From a3d9d75434b9f3d7d32c8f2bdcd3f5787075531a Mon Sep 17 00:00:00 2001
From: Partho Sarthi <psarthi@nvidia.com>
Date: Wed, 8 Apr 2026 10:22:15 -0700
Subject: [PATCH 01/16] Enhance profiling metrics with new OOM detection

Signed-off-by: Partho Sarthi <psarthi@nvidia.com>
---
 .../configs/reports/coreRawMetricsReport.yaml |  18 ++-
 .../profiling/ApplicationSummaryInfo.scala    | 140 +++++++++++-------
 .../tool/profiling/CollectInformation.scala   |  17 ++-
 .../rapids/tool/profiling/ProfileArgs.scala   |   2 +-
 .../profiling/ProfileClassWarehouse.scala     |  51 ++++++-
 .../rapids/tool/profiling/Profiler.scala      |  32 ++--
 .../spark/rapids/tool/tuning/AutoTuner.scala  |   4 +-
 .../rapids/tool/views/OutHeaderRegistry.scala |   4 +-
 .../tool/views/QualRawReportGenerator.scala   |  21 ++-
 .../tool/profiling/OomDetectionSuite.scala    |  62 ++++++++
 .../tool/tuning/BaseAutoTunerSuite.scala      |  12 +-
 .../tool/tuning/ProfilingAutoTunerSuite.scala |   7 +-
 .../tuning/ProfilingAutoTunerSuiteV2.scala    |   4 +-
 13 files changed, 283 insertions(+), 91 deletions(-)
 create mode 100644 core/src/test/scala/com/nvidia/spark/rapids/tool/profiling/OomDetectionSuite.scala

diff --git a/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml b/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
index d9e4b1b6a..a89482665 100644
--- a/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
+++ b/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
@@ -1,4 +1,4 @@
-# Copyright (c) 2025, NVIDIA CORPORATION.
+# Copyright (c) 2025-2026, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -106,6 +106,22 @@ reportDefinitions:
             dataType: Long
             description: >-
               TBD
+          - name: maxTaskInputBytesRead
+            dataType: Double
+            description: >-
+              Maximum per-task input bytes read across all SQLs
+          - name: maxColumnarExchangeDataSizeBytes
+            dataType: Long
+            description: >-
+              Maximum data size from ColumnarExchange metrics (profiling only)
+          - name: scanStagesWithGpuOom
+            dataType: String
+            description: >-
+              Comma-separated stage IDs of scan stages with GPU OOM (profiling only)
+          - name: shuffleStagesWithOom
+            dataType: String
+            description: >-
+              Comma-separated stage IDs of shuffle stages with OOM (profiling only)
       # AppLogPathProfileResults
       - label: coreRawApplicationLogPathMappingCSV
         description: >-
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
index 783e2fc50..4adefb272 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2021-2025, NVIDIA CORPORATION.
+ * Copyright (c) 2021-2026, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -90,8 +90,8 @@ trait AppInfoReadMetrics {
 }
 
 trait AppInfoGpuOomCheck {
-  def hasScanStagesWithGpuOom: Boolean = false
-  def hasShuffleStagesWithOom: Boolean = false
+  def scanStagesWithGpuOom: Set[Long] = Set.empty
+  def shuffleStagesWithOom: Set[Long] = Set.empty
 }
 
 trait AppInfoColumnarExchangeMetrics {
@@ -230,65 +230,112 @@ class SingleAppSummaryInfoProvider(
   }
 
   /**
-   * Check if there are any scan stages with failed tasks due to GPU OOM errors
+   * Returns stage IDs of scan stages with failed tasks due to GPU OOM errors
    * (GpuRetryOOM and GpuSplitAndRetryOOM).
    */
-  override def hasScanStagesWithGpuOom: Boolean = {
-    // If the plugin is not enabled (i.e. non-GPU app), return false
-    if (!app.appInfo.exists(_.pluginEnabled)) {
-      return false
+  override def scanStagesWithGpuOom: Set[Long] = {
+    SingleAppSummaryInfoProvider.computeScanStagesWithGpuOom(
+      app.appInfo.exists(_.pluginEnabled),
+      app.failedTasks, app.stageMetrics, appInfo)
+  }
+
+  /**
+   * Returns stage IDs of failed shuffle stages with OOM errors in the task's end reason.
+   * Note: This check is enabled only if the plugin is enabled (i.e. GPU app) and running on YARN.
+   */
+  override def shuffleStagesWithOom: Set[Long] = {
+    SingleAppSummaryInfoProvider.computeShuffleStagesWithOom(
+      app.appInfo.exists(_.pluginEnabled),
+      getSparkProperty("spark.master"),
+      app.failedStages, app.failedTasks)
+  }
+
+  /**
+   * Get the maximum data size from ColumnarExchange metrics.
+   * This method searches through SQLPlan metrics to find all ColumnarExchange nodes
+   * with "data size" metrics and returns the maximum total value.
+   *
+   * @return Option[Long] containing the maximum data size in bytes, or None if no
+   *         ColumnarExchange "data size" metrics are found
+   */
+  override def getMaxColumnarExchangeDataSizeBytes: Option[Long] = {
+    SingleAppSummaryInfoProvider.computeMaxColumnarExchangeDataSizeBytes(app.sqlMetrics)
+  }
+
+  override def getClassPathEntries: Map[String, String] = {
+    appInfo.classpathEntries
+  }
+}
+
+object SingleAppSummaryInfoProvider {
+  /**
+   * Computes the set of scan stage IDs that had GPU OOM failures.
+   * This static method enables computing the result before ApplicationSummaryInfo is assembled.
+   */
+  def computeScanStagesWithGpuOom(
+      pluginEnabled: Boolean,
+      failedTasks: Seq[FailedTaskProfileResults],
+      stageMetrics: Seq[AccumProfileResults],
+      appInfo: ApplicationInfo): Set[Long] = {
+    if (!pluginEnabled) {
+      return Set.empty
     }
 
     // Find stages with failed tasks due to GPU OOM errors
-    val failedStagesWithGpuOom = app.failedTasks.collect {
-      case task if SparkRapidsOomExceptions.gpuExceptionClassNames
-        .exists(task.endReason.contains) => task.stageId
+    val failedStagesWithGpuOom = failedTasks.collect {
+      case task if SparkRapidsOomExceptions.isGpuOom(task.endReason) => task.stageId.toLong
     }
 
     if (failedStagesWithGpuOom.isEmpty) {
-      return false
+      return Set.empty
     }
 
