指标

概述

Spark 的指标被解耦到对应于 Spark 组件的不同实例中。在每个实例中，您可以配置一组向其报告指标的接收器。目前支持以下实例：

master：Spark 独立主进程。

applications：master 中的一个组件，负责报告各种应用程序。
worker：一个 Spark 独立工作进程。
executor: 一个 Spark 执行器。
driver：Spark 驱动程序进程（创建 SparkContext 的进程）。
shuffleService：Spark 随机播放服务。
applicationMaster: 在 YARN 上运行时的 Spark ApplicationMaster。
mesos_cluster: 在 Mesos 上运行时的 Spark 集群调度器。

每个实例都可以向零个或多个接收器报告。水槽包含在 org.apache.spark.metrics.sink包装中：

ConsoleSink：将指标信息记录到控制台。

CSVSink：定期将指标数据导出到 CSV 文件。

JmxSink：注册指标以在 JMX 控制台中查看。

MetricsServlet：在现有的 Spark UI 中添加一个 servlet，以将指标数据作为 JSON 数据提供。

PrometheusServlet：（实验性）在现有的 Spark UI 中添加一个 servlet，以提供 Prometheus 格式的指标数据。

GraphiteSink：将指标发送到 Graphite 节点。

Slf4jSink：将指标作为日志条目发送到 slf4j。

StatsdSink：将指标发送到 StatsD 节点。

任务配置

方式1：使用docker镜像中的配置文件

monitoring:
  exposeDriverMetrics: true
  exposeExecutorMetrics: true
  prometheus:
    jmxExporterJar: "/opt/spark/jars/jmx_prometheus_javaagent-0.17.2.jar"
    port: 8090
    portName: http-metric
    configFile: "/opt/spark/metrics/conf/prometheus.yaml"

lowercaseOutputName: true
attrNameSnakeCase: true
rules:
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.(BlockManager|DAGScheduler|jvm)\.(\S+)><>Value
    name: spark_driver_$3_$4
    type: GAUGE
    labels:
      app_namespace: "$1"
      app_id: "$2"
  # [ADD]
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.(BlockManager|DAGScheduler|jvm)\.(\S+).(\S+)><>Value
    name: spark_driver_$3_$4_$5
    type: GAUGE
    labels:
      app_namespace: "$1"
      app_id: "$2"
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.(\S+)\.StreamingMetrics\.streaming\.(\S+)><>Value
    name: spark_streaming_driver_$4
    type: GAUGE
    labels:
      app_namespace: "$1"
      app_id: "$2"
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.spark\.streaming\.(\S+)\.(\S+)><>Value
    name: spark_structured_streaming_driver_$4
    type: GAUGE
    labels:
      app_namespace: "$1"
      app_id: "$2"
      query_name: "$3"
  - pattern: metrics<name=(\S+)\.(\S+)\.(\S+)\.executor\.(\S+)><>Value
    name: spark_executor_$4
    type: GAUGE
    labels:
      app_namespace: "$1"
      app_id: "$2"
      executor_id: "$3"
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.DAGScheduler\.(.*)><>Count
    name: spark_driver_DAGScheduler_$3_count
    type: COUNTER
    labels:
      app_namespace: "$1"
      app_id: "$2"
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.HiveExternalCatalog\.(.*)><>Count
    name: spark_driver_HiveExternalCatalog_$3_count
    type: COUNTER
    labels:
      app_namespace: "$1"
      app_id: "$2"
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.CodeGenerator\.(.*)><>Count
    name: spark_driver_CodeGenerator_$3_count
    type: COUNTER
    labels:
      app_namespace: "$1"
      app_id: "$2"
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.LiveListenerBus\.(.*)><>Count
    name: spark_driver_LiveListenerBus_$3_count
    type: COUNTER
    labels:
      app_namespace: "$1"
      app_id: "$2"
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.LiveListenerBus\.(.*)><>Value
    name: spark_driver_LiveListenerBus_$3
    type: GAUGE
    labels:
      app_namespace: "$1"
      app_id: "$2"
  - pattern: metrics<name=(\S+)\.(\S+)\.(.*)\.executor\.(.*)><>Count
    name: spark_executor_$4_count
    type: COUNTER
    labels:
      app_namespace: "$1"
      app_id: "$2"
      executor_id: "$3"
  # [ADD]
  - pattern: metrics<name=(\S+)\.(\S+)\.([0-9]+)\.executor\.(.*)><>Value
    name: spark_executor_$4
    type: GAUGE
    labels:
      app_namespace: "$1"
      app_id: "$2"
      executor_id: "$3"
  - pattern: metrics<name=(\S+)\.(\S+)\.([0-9]+)\.(jvm|NettyBlockTransfer)\.(.*)><>Value
    name: spark_executor_$4_$5
    type: GAUGE
    labels:
      app_namespace: "$1"
      app_id: "$2"
      executor_id: "$3"
  - pattern: metrics<name=(\S+)\.(\S+)\.([0-9]+)\.HiveExternalCatalog\.(.*)><>Count
    name: spark_executor_HiveExternalCatalog_$4_count
    type: COUNTER
    labels:
      app_namespace: "$1"
      app_id: "$2"
      executor_id: "$3"
  - pattern: metrics<name=(\S+)\.(\S+)\.([0-9]+)\.CodeGenerator\.(.*)><>Count
    name: spark_executor_CodeGenerator_$4_count
    type: COUNTER
    labels:
      app_namespace: "$1"
      app_id: "$2"
      executor_id: "$3"

