天达云 科技型企业 | 亚太互联网络信息中心(APNIC)成员 | 注册免费体验
天达云首页 > 互联网学习教程 > 大数据 > Uber jvm profiler如何使用

Uber jvm profiler如何使用
更新：HHH   时间：2023-1-7

这篇文章将为大家详细讲解有关Uber jvm profiler如何使用，小编觉得挺实用的，因此分享给大家做个参考，希望大家阅读完这篇文章后可以有所收获。

背景

uber jvm profiler是用于在分布式监控收集jvm 相关指标，如:cpu/memory/io/gc信息等

安装

确保安装了maven和JDK>=8前提下，直接mvn clean package

java application

  • 说明

    直接以java agent的部署就可以使用

  • 使用

    java -javaagent:jvm-profiler-1.0.0.jar=reporter=com.uber.profiling.reporters.KafkaOutputReporter,brokerList='kafka1:9092',topicPrefix=demo_,tag=tag-demo,metricInterval=5000,sampleInterval=0 -cp target/jvm-profiler-1.0.0.jar

  • 选项解释

参数说明
reporterreporter类别, 此处直接默认为com.uber.profiling.reporters.KafkaOutputReporter就可以
brokerList如reporter为com.uber.profiling.reporters.KafkaOutputReporter,则brokerList为kafka列表,以逗号分隔
topicPrefix如reporter为com.uber.profiling.reporters.KafkaOutputReporter,则topicPrefix为kafka topic的前缀
tagkey为tag的metric，会输出到reporter中
metricIntervalmetric report的频率，根据实际情况设置，单位为ms
sampleIntervaljvm堆栈metrics report的频率，根据实际情况设置，单位为ms

  • 结果展示

  "nonHeapMemoryTotalUsed": 11890584.0,
  "bufferPools": [
      {
          "totalCapacity": 0,
          "name": "direct",
          "count": 0,
          "memoryUsed": 0
      },
      {
          "totalCapacity": 0,
          "name": "mapped",
          "count": 0,
          "memoryUsed": 0
      }
  ],
  "heapMemoryTotalUsed": 24330736.0,
  "epochMillis": 1515627003374,
  "nonHeapMemoryCommitted": 13565952.0,
  "heapMemoryCommitted": 257425408.0,
  "memoryPools": [
      {
          "peakUsageMax": 251658240,
          "usageMax": 251658240,
          "peakUsageUsed": 1194496,
          "name": "Code Cache",
          "peakUsageCommitted": 2555904,
          "usageUsed": 1173504,
          "type": "Non-heap memory",
          "usageCommitted": 2555904
      },
      {
          "peakUsageMax": -1,
          "usageMax": -1,
          "peakUsageUsed": 9622920,
          "name": "Metaspace",
          "peakUsageCommitted": 9830400,
          "usageUsed": 9622920,
          "type": "Non-heap memory",
          "usageCommitted": 9830400
      },
      {
          "peakUsageMax": 1073741824,
          "usageMax": 1073741824,
          "peakUsageUsed": 1094160,
          "name": "Compressed Class Space",
          "peakUsageCommitted": 1179648,
          "usageUsed": 1094160,
          "type": "Non-heap memory",
          "usageCommitted": 1179648
      },
      {
          "peakUsageMax": 1409286144,
          "usageMax": 1409286144,
          "peakUsageUsed": 24330736,
          "name": "PS Eden Space",
          "peakUsageCommitted": 67108864,
          "usageUsed": 24330736,
          "type": "Heap memory",
          "usageCommitted": 67108864
      },
      {
          "peakUsageMax": 11010048,
          "usageMax": 11010048,
          "peakUsageUsed": 0,
          "name": "PS Survivor Space",
          "peakUsageCommitted": 11010048,
          "usageUsed": 0,
          "type": "Heap memory",
          "usageCommitted": 11010048
      },
      {
          "peakUsageMax": 2863661056,
          "usageMax": 2863661056,
          "peakUsageUsed": 0,
          "name": "PS Old Gen",
          "peakUsageCommitted": 179306496,
          "usageUsed": 0,
          "type": "Heap memory",
          "usageCommitted": 179306496
      }
  ],
  "processCpuLoad": 0.0008024004394748531,
  "systemCpuLoad": 0.23138430784607697,
  "processCpuTime": 496918000,
  "appId": null,
  "name": "24103@machine01",
  "host": "machine01",
  "processUuid": "3c2ec835-749d-45ea-a7ec-e4b9fe17c23a",
  "tag": "mytag",
  "gc": [
      {
          "collectionTime": 0,
          "name": "PS Scavenge",
          "collectionCount": 0
      },
      {
          "collectionTime": 0,
          "name": "PS MarkSweep",
          "collectionCount": 0
      }
  ]
}