     // Calculate stageIds of scan stages (i.e. stages with 'scan time' metrics)
-    val scanStages = app.stageMetrics.collect {
+    val scanStages = stageMetrics.collect {
       case metric if IoMetrics.getIoMetricsHelper(appInfo).isScanTimeMetric(metric) =>
-        metric.stageId
+        metric.stageId.toLong
     }.toSet
 
     if (scanStages.isEmpty) {
-      return false
+      return Set.empty
     }
 
-    // Check if any failed GPU OOM stage is also a scan stage
-    failedStagesWithGpuOom.exists(scanStages.contains)
+    // Return scan stages that also had GPU OOM failures
+    failedStagesWithGpuOom.filter(scanStages.contains).toSet
   }
 
   /**
-   * This method checks for failed shuffle stages with OOM errors in the task's end reason.
-   * Note: This check is enabled only if the plugin is enabled (i.e. GPU app) and running on YARN.
-   * """
+   * Computes the set of shuffle stage IDs that had container OOM failures (YARN only).
+   * Detects ExecutorLostFailure with exit code 137 (SIGKILL from container memory enforcement).
+   * See: https://github.com/NVIDIA/spark-rapids-tools/issues/1566
+   * This static method enables computing the result before ApplicationSummaryInfo is assembled.
    */
-  override def hasShuffleStagesWithOom: Boolean = {
-    // If the plugin is not enabled (i.e. non-GPU app) or not running on YARN, return false
-    val sparkMaster = SparkMaster(getSparkProperty("spark.master"))
-    if (!app.appInfo.exists(_.pluginEnabled) || !sparkMaster.contains(Yarn)) {
-      return false
+  def computeShuffleStagesWithOom(
+      pluginEnabled: Boolean,
+      sparkMasterStr: Option[String],
+      failedStages: Seq[FailedStagesProfileResults],
+      failedTasks: Seq[FailedTaskProfileResults]): Set[Long] = {
+    if (!pluginEnabled || !SparkMaster(sparkMasterStr).contains(Yarn)) {
+      return Set.empty
     }
 
     // Get stage IDs of failed shuffle stages
     // Sample stage name: "submitShuffleJob$ at GpuShuffleExchangeExec.scala:53"
-    val failedStagesWithShuffle = app.failedStages.collect {
+    val failedStagesWithShuffle = failedStages.collect {
       case stage if stage.name.contains(SparkRapidsOomExceptions.gpuShuffleClassName) =>
-        stage.stageId
+        stage.stageId.toLong
     }.toSet
 
     if (failedStagesWithShuffle.isEmpty) {
-      return false
+      return Set.empty
     }
 
     // scalastyle:off line.size.limit
     // Check if the failed task's end reason contains OOM errors
-    // Sample end reason for failed tasks on YARN: "ExecutorLostFailure (executor 2 exited caused by one of the running tasks) Reason: Container from a bad node: container_e02_17xxx on host: test-cluster-w-0. Exit status: 137"
+    // Sample end reason for failed tasks on YARN:
+    // "ExecutorLostFailure (executor 2 exited caused by one of the running tasks)
+    //  Reason: Container from a bad node: container_e02_17xxx on host: test-cluster-w-0.
+    //  Exit status: 137"
     // Reference: https://github.com/apache/spark/blob/master/resource-managers/yarn/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocator.scala
     // scalastyle:on line.size.limit
     // Regular expressions to identify OOM failures in task's end reason
@@ -299,30 +346,23 @@ class SingleAppSummaryInfoProvider(
       s"Exit status: ${UnixExitCode.FORCE_KILLED}"
     ).map(_.r)
 
-    // Check if any failed task in shuffle stages have OOM failures
-    app.failedTasks.exists { task =>
-      if (failedStagesWithShuffle.contains(task.stageId)) {
-        // Check if the task failed due to OOM
-        oomFailurePatterns.forall(p => p.findFirstIn(task.endReason).isDefined)
-      } else {
-        // Ignore if the failed task is not in a shuffle stage
-        false
-      }
-    }
+    // Return shuffle stages that had tasks with OOM failures
+    failedTasks.collect {
+      case task if failedStagesWithShuffle.contains(task.stageId.toLong) &&
+        oomFailurePatterns.forall(p => p.findFirstIn(task.endReason).isDefined) =>
+        task.stageId.toLong
+    }.toSet
   }
 
   /**
-   * Get the maximum data size from ColumnarExchange metrics.
-   * This method searches through SQLPlan metrics to find all ColumnarExchange nodes
-   * with "data size" metrics and returns the maximum total value.
-   *
-   * @return Option[Long] containing the maximum data size in bytes, or None if no
-   *         ColumnarExchange "data size" metrics are found
+   * Computes the maximum data size from ColumnarExchange metrics.
+   * This static method enables computing the result before ApplicationSummaryInfo is assembled.
    */
-  override def getMaxColumnarExchangeDataSizeBytes: Option[Long] = {
-    val columnarExchangeDataSizesBytes = app.sqlMetrics.collect {
+  def computeMaxColumnarExchangeDataSizeBytes(
+      sqlMetrics: Seq[SQLAccumProfileResults]): Option[Long] = {
+    val columnarExchangeDataSizesBytes = sqlMetrics.collect {
       case metric if metric.nodeName.contains("ColumnarExchange") &&
-                      metric.name == "data size" =>
+        metric.name == "data size" =>
         metric.total
     }
     if (columnarExchangeDataSizesBytes.nonEmpty) {
@@ -331,8 +371,4 @@ class SingleAppSummaryInfoProvider(
       None
     }
   }
-
-  override def getClassPathEntries: Map[String, String] = {
-    appInfo.classpathEntries
-  }
 }
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/CollectInformation.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/CollectInformation.scala
index 1addbbb48..8b165733f 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/CollectInformation.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/CollectInformation.scala
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2021-2025, NVIDIA CORPORATION.
+ * Copyright (c) 2021-2026, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -37,6 +37,21 @@ class CollectInformation(apps: Seq[ApplicationInfo]) extends Logging {
     ProfInformationView.getRawView(apps)
   }
 