方式2：直接放在配置文件中

monitoring:
  exposeDriverMetrics: true
  exposeExecutorMetrics: true
  metricsProperties: |
    *.sink.jmx.class=org.apache.spark.metrics.sink.JmxSink
    # Enable JvmSource for instance master, worker, driver and executor
    master.source.jvm.class=org.apache.spark.metrics.source.JvmSource
    worker.source.jvm.class=org.apache.spark.metrics.source.JvmSource
    driver.source.jvm.class=org.apache.spark.metrics.source.JvmSource
    executor.source.jvm.class=org.apache.spark.metrics.source.JvmSource
  prometheus:
    jmxExporterJar: "/opt/spark/jars/jmx_prometheus_javaagent-0.17.2.jar"
    port: 8090
    configuration: |
      lowercaseOutputName: true
      attrNameSnakeCase: true
      rules:
        # These come from the application driver if it's a streaming application
        # Example: default/streaming.driver.com.example.ClassName.StreamingMetrics.streaming.lastCompletedBatch_schedulingDelay
        - pattern: metrics<name=(\S+)\.(\S+)\.driver\.(\S+)\.StreamingMetrics\.streaming\.(\S+)><>Value
          name: spark_streaming_driver_$4
          labels:
            app_namespace: "$1"
            app_id: "$2"
        # These come from the application driver if it's a structured streaming application
        # Example: default/streaming.driver.spark.streaming.QueryName.inputRate-total
        - pattern: metrics<name=(\S+)\.(\S+)\.driver\.spark\.streaming\.(\S+)\.(\S+)><>Value
          name: spark_structured_streaming_driver_$4
          labels:
            app_namespace: "$1"
            app_id: "$2"
            query_name: "$3"
        # These come from the application executors
        # Example: default/spark-pi.0.executor.threadpool.activeTasks
        - pattern: metrics<name=(\S+)\.(\S+)\.(\S+)\.executor\.(\S+)><>Value
          name: spark_executor_$4
          type: GAUGE
          labels:
            app_namespace: "$1"
            app_id: "$2"
            executor_id: "$3"
        # These come from the application driver
        # Example: default/spark-pi.driver.DAGScheduler.stage.failedStages
        - pattern: metrics<name=(\S+)\.(\S+)\.driver\.(BlockManager|DAGScheduler|jvm)\.(\S+)><>Value
          name: spark_driver_$3_$4
          type: GAUGE
          labels:
            app_namespace: "$1"
            app_id: "$2"        
        # [ADD]
        - pattern: metrics<name=(\S+)\.(\S+)\.driver\.(BlockManager|DAGScheduler|jvm)\.(\S+).(\S+)><>Value
          name: spark_driver_$3_$4_$5
          type: GAUGE
          labels:
            app_namespace: "$1"
            app_id: "$2"        
        # These come from the application driver
        # Emulate timers for DAGScheduler like messagePRocessingTime
        - pattern: metrics<name=(\S+)\.(\S+)\.driver\.DAGScheduler\.(.*)><>Count
          name: spark_driver_DAGScheduler_$3_count
          type: COUNTER
          labels:
            app_namespace: "$1"
            app_id: "$2"
        # HiveExternalCatalog is of type counter
        - pattern: metrics<name=(\S+)\.(\S+)\.driver\.HiveExternalCatalog\.(.*)><>Count
          name: spark_driver_HiveExternalCatalog_$3_count
          type: COUNTER
          labels:
            app_namespace: "$1"
            app_id: "$2"
        # These come from the application driver
        # Emulate histograms for CodeGenerator
        - pattern: metrics<name=(\S+)\.(\S+)\.driver\.CodeGenerator\.(.*)><>Count
          name: spark_driver_CodeGenerator_$3_count
          type: COUNTER
          labels:
            app_namespace: "$1"
            app_id: "$2"
        # These come from the application driver
        # Emulate timer (keep only count attribute) plus counters for LiveListenerBus
        - pattern: metrics<name=(\S+)\.(\S+)\.driver\.LiveListenerBus\.(.*)><>Count
          name: spark_driver_LiveListenerBus_$3_count
          type: COUNTER
          labels:
            app_namespace: "$1"
            app_id: "$2"
        # Get Gauge type metrics for LiveListenerBus
        - pattern: metrics<name=(\S+)\.(\S+)\.driver\.LiveListenerBus\.(.*)><>Value
          name: spark_driver_LiveListenerBus_$3
          type: GAUGE
          labels:
            app_namespace: "$1"
            app_id: "$2"
        # Executors counters
        - pattern: metrics<name=(\S+)\.(\S+)\.(.*)\.executor\.(.*)><>Count
          name: spark_executor_$4_count
          type: COUNTER
          labels:
            app_namespace: "$1"
            app_id: "$2"
            executor_id: "$3"
        # [ADD]
        - pattern: metrics<name=(\S+)\.(\S+)\.([0-9]+)\.executor\.(.*)><>Value
          name: spark_executor_$4
          type: GAUGE
          labels:
            app_namespace: "$1"
            app_id: "$2"
            executor_id: "$3"
        # These come from the application executors
        # Example: app-20160809000059-0000.0.jvm.threadpool.activeTasks
        - pattern: metrics<name=(\S+)\.(\S+)\.([0-9]+)\.(jvm|NettyBlockTransfer)\.(.*)><>Value
          name: spark_executor_$4_$5
          type: GAUGE
          labels:
            app_namespace: "$1"
            app_id: "$2"
            executor_id: "$3"
        - pattern: metrics<name=(\S+)\.(\S+)\.([0-9]+)\.HiveExternalCatalog\.(.*)><>Count
          name: spark_executor_HiveExternalCatalog_$4_count
          type: COUNTER
          labels:
            app_namespace: "$1"
            app_id: "$2"
            executor_id: "$3"
        # These come from the application driver
        # Emulate histograms for CodeGenerator
        - pattern: metrics<name=(\S+)\.(\S+)\.([0-9]+)\.CodeGenerator\.(.*)><>Count
          name: spark_executor_CodeGenerator_$4_count
          type: COUNTER
          labels:
            app_namespace: "$1"
            app_id: "$2"
            executor_id: "$3