spark application

  • 说明

    和java应用不同，需要把jvm-profiler.jar分发到各个节点上

  • 使用

       --jars hdfs:///public/libs/jvm-profiler-1.0.0.jar   
   --conf spark.driver.extraJavaOptions=-javaagent:jvm-profiler-1.0.0.jar=reporter=com.uber.profiling.reporters.KafkaOutputReporter,brokerList='kafka1:9092',topicPrefix=demo_,tag=tag-demo,metricInterval=5000,sampleInterval=0 
   --conf spark.executor.extraJavaOptions=-javaagent:jvm-profiler-1.0.0.jar=reporter=com.uber.profiling.reporters.KafkaOutputReporter,brokerList='kafka1:9092',topicPrefix=demo_,tag=tag-demo,metricInterval=5000,sampleInterval=0


  • 选项解释

参数说明
reporterreporter类别, 此处直接默认为com.uber.profiling.reporters.KafkaOutputReporter就可以
brokerList如reporter为com.uber.profiling.reporters.KafkaOutputReporter,则brokerList为kafka列表,以逗号分隔
topicPrefix如reporter为com.uber.profiling.reporters.KafkaOutputReporter,则topicPrefix为kafka topic的前缀
tagkey为tag的metric，会输出到reporter中
metricIntervalmetric report的频率，根据实际情况设置，单位为ms
sampleIntervaljvm堆栈metrics report的频率，根据实际情况设置，单位为ms

  • 结果展示

  "nonHeapMemoryTotalUsed": 11890584.0,
  "bufferPools": [
      {
          "totalCapacity": 0,
          "name": "direct",
          "count": 0,
          "memoryUsed": 0
      },
      {
          "totalCapacity": 0,
          "name": "mapped",
          "count": 0,
          "memoryUsed": 0
      }
  ],
  "heapMemoryTotalUsed": 24330736.0,
  "epochMillis": 1515627003374,
  "nonHeapMemoryCommitted": 13565952.0,
  "heapMemoryCommitted": 257425408.0,
  "memoryPools": [
      {
          "peakUsageMax": 251658240,
          "usageMax": 251658240,
          "peakUsageUsed": 1194496,
          "name": "Code Cache",
          "peakUsageCommitted": 2555904,
          "usageUsed": 1173504,
          "type": "Non-heap memory",
          "usageCommitted": 2555904
      },
      {
          "peakUsageMax": -1,
          "usageMax": -1,
          "peakUsageUsed": 9622920,
          "name": "Metaspace",
          "peakUsageCommitted": 9830400,
          "usageUsed": 9622920,
          "type": "Non-heap memory",
          "usageCommitted": 9830400
      },
      {
          "peakUsageMax": 1073741824,
          "usageMax": 1073741824,
          "peakUsageUsed": 1094160,
          "name": "Compressed Class Space",
          "peakUsageCommitted": 1179648,
          "usageUsed": 1094160,
          "type": "Non-heap memory",
          "usageCommitted": 1179648
      },
      {
          "peakUsageMax": 1409286144,
          "usageMax": 1409286144,
          "peakUsageUsed": 24330736,
          "name": "PS Eden Space",
          "peakUsageCommitted": 67108864,
          "usageUsed": 24330736,
          "type": "Heap memory",
          "usageCommitted": 67108864
      },
      {
          "peakUsageMax": 11010048,
          "usageMax": 11010048,
          "peakUsageUsed": 0,
          "name": "PS Survivor Space",
          "peakUsageCommitted": 11010048,
          "usageUsed": 0,
          "type": "Heap memory",
          "usageCommitted": 11010048
      },
      {
          "peakUsageMax": 2863661056,
          "usageMax": 2863661056,
          "peakUsageUsed": 0,
          "name": "PS Old Gen",
          "peakUsageCommitted": 179306496,
          "usageUsed": 0,
          "type": "Heap memory",
          "usageCommitted": 179306496
      }
  ],
  "processCpuLoad": 0.0008024004394748531,
  "systemCpuLoad": 0.23138430784607697,
  "processCpuTime": 496918000,
  "appId": null,
  "name": "24103@machine01",
  "host": "machine01",
  "processUuid": "3c2ec835-749d-45ea-a7ec-e4b9fe17c23a",
  "tag": "mytag",
  "gc": [
      {
          "collectionTime": 0,
          "name": "PS Scavenge",
          "collectionCount": 0
      },
      {
          "collectionTime": 0,
          "name": "PS MarkSweep",
          "collectionCount": 0
      }
  ]
}