+  // Extends AppInfoProfileResults with fields that require aggregate metrics or
+  // cross-referencing (e.g., failed tasks with stage metrics). Fields like
+  // totalCoreSeconds are already populated at the view layer in InformationView.
+  def getExtendedAppInfo(
+      maxTaskInputBytesRead: Double,
+      maxColumnarExchangeDataSizeBytes: Option[Long],
+      scanStagesWithGpuOom: Set[Long],
+      shuffleStagesWithOom: Set[Long]): Seq[AppInfoProfileResults] = {
+    getAppInfo.map(_.copy(
+      maxTaskInputBytesRead = maxTaskInputBytesRead,
+      maxColumnarExchangeDataSizeBytes = maxColumnarExchangeDataSizeBytes,
+      scanStagesWithGpuOom = scanStagesWithGpuOom,
+      shuffleStagesWithOom = shuffleStagesWithOom))
+  }
+
   def getAppLogPath: Seq[AppLogPathProfileResults] = {
     ProfLogPathView.getRawView(apps)
   }
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala
index 81a253ebf..a660c4ac6 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala
@@ -89,7 +89,7 @@ Usage: java -cp rapids-4-spark-tools_2.12-<version>.jar:$SPARK_HOME/jars/*
       descr = "Number of thread to use for parallel processing. The default is the " +
         "number of cores on host divided by 4.")
   val csv: ScallopOption[Boolean] =
-    opt[Boolean](required = false,
+    opt[Boolean](required = false, default = Some(true),
       descr = "Output each table to a CSV file as well creating the summary text file.")
   val timeout: ScallopOption[Long] =
     opt[Long](required = false,
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
index 7bde04197..8a0422d0e 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
@@ -463,7 +463,11 @@ case class AppInfoProfileResults(
     sparkRuntime: SparkRuntime.SparkRuntime,
     sparkVersion: String,
     pluginEnabled: Boolean,
-    totalCoreSeconds: Long)  extends ProfileResult {
+    totalCoreSeconds: Long,
+    maxTaskInputBytesRead: Double = 0.0,
+    maxColumnarExchangeDataSizeBytes: Option[Long] = None,
+    scanStagesWithGpuOom: Set[Long] = Set.empty,
+    shuffleStagesWithOom: Set[Long] = Set.empty)  extends ProfileResult {
   override def outputHeaders: Array[String] = {
     OutHeaderRegistry.outputHeaders("AppInfoProfileResults")
   }
@@ -496,11 +500,23 @@ case class AppInfoProfileResults(
     }
   }
 
+  private def maxColumnarExchangeToStr: String = {
+    maxColumnarExchangeDataSizeBytes.map(_.toString).getOrElse("")
+  }
+
+  private def stageIdsToStr(stageIds: Set[Long]): String = {
+    if (stageIds.isEmpty) "" else stageIds.toSeq.sorted.mkString(",")
+  }
+
   override def convertToSeq(): Array[String] = {
     Array(appName, appIdToStr, attemptIdToStr,
       sparkUser, startTime.toString, endTimeToStr, durToStr,
       durationStr, sparkRuntime.toString, sparkVersion, pluginEnabled.toString,
-      totalCoreSeconds.toString)
+      totalCoreSeconds.toString,
+      f"$maxTaskInputBytesRead%.0f",
+      maxColumnarExchangeToStr,
+      stageIdsToStr(scanStagesWithGpuOom),
+      stageIdsToStr(shuffleStagesWithOom))
   }
 
   override def convertToCSVSeq(): Array[String] = {
@@ -512,7 +528,11 @@ case class AppInfoProfileResults(
       StringUtils.reformatCSVString(sparkRuntime.toString),
       StringUtils.reformatCSVString(sparkVersion),
       pluginEnabled.toString,
-      totalCoreSeconds.toString)
+      totalCoreSeconds.toString,
+      f"$maxTaskInputBytesRead%.0f",
+      maxColumnarExchangeToStr,
+      StringUtils.reformatCSVString(stageIdsToStr(scanStagesWithGpuOom)),
+      StringUtils.reformatCSVString(stageIdsToStr(shuffleStagesWithOom)))
   }
 }
 
@@ -1499,15 +1519,32 @@ case class RecommendedCommentResult(comment: String) {
   override def toString: String = "- %s".format(comment)
 }
 
+// scalastyle:off line.size.limit
 /**
- * Helper object to store the list of SparkRapids OOM exceptions.
+ * Helper object to detect OOM exceptions from SparkRapids event logs.
+ *
+ * GPU OOM class names from spark-rapids-jni:
+ * - GpuOOM -> GpuRetryOOM, GpuSplitAndRetryOOM
+ * See: https://github.com/NVIDIA/spark-rapids-jni/blob/725cd64be2115cd072bf51d7d6c5281d6d08bf4f/src/main/cpp/src/SparkResourceAdaptorJni.cpp#L1313
+ * See: https://github.com/NVIDIA/spark-rapids/blob/79922d62a1c5759963e969018322ad8e544629ff/sql-plugin/src/main/scala/com/nvidia/spark/rapids/RmmRapidsRetryIterator.scala
  */
+// scalastyle:on line.size.limit
 object SparkRapidsOomExceptions {
-  val gpuExceptionClassNames: Set[String] = {
-    Set("GpuSplitAndRetryOOM", "GpuRetryOOM")
-  }
+  // Current JNI: GpuOOM -> GpuRetryOOM, GpuSplitAndRetryOOM
+  // Pre-24.02 JNI: jni.SplitAndRetryOOM, jni.RetryOOM (no Gpu prefix)
+  // Using "jni." prefix to avoid matching CpuSplitAndRetryOOM
+  val gpuExceptionClassNames: Set[String] =
+    Set("GpuSplitAndRetryOOM", "GpuRetryOOM", "GpuOOM",
+        "jni.SplitAndRetryOOM", "jni.RetryOOM")
 
   val gpuShuffleClassName: String = "GpuShuffleExchangeExec"
+
+  /**
+   * Check if a failure reason indicates a GPU OOM error.
+   */
+  def isGpuOom(failureReason: String): Boolean = {
+    gpuExceptionClassNames.exists(failureReason.contains)
+  }
 }
 