镜像配置

根目录创建metrics文件夹

metrics.properties

*.sink.jmx.class=org.apache.spark.metrics.sink.JmxSink
# Enable JvmSource for instance master, worker, driver and executor
master.source.jvm.class=org.apache.spark.metrics.source.JvmSource
worker.source.jvm.class=org.apache.spark.metrics.source.JvmSource
driver.source.jvm.class=org.apache.spark.metrics.source.JvmSource
executor.source.jvm.class=org.apache.spark.metrics.source.JvmSource

prometheus.yaml

---
lowercaseOutputName: true
attrNameSnakeCase: true
rules:
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.(BlockManager|DAGScheduler|jvm)\.(\S+)><>Value
    name: spark_driver_$3_$4
    type: GAUGE
    labels:
      app_namespace: "$1"
      app_id: "$2"
  # [ADD]
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.(BlockManager|DAGScheduler|jvm)\.(\S+).(\S+)><>Value
    name: spark_driver_$3_$4_$5
    type: GAUGE
    labels:
      app_namespace: "$1"
      app_id: "$2"
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.(\S+)\.StreamingMetrics\.streaming\.(\S+)><>Value
    name: spark_streaming_driver_$4
    type: GAUGE
    labels:
      app_namespace: "$1"
      app_id: "$2"
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.spark\.streaming\.(\S+)\.(\S+)><>Value
    name: spark_structured_streaming_driver_$4
    type: GAUGE
    labels:
      app_namespace: "$1"
      app_id: "$2"
      query_name: "$3"
  - pattern: metrics<name=(\S+)\.(\S+)\.(\S+)\.executor\.(\S+)><>Value
    name: spark_executor_$4
    type: GAUGE
    labels:
      app_namespace: "$1"
      app_id: "$2"
      executor_id: "$3"
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.DAGScheduler\.(.*)><>Count
    name: spark_driver_DAGScheduler_$3_count
    type: COUNTER
    labels:
      app_namespace: "$1"
      app_id: "$2"
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.HiveExternalCatalog\.(.*)><>Count
    name: spark_driver_HiveExternalCatalog_$3_count
    type: COUNTER
    labels:
      app_namespace: "$1"
      app_id: "$2"
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.CodeGenerator\.(.*)><>Count
    name: spark_driver_CodeGenerator_$3_count
    type: COUNTER
    labels:
      app_namespace: "$1"
      app_id: "$2"
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.LiveListenerBus\.(.*)><>Count
    name: spark_driver_LiveListenerBus_$3_count
    type: COUNTER
    labels:
      app_namespace: "$1"
      app_id: "$2"
  - pattern: metrics<name=(\S+)\.(\S+)\.driver\.LiveListenerBus\.(.*)><>Value
    name: spark_driver_LiveListenerBus_$3
    type: GAUGE
    labels:
      app_namespace: "$1"
      app_id: "$2"
  - pattern: metrics<name=(\S+)\.(\S+)\.(.*)\.executor\.(.*)><>Count
    name: spark_executor_$4_count
    type: COUNTER
    labels:
      app_namespace: "$1"
      app_id: "$2"
      executor_id: "$3"
  # [ADD]
  - pattern: metrics<name=(\S+)\.(\S+)\.([0-9]+)\.executor\.(.*)><>Value
    name: spark_executor_$4
    type: GAUGE
    labels:
      app_namespace: "$1"
      app_id: "$2"
      executor_id: "$3"
  - pattern: metrics<name=(\S+)\.(\S+)\.([0-9]+)\.(jvm|NettyBlockTransfer)\.(.*)><>Value
    name: spark_executor_$4_$5
    type: GAUGE
    labels:
      app_namespace: "$1"
      app_id: "$2"
      executor_id: "$3"
  - pattern: metrics<name=(\S+)\.(\S+)\.([0-9]+)\.HiveExternalCatalog\.(.*)><>Count
    name: spark_executor_HiveExternalCatalog_$4_count
    type: COUNTER
    labels:
      app_namespace: "$1"
      app_id: "$2"
      executor_id: "$3"
  - pattern: metrics<name=(\S+)\.(\S+)\.([0-9]+)\.CodeGenerator\.(.*)><>Count
    name: spark_executor_CodeGenerator_$4_count
    type: COUNTER
    labels:
      app_namespace: "$1"
      app_id: "$2"
      executor_id: "$3"

Dockerfile里面增加

RUN mkdir -p /opt/spark/metrics/conf
COPY metrics/metrics.properties /opt/spark/metrics/conf
COPY metrics/prometheus.yaml /opt/spark/metrics/conf

可用指标提供者列表

Spark 使用的指标有多种类型：gauge、counter、histogram、meter 和 timer，详见Dropwizard 库文档。以下组件和指标列表报告了名称和有关可用指标的一些详细信息，按组件实例和源命名空间分组。Spark 仪器中最常用的指标时间是仪表和计数器。可以识别计数器，因为它们具有.count后缀。计时器、仪表和直方图在列表中进行了注释，其余列表元素是 gauge 类型的指标。大多数指标在其父组件实例配置后立即激活，一些指标还需要通过额外的配置参数启用，详细信息在列表中报告。

Component instance = Driver

This is the component with the largest amount of instrumented metrics

namespace=BlockManager

disk.diskSpaceUsed_MB
memory.maxMem_MB
memory.maxOffHeapMem_MB
memory.maxOnHeapMem_MB
memory.memUsed_MB
memory.offHeapMemUsed_MB
memory.onHeapMemUsed_MB
memory.remainingMem_MB
memory.remainingOffHeapMem_MB
memory.remainingOnHeapMem_MB

namespace=HiveExternalCatalog

note: these metrics are conditional to a configuration parameter: spark.metrics.staticSources.enabled (default is true)
fileCacheHits.count
filesDiscovered.count
hiveClientCalls.count
parallelListingJobCount.count
partitionsFetched.count

namespace=CodeGenerator

note: these metrics are conditional to a configuration parameter: spark.metrics.staticSources.enabled (default is true)
compilationTime (histogram)
generatedClassSize (histogram)
generatedMethodSize (histogram)
sourceCodeSize (histogram)