分析

  • 已有的reporter

reporter说明
ConsoleOutputReporter默认的repoter，一般用于调试
FileOutputReporter基于文件的reporter,分布式环境下不适用，得设置outputDir
KafkaOutputReporter基于kafka的reporter，正式环境用的多，得设置brokerList，topicPrefix
GraphiteOutputReporter基于Graphite的reporter,需设置graphite.host等配置
RedisOutputReporter基于redis的reporter，构建命令 mvn -P redis clean package
InfluxDBOutputReporter基于InfluxDB的reporter，构建命令mvn -P influxdb clean package，需设置influxdb.host等配置
建议在生产环境下使用KafkaOutputReporter，操作灵活性高，可以结合clickhouse grafana进行指标展示

  • 源码分析

    该jvm-profiler整体是基于java agent实现,项目pom文件 指定了MANIFEST.MF中的Premain-Class项和Agent-Class为com.uber.profiling.Agent 具体的实现类为AgentImpl
    就具体的AgentImpl类的run方法来进行分析

    public void run(Arguments arguments, Instrumentation instrumentation, Collection<AutoCloseable> objectsToCloseOnShutdown) {
      if (arguments.isNoop()) {
          logger.info("Agent noop is true, do not run anything");
          return;
      }

      Reporter reporter = arguments.getReporter();

      String processUuid = UUID.randomUUID().toString();

      String appId = null;

      String appIdVariable = arguments.getAppIdVariable();
      if (appIdVariable != null && !appIdVariable.isEmpty()) {
          appId = System.getenv(appIdVariable);
      }

      if (appId == null || appId.isEmpty()) {
          appId = SparkUtils.probeAppId(arguments.getAppIdRegex());
      }

      if (!arguments.getDurationProfiling().isEmpty()
              || !arguments.getArgumentProfiling().isEmpty()) {
          instrumentation.addTransformer(new JavaAgentFileTransformer(arguments.getDurationProfiling(), arguments.getArgumentProfiling()));
      }

      List<Profiler> profilers = createProfilers(reporter, arguments, processUuid, appId);

      ProfilerGroup profilerGroup = startProfilers(profilers);

      Thread shutdownHook = new Thread(new ShutdownHookRunner(profilerGroup.getPeriodicProfilers(), Arrays.asList(reporter), objectsToCloseOnShutdown));
      Runtime.getRuntime().addShutdownHook(shutdownHook);
  }

     

    • arguments.getReporter() 获取reporter，如果没有设置则设置为reporterConstructor,否则设置为指定的reporter

    • String appId ,设置appId，首先从配置中查找，如果没有设置，再从env中查找，对于spark应用则取spark.app.id的值

    • List<Profiler> profilers = createProfilers(reporter, arguments, processUuid, appId)，创建profilers,默认有CpuAndMemoryProfiler，ThreadInfoProfiler，ProcessInfoProfiler ；
      1.其中CpuAndMemoryProfiler，ThreadInfoProfiler，ProcessInfoProfiler是从JMX中读取数据，ProcessInfoProfiler还会从 /pro读取数据；
      2.如果设置了durationProfiling，argumentProfiling，sampleInterval，ioProfiling，则会增加对应的MethodDurationProfiler(输出方法调用花费的时间)，MethodArgumentProfiler(输出方法参数的值)，StacktraceReporterProfiler，IOProfiler；
      3.MethodArgumentProfiler和MethodDurationProfiler利用javassist第三方字节码编译工具来改写对应的类，具体实现参照JavaAgentFileTransformer
      4.StacktraceReporterProfiler从JMX中读取数据
      5.IOProfiler则是读取本地机器上的/pro文件对应的目录的数据

    • ProfilerGroup profilerGroup = startProfilers(profilers) 开始进行profiler的定时report
      其中还会区分oneTimeProfilers和periodicProfilers，ProcessInfoProfiler就属于oneTimeProfilers，因为process的信息，在运行期间是不会变的，不需要周期行的reporter
      至此，整个流程结束

关于“Uber jvm profiler如何使用”这篇文章就分享到这里了，希望以上内容可以对大家有一定的帮助，使各位可以学到更多知识，如果觉得文章不错，请把它分享出去让更多的人看到。
返回大数据教程...
400-837-6568
7 * 24小时全天全国服务热线400电话
天达云·专业的云计算服务平台 全国400服务电话：400-837-6568
Copyright © 2011-2025 Tdyun.com. All Rights Reserved. 天达云 版权所有 粤ICP备18083267号-7