 /**
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
index 1534fbf2b..3b7bd878e 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
@@ -293,19 +293,33 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
       }
     }
     val analysis = RawMetricProfilerView.getAggMetrics(analyzedApps)
-    val maxTaskInputInfo = if (useAutoTuner) {
-      analysis.maxTaskInputSizes
-    } else {
-      Seq.empty
-    }
+    val maxTaskInputInfo = analysis.maxTaskInputSizes
     val sqlIdAlign = if (outputAlignedSQLIds) {
       collect.getSQLCleanAndAligned
     } else {
       Seq.empty
     }
     val endTime = System.currentTimeMillis()
-    val appInfo = collect.getAppInfo
     val sqlMetrics = collect.getSQLPlanMetrics
+    val stageMetrics = collect.getStageLevelMetrics
+    val failedTasks = healthCheck.getFailedTasks
+    val failedStages = healthCheck.getFailedStages
+
+    // Compute AutoTuner inputs to enrich application_information.csv
+    val singleApp = analyzedApps.head
+    val pluginEnabled = singleApp.gpuMode
+    val maxTaskInput = analysis.maxTaskInputSizes.headOption
+      .map(_.maxTaskInputBytesRead).getOrElse(0.0)
+    val maxColumnarExchange =
+      SingleAppSummaryInfoProvider.computeMaxColumnarExchangeDataSizeBytes(sqlMetrics)
+    val scanOomStages = SingleAppSummaryInfoProvider.computeScanStagesWithGpuOom(
+      pluginEnabled, failedTasks, stageMetrics, singleApp)
+    val shuffleOomStages = SingleAppSummaryInfoProvider.computeShuffleStagesWithOom(
+      pluginEnabled, singleApp.sparkProperties.get("spark.master"),
+      failedStages, failedTasks)
+
+    val appInfo = collect.getExtendedAppInfo(
+      maxTaskInput, maxColumnarExchange, scanOomStages, shuffleOomStages)
     logDebug(s"Time to collect Profiling Info [${appInfo.head.appId}]: ${endTime - startTime}.")
     val appInfoSummary = ApplicationSummaryInfo(
       appInfo = appInfo,
@@ -315,14 +329,14 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
       rapidsProps = collect.getRapidsProperties,
       rapidsJar = collect.getRapidsJARInfo,
       sqlMetrics = sqlMetrics,
-      stageMetrics = collect.getStageLevelMetrics,
+      stageMetrics = stageMetrics,
       jobAggMetrics = analysis.jobAggs,
       stageAggMetrics = analysis.stageAggs,
       sqlTaskAggMetrics = analysis.sqlAggs,
       durAndCpuMet = analysis.sqlDurAggs,
       skewInfo = analysis.taskShuffleSkew,
-      failedTasks = healthCheck.getFailedTasks,
-      failedStages = healthCheck.getFailedStages,
+      failedTasks = failedTasks,
+      failedStages = failedStages,
       failedJobs = healthCheck.getFailedJobs,
       removedBMs = healthCheck.getRemovedBlockManager,
       removedExecutors = healthCheck.getRemovedExecutors,
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/tuning/AutoTuner.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/tuning/AutoTuner.scala
index 90800289e..1eb8dd56e 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/tuning/AutoTuner.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/tuning/AutoTuner.scala
@@ -2177,7 +2177,7 @@ class ProfilingAutoTuner(
     // First, calculate the recommendation based on input sizes
     val calculatedValueFromInputSize = super.calculateMaxPartitionBytesInMB(maxPartitionBytes)
     getPropertyValue("spark.sql.files.maxPartitionBytes") match {
-      case Some(currentValue) if appInfoProvider.hasScanStagesWithGpuOom =>
+      case Some(currentValue) if appInfoProvider.scanStagesWithGpuOom.nonEmpty =>
         // GPU OOM detected. We may want to reduce max partition size.
         val halvedValue = StringUtils.convertToMB(currentValue, Some(ByteUnit.BYTE)) / 2
         // Choose the minimum between the calculated value and half of the current value.
@@ -2198,7 +2198,7 @@ class ProfilingAutoTuner(
    */
   override def recommendShufflePartitionsInternal(): Int = {
     val calculatedValue = super.recommendShufflePartitionsInternal()
-    if (appInfoProvider.hasShuffleStagesWithOom) {
+    if (appInfoProvider.shuffleStagesWithOom.nonEmpty) {
       // Shuffle Stages with Task OOM detected. We may want to increase shuffle partitions.
       val recShufflePartitions = shufflePartitionValue *
         configProvider.getEntry("SHUFFLE_PARTITION_MULTIPLIER").getDefault.toInt
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
index 795de4c6a..4a0bd37be 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
@@ -69,7 +69,9 @@ object OutHeaderRegistry {
       Array("sqlID", "nodeID", "nodeName", "nodeDescription", "reason"),
     "AppInfoProfileResults" ->
       Array("appName", "appId", "attemptId", "sparkUser", "startTime", "endTime", "duration",
-        "durationStr", "sparkRuntime", "sparkVersion", "pluginEnabled", "totalCoreSeconds"),
+        "durationStr", "sparkRuntime", "sparkVersion", "pluginEnabled", "totalCoreSeconds",
+        "maxTaskInputBytesRead", "maxColumnarExchangeDataSizeBytes",
+        "scanStagesWithGpuOom", "shuffleStagesWithOom"),
     "AppLogPathProfileResults" ->
       Array("appName", "appId", "eventLogPath"),
     "FailedTaskProfileResults" ->
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
index 604fbfa7b..a9e146ed6 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2024-2025, NVIDIA CORPORATION.
+ * Copyright (c) 2024-2026, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -71,9 +71,19 @@ object QualRawReportGenerator extends Logging {
     val pWriter =
       new ProfileOutputWriter(metricsDirectory, "profile", 10000000, outputCSV = true)
     try {
+      // Compute aggregate metrics early so maxTaskInputBytesRead is available for
+      // application_information.csv enrichment
+      val aggRawMetrics = QualSparkMetricsAggregator
+        .getAggRawMetrics(app, sqlAnalyzer = Some(sqlPlanAnalyzer))
+      val maxTaskInput = aggRawMetrics.maxTaskInputSizes.headOption
+        .map(_.maxTaskInputBytesRead).getOrElse(0.0)
+
       pWriter.writeText("### A. Information Collected ###")
-      pWriter.writeTable(
-        QualInformationView.getLabel, QualInformationView.getRawView(Seq(app)))
+      // Extend application info with maxTaskInputBytesRead from aggregate metrics.
+      // OOM and ColumnarExchange columns are empty for qualification (CPU event logs).
+      val extendedAppInfo = QualInformationView.getRawView(Seq(app)).map(
+        _.copy(maxTaskInputBytesRead = maxTaskInput))
+      pWriter.writeTable(QualInformationView.getLabel, extendedAppInfo)
       pWriter.writeTable(QualLogPathView.getLabel, QualLogPathView.getRawView(Seq(app)))
       val sqlPlanMetricsResults = generateSQLProcessingView(pWriter, sqlPlanAnalyzer)
       pWriter.writeJsonL(
@@ -96,9 +106,8 @@ object QualRawReportGenerator extends Logging {
         SystemQualPropertiesView.getRawView(Seq(app)),
         Some(SystemQualPropertiesView.getDescription))
       pWriter.writeText("\n### B. Analysis ###\n")
-      constructLabelsMaps(QualSparkMetricsAggregator
-        .getAggRawMetrics(
-          app, sqlAnalyzer = Some(sqlPlanAnalyzer))).foreach { case (label, metrics) =>
+      // Reuse already-computed aggRawMetrics instead of computing again
+      constructLabelsMaps(aggRawMetrics).foreach { case (label, metrics) =>
           pWriter.writeCSVTable(label, metrics)
       }
       pWriter.writeText("\n### C. Health Check###\n")
diff --git a/core/src/test/scala/com/nvidia/spark/rapids/tool/profiling/OomDetectionSuite.scala b/core/src/test/scala/com/nvidia/spark/rapids/tool/profiling/OomDetectionSuite.scala
new file mode 100644
index 000000000..c5f01f22a
--- /dev/null
+++ b/core/src/test/scala/com/nvidia/spark/rapids/tool/profiling/OomDetectionSuite.scala
@@ -0,0 +1,62 @@
+/*
+ * Copyright (c) 2026, NVIDIA CORPORATION.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package com.nvidia.spark.rapids.tool.profiling
+
+import org.scalatest.funsuite.AnyFunSuite
+
+class OomDetectionSuite extends AnyFunSuite {
+
+  // (description, failureReason, expectedIsGpuOom)
+  private val gpuOomTestCases = Seq(
+    ("GpuSplitAndRetryOOM",
+      "com.nvidia.spark.rapids.jni.GpuSplitAndRetryOOM: " +
+        "GPU OutOfMemory: a batch of 1 cannot be split!",
+      true),
+    ("GpuRetryOOM",
+      "com.nvidia.spark.rapids.jni.GpuRetryOOM: GPU OutOfMemory",
+      true),
+    ("GpuOOM base class",
+      "com.nvidia.spark.rapids.jni.GpuOOM: GPU OutOfMemory",
+      true),
+    ("pre-24.02 jni.SplitAndRetryOOM (no Gpu prefix)",
+      "com.nvidia.spark.rapids.jni.SplitAndRetryOOM: " +
+        "GPU OutOfMemory: a batch of 1 cannot be split!",
+      true),
+    ("pre-24.02 jni.RetryOOM (no Gpu prefix)",
+      "com.nvidia.spark.rapids.jni.RetryOOM: GPU OutOfMemory",
+      true),
+    ("CpuSplitAndRetryOOM should not match",
+      "com.nvidia.spark.rapids.jni.CpuSplitAndRetryOOM: " +
+        "CPU OutOfMemory",
+      false),
+    ("CpuRetryOOM should not match",
+      "com.nvidia.spark.rapids.jni.CpuRetryOOM: CPU OutOfMemory",
+      false),
+    ("NullPointerException should not match",
+      "java.lang.NullPointerException: some error",
+      false),
+    ("ExecutorLostFailure should not match",
+      "ExecutorLostFailure (executor 5 exited) Exit status: 137",
+      false)
+  )
+
+  gpuOomTestCases.foreach { case (desc, reason, expected) =>
+    test(s"isGpuOom: $desc") {
+      assert(SparkRapidsOomExceptions.isGpuOom(reason) === expected)
+    }
+  }
+}
diff --git a/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/BaseAutoTunerSuite.scala b/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/BaseAutoTunerSuite.scala
index 9279625f4..ceffc08c7 100644
--- a/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/BaseAutoTunerSuite.scala
+++ b/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/BaseAutoTunerSuite.scala
@@ -49,8 +49,8 @@ class AppInfoProviderMockTest(val maxInput: Double,
     val meanShuffleRead: Double,
     val shuffleStagesWithPosSpilling: Set[Long],
     val shuffleSkewStages: Set[Long],
-    val scanStagesWithGpuOom: Boolean,
-    val shuffleStagesWithOom: Boolean,
+    val scanStagesWithGpuOomSet: Set[Long],
+    val shuffleStagesWithOomSet: Set[Long],
     val maxColumnarExchangeDataSizeBytes: Option[Long] = None)
     extends BaseProfilingAppSummaryInfoProvider {
   override def isAppInfoAvailable = true
@@ -69,8 +69,8 @@ class AppInfoProviderMockTest(val maxInput: Double,
   override def getRedundantReadSize: Long = redundantReadSize
   override def getShuffleStagesWithPosSpilling: Set[Long] = shuffleStagesWithPosSpilling
   override def getShuffleSkewStages: Set[Long] = shuffleSkewStages
-  override def hasScanStagesWithGpuOom: Boolean = scanStagesWithGpuOom
-  override def hasShuffleStagesWithOom: Boolean = shuffleStagesWithOom
+  override def scanStagesWithGpuOom: Set[Long] = scanStagesWithGpuOomSet
+  override def shuffleStagesWithOom: Set[Long] = shuffleStagesWithOomSet
   override def getMaxColumnarExchangeDataSizeBytes: Option[Long] = maxColumnarExchangeDataSizeBytes
 