namespace=DAGScheduler

job.activeJobs
job.allJobs
messageProcessingTime (timer)
stage.failedStages
stage.runningStages
stage.waitingStages

namespace=LiveListenerBus

listenerProcessingTime.org.apache.spark.HeartbeatReceiver (timer)
listenerProcessingTime.org.apache.spark.scheduler.EventLoggingListener (timer)
listenerProcessingTime.org.apache.spark.status.AppStatusListener (timer)
numEventsPosted.count
queue.appStatus.listenerProcessingTime (timer)
queue.appStatus.numDroppedEvents.count
queue.appStatus.size
queue.eventLog.listenerProcessingTime (timer)
queue.eventLog.numDroppedEvents.count
queue.eventLog.size
queue.executorManagement.listenerProcessingTime (timer)

namespace=appStatus (all metrics of type=counter)

note: Introduced in Spark 3.0. Conditional to a configuration parameter:spark.metrics.appStatusSource.enabled (default is false)
stages.failedStages.count
stages.skippedStages.count
stages.completedStages.count
tasks.blackListedExecutors.count // deprecated use excludedExecutors instead
tasks.excludedExecutors.count
tasks.completedTasks.count
tasks.failedTasks.count
tasks.killedTasks.count
tasks.skippedTasks.count
tasks.unblackListedExecutors.count // deprecated use unexcludedExecutors instead
tasks.unexcludedExecutors.count
jobs.succeededJobs
jobs.failedJobs
jobDuration

namespace=AccumulatorSource

note: User-configurable sources to attach accumulators to metric system
DoubleAccumulatorSource
LongAccumulatorSource

namespace=spark.streaming

note: This applies to Spark Structured Streaming only. Conditional to a configuration parameter: spark.sql.streaming.metricsEnabled=true (default is false)
eventTime-watermark
inputRate-total
latency
processingRate-total
states-rowsTotal
states-usedBytes

namespace=JVMCPU

jvmCpuTime

namespace=executor

note: These metrics are available in the driver in local mode only.
A full list of available metrics in this namespace can be found in the corresponding entry for the Executor component instance.

namespace=ExecutorMetrics

note: these metrics are conditional to a configuration parameter: spark.metrics.executorMetricsSource.enabled (default is true)
This source contains memory-related metrics. A full list of available metrics in this namespace can be found in the corresponding entry for the Executor component instance.

namespace=ExecutorAllocationManager

note: these metrics are only emitted when using dynamic allocation. Conditional to a configuration parameter spark.dynamicAllocation.enabled (default is false)
executors.numberExecutorsToAdd
executors.numberExecutorsPendingToRemove
executors.numberAllExecutors
executors.numberTargetExecutors
executors.numberMaxNeededExecutors
executors.numberExecutorsGracefullyDecommissioned.count
executors.numberExecutorsDecommissionUnfinished.count
executors.numberExecutorsExitedUnexpectedly.count
executors.numberExecutorsKilledByDriver.count

namespace=plugin.

Optional namespace(s). Metrics in this namespace are defined by user-supplied code, and configured using the Spark plugin API. See “Advanced Instrumentation” below for how to load custom plugins into Spark.

Component instance = Executor

These metrics are exposed by Spark executors.

namespace=executor (metrics are of type counter or gauge)

spark.executor.metrics.fileSystemSchemes (default: file,hdfs) determines the exposed file system metrics.
notes:
bytesRead.count
bytesWritten.count
cpuTime.count
deserializeCpuTime.count
deserializeTime.count
diskBytesSpilled.count
filesystem.file.largeRead_ops
filesystem.file.read_bytes
filesystem.file.read_ops
filesystem.file.write_bytes
filesystem.file.write_ops
filesystem.hdfs.largeRead_ops
filesystem.hdfs.read_bytes
filesystem.hdfs.read_ops
filesystem.hdfs.write_bytes
filesystem.hdfs.write_ops
jvmGCTime.count
memoryBytesSpilled.count
recordsRead.count
recordsWritten.count
resultSerializationTime.count
resultSize.count
runTime.count
shuffleBytesWritten.count
shuffleFetchWaitTime.count
shuffleLocalBlocksFetched.count
shuffleLocalBytesRead.count
shuffleRecordsRead.count
shuffleRecordsWritten.count
shuffleRemoteBlocksFetched.count
shuffleRemoteBytesRead.count
shuffleRemoteBytesReadToDisk.count
shuffleTotalBytesRead.count
shuffleWriteTime.count
succeededTasks.count
threadpool.activeTasks
threadpool.completeTasks
threadpool.currentPool_size
threadpool.maxPool_size
threadpool.startedTasks

