[SPARK-19378][SS] Ensure continuity of stateOperator and eventTime metrics even if there is no new data in trigger

brkyvz · tdas · commit 081b7addaf95 · 2017-01-31T16:52:53.000-08:00
## What changes were proposed in this pull request? In StructuredStreaming, if a new trigger was skipped because no new data arrived, we suddenly report nothing for the metrics `stateOperator`. We could however easily report the metrics from `lastExecution` to ensure continuity of metrics. ## How was this patch tested? Regression test in `StreamingQueryStatusAndProgressSuite` Author: Burak Yavuz <brkyvz@gmail.com> Closes #16716 from brkyvz/state-agg.
diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/ProgressReporter.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/ProgressReporter.scala
@@ -180,15 +180,38 @@ trait ProgressReporter extends Logging {
     currentStatus = currentStatus.copy(isTriggerActive = false)
   }
 
+  /** Extract statistics about stateful operators from the executed query plan. */
+  private def extractStateOperatorMetrics(hasNewData: Boolean): Seq[StateOperatorProgress] = {
+    if (lastExecution == null) return Nil
+    // lastExecution could belong to one of the previous triggers if `!hasNewData`.
+    // Walking the plan again should be inexpensive.
+    val stateNodes = lastExecution.executedPlan.collect {
+      case p if p.isInstanceOf[StateStoreSaveExec] => p
+    }
+    stateNodes.map { node =>
+      val numRowsUpdated = if (hasNewData) {
+        node.metrics.get("numUpdatedStateRows").map(_.value).getOrElse(0L)
+      } else {
+        0L
+      }
+      new StateOperatorProgress(
+        numRowsTotal = node.metrics.get("numTotalStateRows").map(_.value).getOrElse(0L),
+        numRowsUpdated = numRowsUpdated)
+    }
+  }
+
   /** Extracts statistics from the most recent query execution. */
   private def extractExecutionStats(hasNewData: Boolean): ExecutionStats = {
     val hasEventTime = logicalPlan.collect { case e: EventTimeWatermark => e }.nonEmpty
     val watermarkTimestamp =
       if (hasEventTime) Map("watermark" -> formatTimestamp(offsetSeqMetadata.batchWatermarkMs))
       else Map.empty[String, String]
 
+    // SPARK-19378: Still report metrics even though no data was processed while reporting progress.
+    val stateOperators = extractStateOperatorMetrics(hasNewData)
+
     if (!hasNewData) {
-      return ExecutionStats(Map.empty, Seq.empty, watermarkTimestamp)
+      return ExecutionStats(Map.empty, stateOperators, watermarkTimestamp)
     }
 
     // We want to associate execution plan leaves to sources that generate them, so that we match
@@ -237,16 +260,6 @@ trait ProgressReporter extends Logging {
         Map.empty
       }
 
-    // Extract statistics about stateful operators in the query plan.
-    val stateNodes = lastExecution.executedPlan.collect {
-      case p if p.isInstanceOf[StateStoreSaveExec] => p
-    }
-    val stateOperators = stateNodes.map { node =>
-      new StateOperatorProgress(
-        numRowsTotal = node.metrics.get("numTotalStateRows").map(_.value).getOrElse(0L),
-        numRowsUpdated = node.metrics.get("numUpdatedStateRows").map(_.value).getOrElse(0L))
-    }
-
     val eventTimeStats = lastExecution.executedPlan.collect {
       case e: EventTimeWatermarkExec if e.eventTimeStats.value.count > 0 =>
         val stats = e.eventTimeStats.value
diff --git a/sql/core/src/test/scala/org/apache/spark/sql/streaming/StreamingQueryStatusAndProgressSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/streaming/StreamingQueryStatusAndProgressSuite.scala
@@ -20,16 +20,19 @@ package org.apache.spark.sql.streaming
 import java.util.UUID
 
 import scala.collection.JavaConverters._
+import scala.language.postfixOps
 
 import org.json4s._
 import org.json4s.jackson.JsonMethods._
+import org.scalatest.concurrent.Eventually
+import org.scalatest.time.SpanSugar._
 
 import org.apache.spark.sql.execution.streaming.MemoryStream
 import org.apache.spark.sql.functions._
+import org.apache.spark.sql.internal.SQLConf
 import org.apache.spark.sql.streaming.StreamingQueryStatusAndProgressSuite._
 
-
-class StreamingQueryStatusAndProgressSuite extends StreamTest {
+class StreamingQueryStatusAndProgressSuite extends StreamTest with Eventually {
   implicit class EqualsIgnoreCRLF(source: String) {
     def equalsIgnoreCRLF(target: String): Boolean = {
       source.replaceAll("\r\n|\r|\n", System.lineSeparator) ===
@@ -171,6 +174,41 @@ class StreamingQueryStatusAndProgressSuite extends StreamTest {
       query.stop()
     }
   }
+
+  test("SPARK-19378: Continue reporting stateOp metrics even if there is no active trigger") {
+    import testImplicits._
+
+    withSQLConf(SQLConf.STREAMING_NO_DATA_PROGRESS_EVENT_INTERVAL.key -> "10") {
+      val inputData = MemoryStream[Int]
+
+      val query = inputData.toDS().toDF("value")
+        .select('value)
+        .groupBy($"value")
+        .agg(count("*"))
+        .writeStream
+        .queryName("metric_continuity")
+        .format("memory")
+        .outputMode("complete")
+        .start()
+      try {
+        inputData.addData(1, 2)
+        query.processAllAvailable()
+
+        val progress = query.lastProgress
+        assert(progress.stateOperators.length > 0)
+        // Should emit new progresses every 10 ms, but we could be facing a slow Jenkins
+        eventually(timeout(1 minute)) {
+          val nextProgress = query.lastProgress
+          assert(nextProgress.timestamp !== progress.timestamp)
+          assert(nextProgress.numInputRows === 0)
+          assert(nextProgress.stateOperators.head.numRowsTotal === 2)
+          assert(nextProgress.stateOperators.head.numRowsUpdated === 0)
+        }
+      } finally {
+        query.stop()
+      }
+    }
+  }
 }
 
 object StreamingQueryStatusAndProgressSuite {