   /**
@@ -135,8 +135,8 @@ abstract class BaseAutoTunerSuite extends AnyFunSuite with BeforeAndAfterEach
       meanShuffleRead: Double = 0.0,
       shuffleStagesWithPosSpilling: Set[Long] = Set(),
       shuffleSkewStages: Set[Long] = Set(),
-      scanStagesWithGpuOom: Boolean = false,
-      shuffleStagesWithOom: Boolean = false,
+      scanStagesWithGpuOom: Set[Long] = Set(),
+      shuffleStagesWithOom: Set[Long] = Set(),
       maxColumnarExchangeDataSizeBytes: Option[Long] = None): AppInfoProviderMockTest = {
     new AppInfoProviderMockTest(maxInput, spilledMetrics, jvmGCFractions, propsFromLog,
       sparkVersion, rapidsJars, distinctLocationPct, redundantReadSize, meanInput, meanShuffleRead,
diff --git a/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/ProfilingAutoTunerSuite.scala b/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/ProfilingAutoTunerSuite.scala
index 140399825..9ed73e57d 100644
--- a/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/ProfilingAutoTunerSuite.scala
+++ b/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/ProfilingAutoTunerSuite.scala
@@ -3136,7 +3136,8 @@ class ProfilingAutoTunerSuite extends ProfilingAutoTunerSuiteBase {
           "spark.plugins" -> "com.nvidia.spark.SQLPlugin",
           "spark.rapids.sql.concurrentGpuTasks" -> "4")
       val infoProvider = getMockInfoProvider(0, Seq(0), Seq(0.0),
-        logEventsProps, Some(testSparkVersion), scanStagesWithGpuOom = hasGpuOOm)
+        logEventsProps, Some(testSparkVersion),
+        scanStagesWithGpuOom = if (hasGpuOOm) Set(1L) else Set.empty)
       val platform = PlatformFactory.createInstance(PlatformNames.DATAPROC)
 