namespace=ExecutorMetrics

ProcessTreeJVMVMemory
ProcessTreeJVMRSSMemory
ProcessTreePythonVMemory
ProcessTreePythonRSSMemory
ProcessTreeOtherVMemory
ProcessTreeOtherRSSMemory
note: “ProcessTree” metrics are collected only under certain conditions. The conditions are the logical AND of the following: /proc filesystem exists, spark.executor.processTreeMetrics.enabled=true. “ProcessTree” metrics report 0 when those conditions are not met.
These metrics are conditional to a configuration parameter: spark.metrics.executorMetricsSource.enabled (default value is true)
ExecutorMetrics are updated as part of heartbeat processes scheduled for the executors and for the driver at regular intervals: spark.executor.heartbeatInterval (default value is 10 seconds)
An optional faster polling mechanism is available for executor memory metrics, it can be activated by setting a polling interval (in milliseconds) using the configuration parameter spark.executor.metrics.pollingInterval
notes:
JVMHeapMemory
JVMOffHeapMemory
OnHeapExecutionMemory
OnHeapStorageMemory
OnHeapUnifiedMemory
OffHeapExecutionMemory
OffHeapStorageMemory
OffHeapUnifiedMemory
DirectPoolMemory
MappedPoolMemory
MinorGCCount
MinorGCTime
MajorGCCount
MajorGCTime
“ProcessTree*” metric counters:

namespace=JVMCPU

jvmCpuTime

namespace=NettyBlockTransfer

shuffle-client.usedDirectMemory
shuffle-client.usedHeapMemory
shuffle-server.usedDirectMemory
shuffle-server.usedHeapMemory

namespace=HiveExternalCatalog

note: these metrics are conditional to a configuration parameter: spark.metrics.staticSources.enabled (default is true)
fileCacheHits.count
filesDiscovered.count
hiveClientCalls.count
parallelListingJobCount.count
partitionsFetched.count

namespace=CodeGenerator

note: these metrics are conditional to a configuration parameter: spark.metrics.staticSources.enabled (default is true)
compilationTime (histogram)
generatedClassSize (histogram)
generatedMethodSize (histogram)
sourceCodeSize (histogram)

namespace=plugin.

Optional namespace(s). Metrics in this namespace are defined by user-supplied code, and configured using the Spark plugin API. See “Advanced Instrumentation” below for how to load custom plugins into Spark.

Source = JVM Source

Notes:

Activate this source by setting the relevant metrics.properties file entry or the configuration parameter:spark.metrics.conf.*.source.jvm.class=org.apache.spark.metrics.source.JvmSource
These metrics are conditional to a configuration parameter: spark.metrics.staticSources.enabled (default is true)
This source is available for driver and executor instances and is also available for other instances.
This source provides information on JVM metrics using the Dropwizard/Codahale Metric Sets for JVM instrumentation and in particular the metric sets BufferPoolMetricSet, GarbageCollectorMetricSet and MemoryUsageGaugeSet.

Component instance = applicationMaster

Note: applies when running on YARN

numContainersPendingAllocate
numExecutorsFailed
numExecutorsRunning
numLocalityAwareTasks
numReleasedContainers

Component instance = mesos_cluster

Note: applies when running on mesos

waitingDrivers
launchedDrivers
retryDrivers

Component instance = master

Note: applies when running in Spark standalone as master

workers
aliveWorkers
apps
waitingApps

Component instance = ApplicationSource

Note: applies when running in Spark standalone as master

status
runtime_ms
cores

Component instance = worker

Note: applies when running in Spark standalone as worker

executors
coresUsed
memUsed_MB
coresFree
memFree_MB

Component instance = shuffleService

Note: applies to the shuffle service

blockTransferRate (meter) - rate of blocks being transferred
blockTransferMessageRate (meter) - rate of block transfer messages, i.e. if batch fetches are enabled, this represents number of batches rather than number of blocks
blockTransferRateBytes (meter)
blockTransferAvgTime_1min (gauge - 1-minute moving average)
numActiveConnections.count
numRegisteredConnections.count
numCaughtExceptions.count
openBlockRequestLatencyMillis (histogram)
registerExecutorRequestLatencyMillis (histogram)
registeredExecutorsSize
shuffle-server.usedDirectMemory
shuffle-server.usedHeapMemory

参考：

https://spark.apache.org/docs/3.2.2/monitoring.html

spark 监控梳理

指标

概述