Spring Boot默認(rèn)的指標(biāo)數(shù)據(jù)從哪來(lái)的?
您是否注意到 Spring Boot 和 Micrometer 為您的應(yīng)用生成的所有默認(rèn)指標(biāo)?如果沒(méi)有 - 您可以將 actuator 依賴(lài)項(xiàng)添加到項(xiàng)目中,然后點(diǎn)擊 / actuator / metrics 端點(diǎn),在那里您將找到有關(guān) JVM 、進(jìn)程、Tomcat、流量等的有用信息。
然后,添加一些緩存,數(shù)據(jù)源 或 JPA 依賴(lài)項(xiàng),甚至?xí)霈F(xiàn)更多指標(biāo)。如果您想知道它們是如何結(jié)束的,我們可以在哪里找到關(guān)于它們所描述的參數(shù)的解釋?zhuān)敲催@篇文章就是為您準(zhǔn)備的。
顯示指標(biāo)
為了讓它井然有序,讓我們從如何在 Spring Boot 應(yīng)用程序中顯示指標(biāo)開(kāi)始。如果您已經(jīng)知道了,可以跳過(guò)這一部分。
Spring Boot中的指標(biāo)由 micrometer.io 處理。但是,如果您使用 actuator ,則不需要向項(xiàng)目添加 micrometer 依賴(lài)項(xiàng),因?yàn)?actuator 已經(jīng)依賴(lài)于它。即使您對(duì)它提供的端點(diǎn)不感興趣,也希望您使用 actuator ,因?yàn)檫@是通過(guò)其 AutoConfigurations 注冊(cè)許多指標(biāo)的模塊。稍后我們會(huì)詳細(xì)討論。
因此,首先,只需將執(zhí)行器依賴(lài)項(xiàng)添加到項(xiàng)目中(這里是 build.gradle.kts )
- dependencies {
- implementation("org.springframework.boot:spring-boot-starter-actuator")
- }
并在執(zhí)行器端點(diǎn)中顯示指標(biāo)名稱(chēng),點(diǎn)擊 http://localhost:8080/actuator/metrics.
- {
- "names": [
- "jvm.threads.states",
- "process.files.max",
- "jvm.memory.used",
- "jvm.gc.memory.promoted",
- "jvm.memory.max",
- "system.load.average.1m",
- ...
- ]
- }
然后,要查看詳細(xì)信息,請(qǐng)?jiān)?URL 路徑中添加指標(biāo)名稱(chēng),例如:http://localhost:8080/actuator/metrics/system.cpu.count.
- {
- "name": "system.cpu.count",
- "description": "The number of processors available to the Java virtual machine",
- "baseUnit": null,
- "measurements": [
- {
- "statistic": "VALUE",
- "value": 8
- }
- ],
- "availableTags": [
- ]
- }
通過(guò)提供特定的儀表注冊(cè)表,可以定期將這些指標(biāo)發(fā)送到您選擇的指標(biāo)系統(tǒng)( Prometheus,New Relic,CloudWatch,Graphite 等)。讓我們用最簡(jiǎn)單的注冊(cè)表來(lái)做 - LoggingMeterRegistry,它只是定期記錄所有指標(biāo)。
- @Configuration
- class MetricsConfig {
- @Bean
- LoggingMeterRegistry loggingMeterRegistry() {
- return new LoggingMeterRegistry();
- }
- }
現(xiàn)在,指標(biāo)也顯示在日志中:
- 2019-07-17 11:07:09.406 INFO 91283 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : jvm.buffer.count{id=direct} value=0 buffers
- 2019-07-17 11:07:09.406 INFO 91283 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : jvm.buffer.count{id=mapped} value=0 buffers
- 2019-07-17 11:07:09.406 INFO 91283 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : jvm.buffer.memory.used{id=direct} value=0 B
- 2019-07-17 11:07:09.406 INFO 91283 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : jvm.buffer.memory.used{id=mapped} value=0 B
- 2019-07-17 11:07:09.408 INFO 91283 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : jvm.classes.loaded{} value=8530 classes
- 2019-07-17 11:07:09.408 INFO 91283 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : jvm.gc.live.data.size{} value=0 B
- 2019-07-17 11:07:09.408 INFO 91283 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : jvm.gc.max.data.size{} value=0 B
- 2019-07-17 11:07:09.410 INFO 91283 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : jvm.memory.committed{area=nonheap,id=Compressed Class Space} value=6.25 MiB
- 2019-07-17 11:07:09.410 INFO 91283 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : jvm.memory.committed{area=heap,id=G1 Eden Space} value=168 MiB
- ...