       // Configure cluster info using Platform's existing method
@@ -3288,7 +3289,7 @@ class ProfilingAutoTunerSuite extends ProfilingAutoTunerSuiteBase {
         "spark.plugins" -> "com.nvidia.spark.SQLPlugin",
         "spark.rapids.sql.concurrentGpuTasks" -> "4")
     val infoProvider = getMockInfoProvider(0, Seq(0), Seq(0.0),
-      logEventsProps, Some(testSparkVersion), shuffleStagesWithOom = true)
+      logEventsProps, Some(testSparkVersion), shuffleStagesWithOom = Set(1L))
     val platform = PlatformFactory.createInstance(PlatformNames.DATAPROC)
 
     // Configure cluster info using Platform's existing method
@@ -3376,7 +3377,7 @@ class ProfilingAutoTunerSuite extends ProfilingAutoTunerSuiteBase {
         "spark.plugins" -> "com.nvidia.spark.SQLPlugin",
         "spark.rapids.sql.concurrentGpuTasks" -> "4")
     val infoProvider = getMockInfoProvider(0, Seq(0), Seq(0.0),
-      logEventsProps, Some(testSparkVersion), shuffleStagesWithOom = true,
+      logEventsProps, Some(testSparkVersion), shuffleStagesWithOom = Set(1L),
       meanInput = 50000, meanShuffleRead = 80000)
     val platform = PlatformFactory.createInstance(PlatformNames.DATAPROC)
 
diff --git a/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/ProfilingAutoTunerSuiteV2.scala b/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/ProfilingAutoTunerSuiteV2.scala
index 834b95480..db47aefe1 100644
--- a/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/ProfilingAutoTunerSuiteV2.scala
+++ b/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/ProfilingAutoTunerSuiteV2.scala
@@ -1692,8 +1692,8 @@ class ProfilingAutoTunerSuiteV2 extends ProfilingAutoTunerSuiteBase {
       redundantReadSize = 0L,
       shuffleStagesWithPosSpilling = Set.empty,
       shuffleSkewStages = Set.empty,
-      scanStagesWithGpuOom = false,
-      shuffleStagesWithOom = false,
+      scanStagesWithGpuOom = Set.empty,
+      shuffleStagesWithOom = Set.empty,
       maxColumnarExchangeDataSizeBytes = Some(1000L * 1024 * 1024 * 1024) // 1000GB
     )
     val platform = PlatformFactory.createInstance(PlatformNames.DATAPROC, Some(targetClusterInfo))

From 091ae29fa1cb7266fac88a930648a23ad7325544 Mon Sep 17 00:00:00 2001
From: Partho Sarthi <psarthi@nvidia.com>
Date: Wed, 8 Apr 2026 10:23:33 -0700
Subject: [PATCH 02/16] Update copyright years in ProfileArgs.scala and
 OutHeaderRegistry.scala to reflect 2026

---
 .../com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala    | 2 +-
 .../com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala  | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala
index a660c4ac6..076a50dd3 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2021-2025, NVIDIA CORPORATION.
+ * Copyright (c) 2021-2026, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
index 4a0bd37be..cf9b0e2bb 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2025, NVIDIA CORPORATION.
+ * Copyright (c) 2025-2026, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.

From e8a6830603bd2acdba0f07bdfb0169f3c313e610 Mon Sep 17 00:00:00 2001
From: Partho Sarthi <psarthi@nvidia.com>
Date: Wed, 8 Apr 2026 14:58:38 -0700
Subject: [PATCH 03/16] Update data types and improve OOM error handling in
 profiling metrics

- Changed data type of `maxTaskInputBytesRead` from Double to Long in `coreRawMetricsReport.yaml`.
- Refactored methods in `ApplicationSummaryInfo.scala` to read pre-computed values for GPU OOM error handling, enhancing performance and reducing duplicate computations.
- Updated comments in `ProfileClassWarehouse.scala` for clarity on GPU OOM exception handling.

