004#write

接口輸出案例

公共配置的tag，所有的指標(biāo)都會帶有該tag：application=myAppName
指標(biāo)名稱：ringbuffer_remaining
指標(biāo)tag：type=my_tag_1
指標(biāo)類型：gauge

# HELP ringbuffer_remaining  
# TYPE ringbuffer_remaining gauge
ringbuffer_remaining{application="myAppName",type="my_tag_1",} 1024.0

采樣取數(shù)邏輯

Gauge

結(jié)合前面Gague使用案例的代碼
io.micrometer.core.instrument.internal.DefaultGauge

1. ref：對應(yīng)ringbuffer實(shí)例的弱引用
2. value：對應(yīng)RingBuffer::remainingCapacity方法

取樣邏輯即直接調(diào)用實(shí)例響應(yīng)方法返回的結(jié)果作為打點(diǎn)value

public class DefaultGauge<T> extends AbstractMeter implements Gauge {
    ...
    private final WeakReference<T> ref;
    private final ToDoubleFunction<T> value;
    ...
    @Override
    public double value() {
        T obj = ref.get();
        if (obj != null) {
            try {
                return value.applyAsDouble(obj);
            }
            catch (Throwable ex) {
                logger.log("Failed to apply the value function for the gauge '" + getId().getName() + "'.", ex);
            }
        }
        return Double.NaN;
    }
    ...
}

Timer

io.micrometer.prometheus.PrometheusTimer

1. count：LongAdder，遞增計(jì)數(shù)器
2. totalTime：LongAdder，任務(wù)耗時(shí)累加結(jié)果
3. max：io.micrometer.core.instrument.distribution.TimeWindowMax，簡化版的ringbuffer，用于記錄時(shí)間窗口中的最大值
4. histogramFlavor：直方圖風(fēng)味（類型），當(dāng)前版本只有兩種：Prometheus/VictoriaMetrics
5. histogram
   1. Prometheus類型：io.micrometer.core.instrument.distribution.TimeWindowFixedBoundaryHistogram#TimeWindowFixedBoundaryHistogram
   2. VictoriaMetrics類型：io.micrometer.core.instrument.distribution.FixedBoundaryVictoriaMetricsHistogram#FixedBoundaryVictoriaMetricsHistogram

取樣邏輯即監(jiān)控的方法實(shí)際調(diào)用時(shí)就會觸發(fā)打點(diǎn)記錄。取樣邏輯只是在接口拉取數(shù)據(jù)時(shí)調(diào)用實(shí)例實(shí)現(xiàn)的接口方法拍一個(gè)樣本快照

1. io.micrometer.core.instrument.distribution.HistogramSupport#takeSnapshot()
2. io.micrometer.prometheus.PrometheusTimer#takeSnapshot
3. io.micrometer.core.instrument.AbstractTimer#takeSnapshot
4. 如果histogram != null則追加histogramCounts數(shù)據(jù)

--io.micrometer.core.instrument.AbstractTimer#takeSnapshot
    @Override
    public HistogramSnapshot takeSnapshot() {
        return histogram.takeSnapshot(count(), totalTime(TimeUnit.NANOSECONDS), max(TimeUnit.NANOSECONDS));
    }
--io.micrometer.prometheus.PrometheusTimer#takeSnapshot
    @Override
    public HistogramSnapshot takeSnapshot() {
        HistogramSnapshot snapshot = super.takeSnapshot();
        if (histogram == null) {
            return snapshot;
        }
        return new HistogramSnapshot(snapshot.count(),
                snapshot.total(),
                snapshot.max(),
                snapshot.percentileValues(),
                histogramCounts(),
                snapshot::outputSummary);
    }

時(shí)間窗口

io.micrometer.core.instrument.distribution.TimeWindowMax

1. rotatingUpdater：AtomicIntegerFieldUpdater，rotating標(biāo)志符原子更新方法
2. clock：Clock，系統(tǒng)時(shí)鐘，返回當(dāng)前系統(tǒng)時(shí)間戳
3. durationBetweenRotatesMills：long，滾動(dòng)步進(jìn)大小
4. ringBuffer：AtomicLong[]，隊(duì)列
5. currentBucket：int，隊(duì)列當(dāng)前游標(biāo)
6. lastRotateTimestampMills：上一次rotate的時(shí)間戳
7. rotating：int，標(biāo)志符，0 - not rotating, 1 - rotating

每次寫入record，或者查詢poll，都會提前校驗(yàn)下是否需要翻轉(zhuǎn)，調(diào)用rotate方法
io.micrometer.core.instrument.distribution.TimeWindowMax#rotate

1. wallTime=系統(tǒng)當(dāng)前時(shí)間
2. timeSinceLastRotateMillis = wallTime - lastRotateTimestampMillis，即：當(dāng)前時(shí)間距離上次翻轉(zhuǎn)的時(shí)間間隔
3. 如果低于步進(jìn)，直接返回不需要翻轉(zhuǎn)：timeSinceLastRotateMillis < durationBetweenRotatesMillis
4. 否則更新標(biāo)志符，表示當(dāng)前正在翻轉(zhuǎn)，需要阻塞等待下
5. 如果timeSinceLastRotateMillis已經(jīng)超出整個(gè)隊(duì)列的長度了：timeSinceLastRotateMillis >= durationBetweenRotatesMillis * ringBuffer.length
   1. 那么直接重置隊(duì)列返回即可
   2. 遍歷ringBuffer所有位置設(shè)置為0
   3. currentBucket更新為0
   4. 更新上次翻轉(zhuǎn)時(shí)間：lastRotateTimestampMillis = wallTime - timeSinceLastRotateMillis % durationBetweenRotatesMillis
6. 否則，將當(dāng)前時(shí)間與上次翻轉(zhuǎn)時(shí)間之間已經(jīng)超時(shí)的bucket重置為0

