Apache Flink 1.20中Graphite Metric Reporter端口值无法识别求助

我在Apache Flink 1.20中配置Graphite作为指标上报数据库，配置内容如下：

metrics:
  scope:
     operator: <job_name>.<operator_name>
     task: <job_name>.<task_name>
  reporters: grph,prom
  reporter:
    grph:
      factory:
        class: org.apache.flink.metrics.graphite.GraphiteReporterFactory
      host: localhost3
      port: 2001
      protocol: TCP
      interval: 10 SECONDS
    prom:
      factory:
        class: org.apache.flink.metrics.prometheus.PrometheusReporterFactory
      port: 9249-9251
      scope:
        operator: <job_name>.<operator_name>
        task: <job_name>.<task_name>

启动集群时出现异常，报错信息显示主机名已解析但端口无效：

2025-03-12 19:31:26,025 ERROR org.apache.flink.runtime.metrics.ReporterSetup                   [] - Could not instantiate MetricReporter grph. Metrics might not be exposed/reported.
java.lang.IllegalArgumentException: Invalid host/port configuration. Host: localhost3 
Port: -1
        at     org.apache.flink.metrics.graphite.GraphiteReporter.getReporter(GraphiteReporter.java:51) ~    [?:?]
        at     org.apache.flink.dropwizard.ScheduledDropwizardReporter.open(ScheduledDropwizardReporter.ja    va:113) ~[?:?]
        at     org.apache.flink.runtime.metrics.ReporterSetup.createReporterSetup(ReporterSetup.java:185)     ~[flink-dist-1.20.0.jar:1.20.0]
        at     org.apache.flink.runtime.metrics.ReporterSetup.lambda$setupReporters$2(ReporterSetup.java:3    60) ~[flink-dist-1.20.0.jar:1.20.0]

但Flink启动日志显示配置值已被正确加载：

INFO  [] - Loading configuration property: metrics.reporter.grph.port, 2001
INFO  [] - Loading configuration property: metrics.reporter.grph.host, localhost3
INFO  [] - Loading configuration property: metrics.reporter.grph.interval, 10 SECONDS
INFO  [] - Loading configuration property: metrics.reporter.grph.interval, 10 SECONDS

请问为何端口值无法被正确解析，反而显示为默认值-1？

问题根源 #

Flink的GraphiteReporter在读取配置时，要求host、port、protocol等Graphite专属参数必须嵌套在graphite子节点下，而非直接放在reporter.grph层级。当前配置把这些参数直接放在grph节点下，导致GraphiteReporter无法识别并读取端口值，最终使用默认的-1。

修复后的配置 #

调整Graphite相关参数的层级结构，将host、port、protocol放入graphite子节点中：

metrics:
  scope:
     operator: <job_name>.<operator_name>
     task: <job_name>.<task_name>
  reporters: grph,prom
  reporter:
    grph:
      factory:
        class: org.apache.flink.metrics.graphite.GraphiteReporterFactory
      graphite:
        host: localhost3
        port: 2001
        protocol: TCP
      interval: 10 SECONDS
    prom:
      factory:
        class: org.apache.flink.metrics.prometheus.PrometheusReporterFactory
      port: 9249-9251
      scope:
        operator: <job_name>.<operator_name>
        task: <job_name>.<task_name>

验证说明 #

• Flink官方定义的GraphiteReporter配置项路径均为metrics.reporter.<reporter-name>.graphite.*，只有符合该路径的参数才会被Reporter读取
• 启动日志中显示的metrics.reporter.grph.port只是Flink加载了配置项，但该参数并未被GraphiteReporter识别，因此被忽略

内容的提问来源于stack exchange，提问作者Kenank