These changes aim to improve the accuracy and efficiency of profiling metrics related to GPU memory management.
---
 .../configs/reports/coreRawMetricsReport.yaml |  2 +-
 .../profiling/ApplicationSummaryInfo.scala    | 26 +++++++------------
 .../profiling/ProfileClassWarehouse.scala     |  5 ++--
 3 files changed, 13 insertions(+), 20 deletions(-)

diff --git a/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml b/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
index a89482665..eee4a919a 100644
--- a/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
+++ b/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
@@ -107,7 +107,7 @@ reportDefinitions:
             description: >-
               TBD
           - name: maxTaskInputBytesRead
-            dataType: Double
+            dataType: Long
             description: >-
               Maximum per-task input bytes read across all SQLs
           - name: maxColumnarExchangeDataSizeBytes
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
index 4adefb272..8be679802 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
@@ -230,36 +230,28 @@ class SingleAppSummaryInfoProvider(
   }
 
   /**
-   * Returns stage IDs of scan stages with failed tasks due to GPU OOM errors
-   * (GpuRetryOOM and GpuSplitAndRetryOOM).
+   * Returns stage IDs of scan stages with failed tasks due to GPU OOM errors.
+   * Reads from the pre-computed values in AppInfoProfileResults to avoid
+   * duplicate computation (Profiler.processApp already computes these).
    */
   override def scanStagesWithGpuOom: Set[Long] = {
-    SingleAppSummaryInfoProvider.computeScanStagesWithGpuOom(
-      app.appInfo.exists(_.pluginEnabled),
-      app.failedTasks, app.stageMetrics, appInfo)
+    app.appInfo.headOption.map(_.scanStagesWithGpuOom).getOrElse(Set.empty)
   }
 
   /**
    * Returns stage IDs of failed shuffle stages with OOM errors in the task's end reason.
-   * Note: This check is enabled only if the plugin is enabled (i.e. GPU app) and running on YARN.
+   * Reads from the pre-computed values in AppInfoProfileResults.
    */
   override def shuffleStagesWithOom: Set[Long] = {
-    SingleAppSummaryInfoProvider.computeShuffleStagesWithOom(
-      app.appInfo.exists(_.pluginEnabled),
-      getSparkProperty("spark.master"),
-      app.failedStages, app.failedTasks)
+    app.appInfo.headOption.map(_.shuffleStagesWithOom).getOrElse(Set.empty)
   }
 
   /**
-   * Get the maximum data size from ColumnarExchange metrics.
-   * This method searches through SQLPlan metrics to find all ColumnarExchange nodes
-   * with "data size" metrics and returns the maximum total value.
-   *
-   * @return Option[Long] containing the maximum data size in bytes, or None if no
-   *         ColumnarExchange "data size" metrics are found
+   * Returns the maximum data size from ColumnarExchange metrics.
+   * Reads from the pre-computed value in AppInfoProfileResults.
    */
   override def getMaxColumnarExchangeDataSizeBytes: Option[Long] = {
-    SingleAppSummaryInfoProvider.computeMaxColumnarExchangeDataSizeBytes(app.sqlMetrics)
+    app.appInfo.headOption.flatMap(_.maxColumnarExchangeDataSizeBytes)
   }
 
   override def getClassPathEntries: Map[String, String] = {
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
index 8a0422d0e..c990c2b88 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
@@ -1532,9 +1532,10 @@ case class RecommendedCommentResult(comment: String) {
 object SparkRapidsOomExceptions {
   // Current JNI: GpuOOM -> GpuRetryOOM, GpuSplitAndRetryOOM
   // Pre-24.02 JNI: jni.SplitAndRetryOOM, jni.RetryOOM (no Gpu prefix)
-  // Using "jni." prefix to avoid matching CpuSplitAndRetryOOM
+  // Using "jni." prefix to avoid matching CpuSplitAndRetryOOM / CpuRetryOOM
+  // Using "jni.GpuOOM" (anchored) for the base class to avoid partial matches
   val gpuExceptionClassNames: Set[String] =
-    Set("GpuSplitAndRetryOOM", "GpuRetryOOM", "GpuOOM",
+    Set("GpuSplitAndRetryOOM", "GpuRetryOOM", "jni.GpuOOM",
         "jni.SplitAndRetryOOM", "jni.RetryOOM")
 
   val gpuShuffleClassName: String = "GpuShuffleExchangeExec"

From 412b67984a98f15ef8bddfd7660b627c57d7c489 Mon Sep 17 00:00:00 2001
From: Partho Sarthi <psarthi@nvidia.com>
Date: Thu, 16 Apr 2026 07:11:06 -0700
Subject: [PATCH 04/16] Extract tuning metrics into separate
 app_tuning_metrics.csv

Move 4 AutoTuner input fields (maxTaskInputBytesRead,
maxColumnarExchangeDataSizeBytes, scanStagesWithGpuOom,
shuffleStagesWithOom) from application_information.csv into a new
app_tuning_metrics.csv file. These are tuning-specific signals that
don't belong in the core application info table.

Also addresses review feedback:
- Anchor GpuOOM pattern as jni.GpuOOM to avoid partial matches
- Change maxTaskInputBytesRead YAML dataType from Double to Long

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Partho Sarthi <psarthi@nvidia.com>
---
 .../configs/reports/coreRawMetricsReport.yaml | 43 ++++++-----
 .../profiling/ApplicationSummaryInfo.scala    | 27 +++----
 .../tool/profiling/CollectInformation.scala   | 15 ----
 .../profiling/ProfileClassWarehouse.scala     | 71 +++++++++++++------
 .../rapids/tool/profiling/Profiler.scala      | 15 ++--
 .../rapids/tool/views/OutHeaderRegistry.scala |  5 +-
 .../tool/views/QualRawReportGenerator.scala   | 15 ++--
 .../spark/rapids/tool/views/package.scala     |  1 +
 .../tool/profiling/ApplicationInfoSuite.scala |  8 +--
 9 files changed, 111 insertions(+), 89 deletions(-)

