[SPARK-40506] Spark Streaming metrics name don't need application name

• Type: Improvement
• Status: In Progress
• Priority: Trivial
• Resolution: Unresolved
• Affects Version/s: 3.2.2
• Component/s: DStreams
• Assignee: Unassigned
• Reporter: 王俊博
• Created: 20/Sep/22 12:15
• Updated: 21/Sep/22 04:53

Description

Spark  StreamingSource  Metrics sourceName is inappropriate.The label now looks like `application_xxxxx_xxxx_driver_NetworkWordCount_StreamingMetrics_streaming_lastCompletedBatch_processingEndTime `, instead of `application_xxxxx_xxxx_driver_StreamingMetrics_streaming_lastCompletedBatch_processingEndTime`, the Spark app name is not need.

This makes it hard to use metrics for different Spark applications over time. And this makes the metrics sourceName standard inconsistent.

//代码占位符

private[streaming] class StreamingSource(ssc: StreamingContext) extends Source {
  override val metricRegistry = new MetricRegistry
  override val sourceName = "%s.StreamingMetrics".format(ssc.sparkContext.appName)
....
}

And for example, other metrics sourceName don't have appName.

//代码占位符
private[spark] class LiveListenerBusMetrics(conf: SparkConf)
  extends Source with Logging {

  override val sourceName: String = "LiveListenerBus"
  override val metricRegistry: MetricRegistry = new MetricRegistry
...
}

Issue Links

• links to: [Github] Pull Request #37951 (Kwafoor)