Micrometer 自定义 metrics-技术圈

点击上方“程序员大白”，选择“星标”公众号

重磅干货，第一时间送达

Micrometer 提供了基于 Java的 monitor facade，其与 springboot 应用和 prometheus 的集成方式如下图展示

上图中展示得很清楚，应用通过Micrometer采集和暴露监控端点给prometheus，prometheus通过pull模式来采集监控时序数据信息。之后作为数据源提供给grafana进行展示。

Micrometer支持的度量方式及在springboot中的应用示例

Counter
Counter(计数器)简单理解就是一种只增不减的计数器。它通常用于记录服务的请求数量、完成的任务数量、错误的发生数量等等。

package com.olive.producter.monitor;
import org.springframework.stereotype.Service;import io.micrometer.core.instrument.Counter;import io.micrometer.core.instrument.Metrics;
@Service("collectorService")public class CollectorService {
static final Counter userCounter = Metrics.counter("user.counter.total", "services", "demo");
public void processCollectResult() throws InterruptedException {
while (true) {            userCounter.increment(1D);        }    }}

Gauge
Gauge(仪表)是一个表示单个数值的度量，它可以表示任意地上下移动的数值测量。Gauge通常用于变动的测量值，如当前的内存使用情况，同时也可以测量上下移动的"计数"，比如队列中的消息数量。

package com.olive.producter.monitor;
import java.util.ArrayList;import java.util.List;import java.util.concurrent.atomic.AtomicInteger;import org.springframework.stereotype.Component;import io.micrometer.core.instrument.Gauge;import io.micrometer.core.instrument.ImmutableTag;import io.micrometer.core.instrument.Metrics;import io.micrometer.core.instrument.Tag;import io.micrometer.core.instrument.simple.SimpleMeterRegistry;
@Component("passCaseMetric")public class PassCaseMetric {
List init() {ArrayList list = new ArrayList() {        };        list.add(new ImmutableTag("service", "demo"));return list;    }
AtomicInteger atomicInteger = new AtomicInteger(0);
Gauge passCaseGuage = Gauge.builder("pass.cases.guage", atomicInteger, AtomicInteger::get).tag("service", "demo")            .description("pass cases guage of demo").register(new SimpleMeterRegistry());
AtomicInteger passCases = Metrics.gauge("pass.cases.guage.value", init(), atomicInteger);
public void handleMetrics() {
while (true) {if (System.currentTimeMillis() % 2 == 0) {                passCases.addAndGet(100);System.out.println("ADD + " + passCaseGuage.measure() + " : " + passCases);            } else {                int val = passCases.addAndGet(-100);if (val < 0) {                    passCases.set(1);                }System.out.println("DECR - " + passCaseGuage.measure() + " : " + passCases);            }        }
    }
}

增加一个controller，触发他们：

package com.olive.producter.web;
import org.springframework.beans.factory.annotation.Autowired;import org.springframework.web.bind.annotation.RequestMapping;import org.springframework.web.bind.annotation.RequestMethod;import org.springframework.web.bind.annotation.RestController;import com.olive.producter.monitor.CollectorService;import com.olive.producter.monitor.PassCaseMetric;
@RestController@RequestMapping("/monitor")public class MonitorController {
    @Autowired    CollectorService collectorService;
    @Autowired    PassCaseMetric passCaseMetric;
    @RequestMapping(value = "/counter", method = RequestMethod.GET)    public String counter() throws InterruptedException {collectorService.processCollectResult();return "+1";    }
    @RequestMapping(value = "/gauge", method = RequestMethod.GET)    public String gauge() throws InterruptedException {passCaseMetric.handleMetrics();return "+gauge";    }
}

启动springboot应用，可以在http://host:port/actuator/prometheus 看到端点收集到的数据。其他的也是类似的不再一一截图展示。

这里使用了一个true的循环用来展示不断更新的效果。

同样的可以在grafana中看到监控展示信息

Timer
Timer(计时器)同时测量一个特定的代码逻辑块的调用(执行)速度和它的时间分布。简单来说，就是在调用结束的时间点记录整个调用块执行的总时间，适用于测量短时间执行的事件的耗时分布，例如消息队列消息的消费速率。

@Testpublic void testTimerSample()        Timer timer = Timer.builder("timer")                .tag("timer", "timersample")                .description("timer sample test.")                .register(new SimpleMeterRegistry());
for(int i=0; i<2; i++) {            timer.record(() -> {try {                    TimeUnit.SECONDS.sleep(2);                }catch (InterruptedException e){
                }
            });        }
        System.out.println(timer.count());        System.out.println(timer.measure());        System.out.println(timer.totalTime(TimeUnit.SECONDS));        System.out.println(timer.mean(TimeUnit.SECONDS));        System.out.println(timer.max(TimeUnit.SECONDS));}

响应数据

2[Measurement{statistic='COUNT', value=2.0}, Measurement{statistic='TOTAL_TIME', value=4.005095763}, Measurement{statistic='MAX', value=2.004500494}]4.0050957632.00254788152.004500494