int iterations = 0;
do {
    ringBuffer[currentBucket].set(0);
    if (++currentBucket >= ringBuffer.length) {
        currentBucket = 0;
    }
    timeSinceLastRotateMillis -= durationBetweenRotatesMillis;
    lastRotateTimestampMillis += durationBetweenRotatesMillis;
} while (timeSinceLastRotateMillis >= durationBetweenRotatesMillis && ++iterations < ringBuffer.length);

例如：當(dāng)前時(shí)間為4，上次翻轉(zhuǎn)時(shí)間為2，隊(duì)列大小為3，durationBetweenRotatesMillis=1，currentBucket=1，那么timeSinceLastRotateMillis=4-2=2
循環(huán)第1輪

1. 更新ringBuffer[1]=0
2. 更新currentBucket=2
3. 更新timeSinceLastRotateMillis=2-1
4. 更新lastRotateTimestampMillis=2+1
5. 更新iterations=1

循環(huán)第2輪

1. 更新ringBuffer[2]=0
2. 更新currentBucket=3
   1. currentBucket>=隊(duì)列長度
   2. 重置currentBucket=0
3. 更新timeSinceLastRotateMillis=1-1
4. 更新lastRotateTimestampMillis=3+1
5. 更新iterations=2，此時(shí)timeSinceLastRotateMillis=0，小于durationBetweenRotatesMillis，結(jié)束循環(huán)

一次旋轉(zhuǎn)圖例

當(dāng)發(fā)現(xiàn)上次旋轉(zhuǎn)時(shí)間（lastRotateTimestampMillis）已經(jīng)落后當(dāng)前時(shí)間（wallTime）4個(gè)單位后，lastRotateTimestampMillis向右移動(dòng)4個(gè)時(shí)間單位，currentBucket也向右移動(dòng)4個(gè)單位。但是因?yàn)閏urrentBucket是數(shù)組的index，當(dāng)越界的時(shí)候就移動(dòng)到0繼續(xù)（一個(gè)環(huán)）。例如下圖：
currentBucket向右移動(dòng)4個(gè)單位，隊(duì)列長度為3，當(dāng)前index=0，那么移動(dòng)后index=2（轉(zhuǎn)了一圈）

總結(jié)

Micrometer可以整合Prometheus也可以整合influxDB等時(shí)序數(shù)據(jù)庫，主要作用就是橋接，類似于Slf4j與log4j，logback的關(guān)系。提供一個(gè)通用的打點(diǎn)能力，并將打點(diǎn)數(shù)據(jù)對接到相應(yīng)的時(shí)序數(shù)據(jù)庫，用戶只需要關(guān)心何時(shí)打點(diǎn)即可。例如：

1. 橋接包中的io.micrometer.prometheus.PrometheusMeterRegistry，將打點(diǎn)數(shù)據(jù)橋接至io.prometheus.client.CollectorRegistry
2. 橋接包中的io.micrometer.influx.InfluxMeterRegistry，將打點(diǎn)數(shù)據(jù)按照influx協(xié)議橋接push至influxDB。
   1. 默認(rèn)push頻率為1分鐘一次，可以按需配置：io.micrometer.core.instrument.push.PushRegistryConfig#step
   2. 線程池默認(rèn)為單線程：java.util.concurrent.Executors#newSingleThreadScheduledExecutor(java.util.concurrent.ThreadFactory)
   3. 線程池線程命名規(guī)則針對influxDB實(shí)現(xiàn)：influx-metrics-publisher

actuator就像是啟動(dòng)器，會將對接具體時(shí)序數(shù)據(jù)庫所需要的配置自動(dòng)化，例如指標(biāo)矩陣相關(guān)的：Prometheus曝光web接口的相關(guān)配置，influx相關(guān)配置，micrometer metrics等等相關(guān)配置

1. org.springframework.boot.actuate.autoconfigure.metrics.JvmMetricsAutoConfiguration
2. org.springframework.boot.actuate.autoconfigure.metrics.KafkaMetricsAutoConfiguration
3. org.springframework.boot.actuate.autoconfigure.metrics.Log4J2MetricsAutoConfiguration
4. org.springframework.boot.actuate.autoconfigure.metrics.LogbackMetricsAutoConfiguration
5. org.springframework.boot.actuate.autoconfigure.metrics.SystemMetricsAutoConfiguration

最終可以通過Grafana等報(bào)表工具對接打點(diǎn)數(shù)據(jù)源展示圖表。常見的架構(gòu)有：Micrometer-》Prometheus-〉Grafana
注意：前端頁面渲染存在瓶頸，例如一個(gè)指標(biāo)的tag如果太多會導(dǎo)致報(bào)表非常的卡頓，一般5k個(gè)tag就會有感知，1W+會明顯影響使用

到此這篇關(guān)于spring-boot整合Micrometer+Prometheus的文章就介紹到這了,更多相關(guān)spring-boot整合Micrometer+Prometheus內(nèi)容請搜索腳本之家以前的文章或繼續(xù)瀏覽下面的相關(guān)文章希望大家以后多多支持腳本之家！

最新評論