Add outputSize and name to the Stage metadata

Add output size and details to stage

Author: msafonov

.github/workflows/ci.yml

@@ -22,7 +22,7 @@ jobs:
         uses: actions/checkout@v3
 
       - name: Validate Gradle Wrapper
-        uses: gradle/wrapper-validation-action@v1
+        uses: gradle/actions/wrapper-validation@v3
 
       - name: Configure JDK
         uses: actions/setup-java@v3

lib/build.gradle

@@ -66,7 +66,7 @@ publishing {
         maven(MavenPublication) {
             groupId = 'com.joom.spark'
             artifactId = "spark-platform_$scalaVersion"
-            version = '0.4.9'
+            version = '0.4.10'
 
             from components.java
 

lib/src/main/scala/com/joom/spark/monitoring/Parts.scala

@@ -34,10 +34,12 @@ case class StageSummary(
                          memorySpillGB: Double,
                          diskSpillGB: Double,
                          inputGB: Double,
+                         outputGB: Double,
                          shuffleWriteGB: Double,
                          peakExecutionMemoryGB: Double,
                          properties: Map[String, String],
                          endTs: Long,
+                         name: Option[String],
                        )
 
 case class ExecutorMetric(

lib/src/main/scala/com/joom/spark/monitoring/StatsReportingSparkListener.scala

@@ -47,7 +47,8 @@ class StatsReportingSparkListener(sparkConf: SparkConf, apiKey: String,
                         var sent: Boolean = false,
                         var startedTaskCount: Int = 0,
                         var failureReason: Option[String] = None,
-                        var properties: Map[String, String] = Map())
+                        var properties: Map[String, String] = Map(),
+                        var name: Option[String] = None)
   private val tasksPerStage = mutable.Map[StageFullId, mutable.ArrayBuffer[(TaskMetrics, TaskEndReason)]]()
   private val stageState = mutable.Map[StageFullId, StageState]()
   private val appStart: Instant = Instant.now()
@@ -155,7 +156,9 @@ class StatsReportingSparkListener(sparkConf: SparkConf, apiKey: String,
     }.toMap
 
     val startTime = stageSubmitted.stageInfo.submissionTime.map(Instant.ofEpochMilli).getOrElse(Instant.now())
-    stageState.getOrElseUpdate(stageFullId, StageState(startTime)).properties = properties
+    val state = stageState.getOrElseUpdate(stageFullId, StageState(startTime))
+    state.properties = properties
+    state.name = Some(stageSubmitted.stageInfo.name)
   }
 
   override def onStageCompleted(stageCompleted: SparkListenerStageCompleted): Unit = {
@@ -236,7 +239,7 @@ class StatsReportingSparkListener(sparkConf: SparkConf, apiKey: String,
         val failureReason = state.failureReason
         val startTime = state.startTime
         val summary = summarizeStage(appId, stageFullId.stageId, stageFullId.attemptNumber, success, failureReason,
-          startTime, tasks.toSeq, state.properties, endTs)
+          startTime, tasks.toSeq, state.properties, endTs, state.name)
 
         implicit val codec: JsonValueCodec[StageSummary] = JsonCodecMaker.make
         send("stages", summary.get)
@@ -342,7 +345,9 @@ object StatsReportingSparkListener {
                              failureReason: Option[String], startTime: Instant,
                              rawTaskMetrics: Seq[(TaskMetrics, TaskEndReason)],
                              properties: Map[String, String],
-                             endTime: Option[Long]): Option[StageSummary] = {
+                             endTime: Option[Long],
+                             name: Option[String],
+                            ): Option[StageSummary] = {
     val taskMetrics = rawTaskMetrics.map(_._1)
       .filter(_ != null) // For failed tasks, there will be 'null' TaskMetrics instances.
     val runTimes = taskMetrics.map(_.executorRunTime.toDouble / 1000.0)
@@ -375,10 +380,12 @@ object StatsReportingSparkListener {
       memorySpillGB = taskMetrics.map(_.memoryBytesSpilled.toDouble).sum / GiB,
       diskSpillGB = taskMetrics.map(_.diskBytesSpilled).sum / GiB,
       inputGB = taskMetrics.map(_.inputMetrics.bytesRead).sum / GiB,
+      outputGB = taskMetrics.map(_.outputMetrics.bytesWritten).sum / GiB,
       shuffleWriteGB = taskMetrics.map(_.shuffleWriteMetrics.bytesWritten).sum / GiB,
       peakExecutionMemoryGB = taskMetrics.map(_.peakExecutionMemory).sum / GiB,
       properties = properties,
       endTs = endTs,
+      name = name,
     ))
   }
 }