Summary
Summary(摘要)用于跟踪事件的分布。它类似于一个计时器，但更一般的情况是，它的大小并不一定是一段时间的测量值。在Micrometer中，对应的类是DistributionSummary，它的用法有点像Timer，但是记录的值是需要直接指定，而不是通过测量一个任务的执行时间。

@Testpublic void testSummary()
        DistributionSummary summary = DistributionSummary.builder("summary")                .tag("summary", "summarySample")                .description("summary sample test")                .register(new SimpleMeterRegistry());
        summary.record(2D);        summary.record(3D);        summary.record(4D);
        System.out.println(summary.count());        System.out.println(summary.measure());        System.out.println(summary.max());        System.out.println(summary.mean());        System.out.println(summary.totalAmount());}

响应数据：

3[Measurement{statistic='COUNT', value=3.0}, Measurement{statistic='TOTAL', value=9.0}, Measurement{statistic='MAX', value=4.0}]4.03.09.0

序

本文主要研究下如何使用自定义Micrometer的metrics

实例

DemoMetrics

public class DemoMetrics implements MeterBinder {    AtomicInteger count = new AtomicInteger(0);
@Overridepublic void bindTo(MeterRegistry meterRegistry) {        Gauge.builder("demo.count", count, c -> c.incrementAndGet())                .tags("host", "localhost")                .description("demo of custom meter binder")                .register(meterRegistry);    }}

这里实现了MeterBinder接口的bindTo方法，将要采集的指标注册到MeterRegistry

注册

原始方式

new DemoMetrics().bindTo(registry);

springboot autoconfigure

@Beanpublic DemoMetrics demoMetrics(){return new DemoMetrics();}

在springboot只要标注下bean，注入到spring容器后，springboot会自动注册到registry。springboot已经帮你初始化了包括UptimeMetrics等一系列metrics。详见源码解析部分。

验证

curl -i http://localhost:8080/actuator/metrics/demo.count

返回实例

{"name": "demo.count","measurements": [    {"statistic": "VALUE","value": 6    }  ],"availableTags": [    {"tag": "host","values": ["localhost"      ]    }  ]}

源码解析

MetricsAutoConfiguration

spring-boot-actuator-autoconfigure-2.0.0.RELEASE-sources.jar!/org/springframework/boot/actuate/autoconfigure/metrics/MetricsAutoConfiguration.java

@Configuration@ConditionalOnClass(Timed.class)@EnableConfigurationProperties(MetricsProperties.class)@AutoConfigureBefore(CompositeMeterRegistryAutoConfiguration.class)public class MetricsAutoConfiguration {
@Bean@ConditionalOnMissingBeanpublic Clock micrometerClock() {return Clock.SYSTEM;    }
@Beanpublic static MeterRegistryPostProcessor meterRegistryPostProcessor(            ApplicationContext context) {return new MeterRegistryPostProcessor(context);    }
@Bean@Order(0)public PropertiesMeterFilter propertiesMeterFilter(MetricsProperties properties) {return new PropertiesMeterFilter(properties);    }
@Configuration@ConditionalOnProperty(value = "management.metrics.binders.jvm.enabled", matchIfMissing = true)static class JvmMeterBindersConfiguration {
@Bean@ConditionalOnMissingBeanpublic JvmGcMetrics jvmGcMetrics() {return new JvmGcMetrics();        }
@Bean@ConditionalOnMissingBeanpublic JvmMemoryMetrics jvmMemoryMetrics() {return new JvmMemoryMetrics();        }
@Bean@ConditionalOnMissingBeanpublic JvmThreadMetrics jvmThreadMetrics() {return new JvmThreadMetrics();        }
@Bean@ConditionalOnMissingBeanpublic ClassLoaderMetrics classLoaderMetrics() {return new ClassLoaderMetrics();        }
    }
@Configurationstatic class MeterBindersConfiguration {
@Bean@ConditionalOnClass(name = { "ch.qos.logback.classic.LoggerContext","org.slf4j.LoggerFactory" })@Conditional(LogbackLoggingCondition.class)@ConditionalOnMissingBean(LogbackMetrics.class)@ConditionalOnProperty(value = "management.metrics.binders.logback.enabled", matchIfMissing = true)public LogbackMetrics logbackMetrics() {return new LogbackMetrics();        }
@Bean@ConditionalOnProperty(value = "management.metrics.binders.uptime.enabled", matchIfMissing = true)@ConditionalOnMissingBeanpublic UptimeMetrics uptimeMetrics() {return new UptimeMetrics();        }
@Bean@ConditionalOnProperty(value = "management.metrics.binders.processor.enabled", matchIfMissing = true)@ConditionalOnMissingBeanpublic ProcessorMetrics processorMetrics() {return new ProcessorMetrics();        }
@Bean@ConditionalOnProperty(name = "management.metrics.binders.files.enabled", matchIfMissing = true)@ConditionalOnMissingBeanpublic FileDescriptorMetrics fileDescriptorMetrics() {return new FileDescriptorMetrics();        }
    }
static class LogbackLoggingCondition extends SpringBootCondition {
@Overridepublic ConditionOutcome getMatchOutcome(ConditionContext context,                AnnotatedTypeMetadata metadata) {            ILoggerFactory loggerFactory = LoggerFactory.getILoggerFactory();            ConditionMessage.Builder message = ConditionMessage                    .forCondition("LogbackLoggingCondition");if (loggerFactory instanceof LoggerContext) {return ConditionOutcome.match(                        message.because("ILoggerFactory is a Logback LoggerContext"));            }return ConditionOutcome                    .noMatch(message.because("ILoggerFactory is an instance of "                            + loggerFactory.getClass().getCanonicalName()));        }
    }
}