指標(biāo)供應(yīng)
那么,如何提供這些指標(biāo)呢?一個(gè)示例可能是 WebMvcMetricsFilter ,向所有 Spring Web MVC 端點(diǎn)添加性能指標(biāo) (http.server.requests metric) 。
但是這個(gè)例子很簡(jiǎn)單。當(dāng)所有請(qǐng)求都由 Spring 框架處理時(shí),在內(nèi)部添加調(diào)用生成指標(biāo)是沒(méi)有必要的(只檢查 WebMvcMetricsFilter.record() 方法)。
但是,如果您使用純 ehcache 或 hibernate 或其他數(shù)據(jù)源,然后生成指標(biāo),情況又會(huì)如何呢?
那么 cache. * 指標(biāo)呢,即使我 @Autowired 純 net.sf.ehcache.Cache 也會(huì)生成?
那么 hibernate. * 指標(biāo)呢,即使我 @Autowired 純 org.hibernate.SessionFactory 也會(huì)生成?
然后, jvm.* , process.* , tomcat.* 等如何自動(dòng)生成?
它似乎比人們想象的更簡(jiǎn)單,因?yàn)檫@些統(tǒng)計(jì)數(shù)據(jù)是由受監(jiān)控的組件本身提供的。有時(shí),它將直接提供,例如cache.getStatistics() 為 EhCache 提供 StatisticsGateway *,*或 sessionFactory.getStatistics() 為 Hibernate SessionFactory 提供 statistics 等等。
有時(shí),這可以通過(guò)其他方式實(shí)現(xiàn),比如托管 bean 。例如,將 RuntimeMXBean 用于 JVM process.* 指標(biāo)以及 將(如GlobalRequestProcessor, Servlet 等) Tomcat mbeans 用于 tomcat. * 指標(biāo)
為了訪問(wèn)這些統(tǒng)計(jì)數(shù)據(jù)并將其轉(zhuǎn)換為特定指標(biāo),Micrometer 引入了 MeterBinder 的概念。
檢查 MeterBinder implementation 層次結(jié)構(gòu),您將了解更多關(guān)于可用的指標(biāo)組的信息。
Micrometer MeterBinders
您也可以直接在 micrometer repo 上檢查。
打開(kāi),例如, EhCache2Metrics ,您將找到 Ehcache 統(tǒng)計(jì)信息映射到特定 Micrometer 指標(biāo)的內(nèi)容和方式。
- cache.size -> StatisticsGateway:getSize cache.gets{result=miss} -> StatisticsGateway:cacheMissCount cache.gets{result=hit} -> StatisticsGateway:cacheHitCount cache.puts -> StatisticsGateway:cachePutCount cache.evictions -> StatisticsGateway:cacheEvictedCount cache.remoteSize -> StatisticsGateway::getRemoteSize cache.removals -> StatisticsGateway::cacheRemoveCount cache.puts.added{result=added} -> StatisticsGateway::cachePutAddedCount cache.puts.added{result=updated} -> StatisticsGateway::cachePutAddedCount cache.misses{reason=expired} -> StatisticsGateway::cacheMissExpiredCount) cache.misses{reason=notFound} -> StatisticsGateway::cacheMissNotFoundCount) cache.xa.commits{result=readOnly} -> StatisticsGateway::xaCommitReadOnlyCount cache.xa.commits{result=exception} -> StatisticsGateway::xaCommitExceptionCount cache.xa.commits{result=committed} -> StatisticsGateway::xaCommitCommittedCount cache.xa.rollbacks{result=exception} -> StatisticsGateway::xaRollbackExceptionCount cache.xa.rollbacks{result=success} -> StatisticsGateway::xaRollbackSuccessCount cache.xa.recoveries{result=nothing} -> StatisticsGateway::xaRecoveryNothingCount cache.xa.recoveries{result=success} -> StatisticsGateway::xaRecoveryRecoveredCount cache.local.offheap.size -> StatisticsGateway::getLocalOffHeapSize) cache.local.heap.size -> StatisticsGateway::getLocalHeapSizeInBytes cache.local.disk.size -> StatisticsGateway::getLocalDiskSizeInBytes
注冊(cè) MeterBinders 是非常簡(jiǎn)單的,示例可以在 micrometer 文檔 中被找到。
記住,您可以手動(dòng)操作:
- new ClassLoaderMetrics().bindTo(registry);
- new JvmMemoryMetrics().bindTo(registry);
- new EhCache2Metrics(cache, Tags.of("name", cache.getName())).bindTo(registry)
- new TomcatMetrics(manager, tags).bindTo(registry)
- ...