可以看到这里注册了好多metrics，比如UptimeMetrics，JvmGcMetrics，ProcessorMetrics，FileDescriptorMetrics等
这里重点看使用@Bean标注了MeterRegistryPostProcessor

MeterRegistryPostProcessor

spring-boot-actuator-autoconfigure-2.0.0.RELEASE-sources.jar!/org/springframework/boot/actuate/autoconfigure/metrics/MeterRegistryPostProcessor.java

class MeterRegistryPostProcessor implements BeanPostProcessor {
private final ApplicationContext context;
private volatile MeterRegistryConfigurer configurer;
    MeterRegistryPostProcessor(ApplicationContext context) {this.context = context;    }
@Overridepublic Object postProcessAfterInitialization(Object bean, String beanName)            throws BeansException {if (bean instanceof MeterRegistry) {            getConfigurer().configure((MeterRegistry) bean);        }return bean;    }
@SuppressWarnings("unchecked")private MeterRegistryConfigurer getConfigurer() {if (this.configurer == null) {this.configurer = new MeterRegistryConfigurer(beansOfType(MeterBinder.class),                    beansOfType(MeterFilter.class),                    (Collection>) (Object) beansOfType(                            MeterRegistryCustomizer.class),this.context.getBean(MetricsProperties.class).isUseGlobalRegistry());        }return this.configurer;    }
private  Collection beansOfType(Class type) {return this.context.getBeansOfType(type).values();    }
}

可以看到这里new了一个MeterRegistryConfigurer，重点注意这里使用beansOfType(MeterBinder.class)方法的返回值给其构造器

MeterRegistryConfigurer

spring-boot-actuator-autoconfigure-2.0.0.RELEASE-sources.jar!/org/springframework/boot/actuate/autoconfigure/metrics/MeterRegistryConfigurer.java

class MeterRegistryConfigurer {
private final Collection> customizers;
private final Collection filters;
private final Collection binders;
private final boolean addToGlobalRegistry;
    MeterRegistryConfigurer(Collection binders,            Collection filters,            Collection> customizers,            boolean addToGlobalRegistry) {this.binders = (binders != null ? binders : Collections.emptyList());this.filters = (filters != null ? filters : Collections.emptyList());this.customizers = (customizers != null ? customizers : Collections.emptyList());this.addToGlobalRegistry = addToGlobalRegistry;    }
    void configure(MeterRegistry registry) {if (registry instanceof CompositeMeterRegistry) {return;        }// Customizers must be applied before binders, as they may add custom// tags or alter timer or summary configuration.        customize(registry);        addFilters(registry);        addBinders(registry);if (this.addToGlobalRegistry && registry != Metrics.globalRegistry) {            Metrics.addRegistry(registry);        }    }
@SuppressWarnings("unchecked")private void customize(MeterRegistry registry) {        LambdaSafe.callbacks(MeterRegistryCustomizer.class, this.customizers, registry)                .withLogger(MeterRegistryConfigurer.class)                .invoke((customizer) -> customizer.customize(registry));    }
private void addFilters(MeterRegistry registry) {this.filters.forEach(registry.config()::meterFilter);    }
private void addBinders(MeterRegistry registry) {this.binders.forEach((binder) -> binder.bindTo(registry));    }
}

可以看到configure方法里头调用了addBinders，也就是把托管给spring容器的MeterBinder实例bindTo到meterRegistry

小结

springboot2 引入的 Micrometer，自定义metrics只需要实现MeterBinder接口，然后托管给spring即可，springboot的autoconfigure帮你自动注册到meterRegistry。

source: //www.cnblogs.com/duanxz/p/10179946.html

推荐阅读

13个你一定要知道的PyTorch特性

解读：为什么要做特征归一化/标准化？

一文搞懂 PyTorch 内部机制

张一鸣：每个逆袭的年轻人，都具备的底层能力

关于程序员大白

程序员大白是一群哈工大，东北大学，西湖大学和上海交通大学的硕士博士运营维护的号，大家乐于分享高质量文章，喜欢总结知识，欢迎关注[程序员大白]，大家一起学习进步！