或者,您可以使用 Spring Boot ,它會(huì)在引擎下為您做這件事。
正如我之前提到的,actuator 將提供許多 AutoConfiguration s 和 MetricsBinders ,只要添加給定的依賴(lài)項(xiàng),它就會(huì)注冊(cè) MeterBinders 。
例如, TomcatMetricsBinder 將注冊(cè) TomcatMetrics (為您的嵌入式容器)。MeterRegistryConfigurer 將注冊(cè) JVM 、運(yùn)行時(shí)間 和其他系統(tǒng)指標(biāo)。
現(xiàn)在,假設(shè)您想在您的應(yīng)用程序中使用 Ehcache 。您可以添加兩個(gè)依賴(lài)項(xiàng):
- implementation("org.springframework.boot:spring-boot-starter-cache")
- implementation("net.sf.ehcache:ehcache")
然后注冊(cè)緩存(您也可以通過(guò) ehcache.xml 來(lái)實(shí)現(xiàn))
- @Bean
- Cache playCache(EhCacheCacheManager cacheManager) {
- CacheConfiguration cacheConfiguration = new CacheConfiguration()
- .name(CACHE_NAME)
- .maxEntriesLocalHeap(MAX_ELEMENTS_IN_MEMORY);
- Cache cache = new Cache(cacheConfiguration);
- cacheManager.getCacheManager().addCache(cache);
- cacheManager.initializeCaches();
- return cache;
- }
現(xiàn)在, CacheMetricsRegistrarConfiguration 將通過(guò) Spring 緩存管理器為每一個(gè)緩存管理注冊(cè) EhCache2Metrics 。
如果您不想使用 Spring 緩存管理,您也可以自己注冊(cè) EhCache2Metrics 。
現(xiàn)在,啟動(dòng)應(yīng)用程序,您將看到其他 ehcache 指標(biāo)。
- 2019-07-17 13:08:45.113 INFO 93052 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : cache.gets{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache,result=hit} throughput=12.95/s
- 2019-07-17 13:08:45.124 INFO 93052 --- [ Thread-4] i.m.c.i.logging.LoggingMeterRegistry : cache.misses{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache,reason=notFound} throughput=3.7/s
- 2019-07-17 13:08:45.124 INFO 93052 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : cache.gets{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache,result=miss} throughput=3.7/s
- 2019-07-17 13:08:48.840 INFO 93052 --- [ Thread-4] i.m.c.i.logging.LoggingMeterRegistry : cache.puts{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache} throughput=16.65/s
- 2019-07-17 13:08:48.840 INFO 93052 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : cache.misses{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache,reason=notFound} throughput=3.7/s
- 2019-07-17 13:08:48.841 INFO 93052 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : cache.puts{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache} throughput=16.65/s
- 2019-07-17 13:08:48.841 INFO 93052 --- [ Thread-4] i.m.c.i.logging.LoggingMeterRegistry : cache.puts.added{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache,result=updated} throughput=0.116667/s
- 2019-07-17 13:08:48.841 INFO 93052 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : cache.puts.added{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache,result=updated} throughput=0.116667/s
- 2019-07-17 13:08:48.841 INFO 93052 --- [ Thread-4] i.m.c.i.logging.LoggingMeterRegistry : cache.puts.added{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache,result=added} throughput=0.116667/s
- 2019-07-17 13:08:48.842 INFO 93052 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : cache.puts.added{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache,result=added} throughput=0.116667/s
- 2019-07-17 13:08:48.847 INFO 93052 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : cache.local.disk.size{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache} value=0 B
- 2019-07-17 13:08:48.847 INFO 93052 --- [ Thread-4] i.m.c.i.logging.LoggingMeterRegistry : cache.local.disk.size{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache} value=0 B
- 2019-07-17 13:08:48.908 INFO 93052 --- [ Thread-4] i.m.c.i.logging.LoggingMeterRegistry : cache.local.heap.size{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache} value=1.039062 KiB
- 2019-07-17 13:08:48.908 INFO 93052 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : cache.local.heap.size{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache} value=1.039062 KiB
- 2019-07-17 13:08:48.909 INFO 93052 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : cache.local.offheap.size{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache} value=0 B
- 2019-07-17 13:08:48.909 INFO 93052 --- [ Thread-4] i.m.c.i.logging.LoggingMeterRegistry : cache.local.offheap.size{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache} value=0 B
- 2019-07-17 13:08:48.909 INFO 93052 --- [ Thread-4] i.m.c.i.logging.LoggingMeterRegistry : cache.remoteSize{} value=0
- 2019-07-17 13:08:48.909 INFO 93052 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : cache.remoteSize{} value=0
- 2019-07-17 13:08:48.909 INFO 93052 --- [ Thread-4] i.m.c.i.logging.LoggingMeterRegistry : cache.size{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache} value=7
- 2019-07-17 13:08:48.909 INFO 93052 --- [trics-publisher] i.m.c.i.logging.LoggingMeterRegistry : cache.size{cache=playCache,cacheManagercacheManager=cacheManager,name=playCache} value=7
在這種情況下,指標(biāo)上下文中每個(gè)組件的職責(zé)可歸納為:
Ehcache 指標(biāo)架構(gòu)
您可以在 此處 提供的示例應(yīng)用中查看所有這些概念。

