diff --git a/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml b/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
index eee4a919a..16d35afed 100644
--- a/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
+++ b/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
@@ -106,22 +106,6 @@ reportDefinitions:
             dataType: Long
             description: >-
               TBD
-          - name: maxTaskInputBytesRead
-            dataType: Long
-            description: >-
-              Maximum per-task input bytes read across all SQLs
-          - name: maxColumnarExchangeDataSizeBytes
-            dataType: Long
-            description: >-
-              Maximum data size from ColumnarExchange metrics (profiling only)
-          - name: scanStagesWithGpuOom
-            dataType: String
-            description: >-
-              Comma-separated stage IDs of scan stages with GPU OOM (profiling only)
-          - name: shuffleStagesWithOom
-            dataType: String
-            description: >-
-              Comma-separated stage IDs of shuffle stages with OOM (profiling only)
       # AppLogPathProfileResults
       - label: coreRawApplicationLogPathMappingCSV
         description: >-
@@ -1442,3 +1426,30 @@ reportDefinitions:
         fileName: profile.log
         fileFormat: TXT
         scope: per-app
+      # AppTuningMetricsProfileResult
+      - label: coreRawAppTuningMetricsCSV
+        description: >-
+          Per-app metrics used by the AutoTuner for partition sizing and cluster sizing.
+        fileName: app_tuning_metrics.csv
+        scope: per-app
+        columns:
+          - name: appId
+            dataType: String
+            description: >-
+              Application ID
+          - name: maxTaskInputBytesRead
+            dataType: Long
+            description: >-
+              Maximum per-task input bytes read across all SQLs
+          - name: maxColumnarExchangeDataSizeBytes
+            dataType: Long
+            description: >-
+              Maximum data size from ColumnarExchange metrics (profiling only)
+          - name: scanStagesWithGpuOom
+            dataType: String
+            description: >-
+              Comma-separated stage IDs of scan stages with GPU OOM (profiling only)
+          - name: shuffleStagesWithOom
+            dataType: String
+            description: >-
+              Comma-separated stage IDs of shuffle stages with OOM (profiling only)
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
index 8be679802..3c50ee477 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
@@ -54,7 +54,8 @@ case class ApplicationSummaryInfo(
     sqlCleanedAlignedIds: Seq[SQLCleanAndAlignIdsProfileResult],
     sparkRapidsBuildInfo: Seq[SparkRapidsBuildInfoEvent],
     writeOpsInfo: Seq[WriteOpProfileResult],
-    sqlPlanInfo: Seq[SQLPlanInfoProfileResult])
+    sqlPlanInfo: Seq[SQLPlanInfoProfileResult],
+    appTuningMetrics: Seq[AppTuningMetricsProfileResult] = Seq.empty)
 
 trait AppInfoPropertyGetter {
   // returns all the properties (i.e., spark)
@@ -229,29 +230,21 @@ class SingleAppSummaryInfoProvider(
     }
   }
 
-  /**
-   * Returns stage IDs of scan stages with failed tasks due to GPU OOM errors.
-   * Reads from the pre-computed values in AppInfoProfileResults to avoid
-   * duplicate computation (Profiler.processApp already computes these).
-   */
   override def scanStagesWithGpuOom: Set[Long] = {
-    app.appInfo.headOption.map(_.scanStagesWithGpuOom).getOrElse(Set.empty)
+    SingleAppSummaryInfoProvider.computeScanStagesWithGpuOom(
+      app.appInfo.exists(_.pluginEnabled),
+      app.failedTasks, app.stageMetrics, appInfo)
   }
 
-  /**
-   * Returns stage IDs of failed shuffle stages with OOM errors in the task's end reason.
-   * Reads from the pre-computed values in AppInfoProfileResults.
-   */
   override def shuffleStagesWithOom: Set[Long] = {
-    app.appInfo.headOption.map(_.shuffleStagesWithOom).getOrElse(Set.empty)
+    SingleAppSummaryInfoProvider.computeShuffleStagesWithOom(
+      app.appInfo.exists(_.pluginEnabled),
+      getSparkProperty("spark.master"),
+      app.failedStages, app.failedTasks)
   }
 
-  /**
-   * Returns the maximum data size from ColumnarExchange metrics.
-   * Reads from the pre-computed value in AppInfoProfileResults.
-   */
   override def getMaxColumnarExchangeDataSizeBytes: Option[Long] = {
-    app.appInfo.headOption.flatMap(_.maxColumnarExchangeDataSizeBytes)
+    SingleAppSummaryInfoProvider.computeMaxColumnarExchangeDataSizeBytes(app.sqlMetrics)
   }
 
   override def getClassPathEntries: Map[String, String] = {
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/CollectInformation.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/CollectInformation.scala
index 8b165733f..e04a74ea0 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/CollectInformation.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/CollectInformation.scala
@@ -37,21 +37,6 @@ class CollectInformation(apps: Seq[ApplicationInfo]) extends Logging {
     ProfInformationView.getRawView(apps)
   }
 
-  // Extends AppInfoProfileResults with fields that require aggregate metrics or
-  // cross-referencing (e.g., failed tasks with stage metrics). Fields like
-  // totalCoreSeconds are already populated at the view layer in InformationView.
-  def getExtendedAppInfo(
-      maxTaskInputBytesRead: Double,
-      maxColumnarExchangeDataSizeBytes: Option[Long],
-      scanStagesWithGpuOom: Set[Long],
-      shuffleStagesWithOom: Set[Long]): Seq[AppInfoProfileResults] = {
-    getAppInfo.map(_.copy(
-      maxTaskInputBytesRead = maxTaskInputBytesRead,
-      maxColumnarExchangeDataSizeBytes = maxColumnarExchangeDataSizeBytes,
-      scanStagesWithGpuOom = scanStagesWithGpuOom,
-      shuffleStagesWithOom = shuffleStagesWithOom))
-  }
-
   def getAppLogPath: Seq[AppLogPathProfileResults] = {
     ProfLogPathView.getRawView(apps)
   }
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
index c990c2b88..7079220af 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
@@ -463,11 +463,7 @@ case class AppInfoProfileResults(
     sparkRuntime: SparkRuntime.SparkRuntime,
     sparkVersion: String,
     pluginEnabled: Boolean,
-    totalCoreSeconds: Long,
-    maxTaskInputBytesRead: Double = 0.0,
-    maxColumnarExchangeDataSizeBytes: Option[Long] = None,
-    scanStagesWithGpuOom: Set[Long] = Set.empty,
-    shuffleStagesWithOom: Set[Long] = Set.empty)  extends ProfileResult {
+    totalCoreSeconds: Long)  extends ProfileResult {
   override def outputHeaders: Array[String] = {
     OutHeaderRegistry.outputHeaders("AppInfoProfileResults")
   }
@@ -500,23 +496,11 @@ case class AppInfoProfileResults(
     }
   }
 
-  private def maxColumnarExchangeToStr: String = {
-    maxColumnarExchangeDataSizeBytes.map(_.toString).getOrElse("")
-  }
-
-  private def stageIdsToStr(stageIds: Set[Long]): String = {
-    if (stageIds.isEmpty) "" else stageIds.toSeq.sorted.mkString(",")
-  }
-
   override def convertToSeq(): Array[String] = {
     Array(appName, appIdToStr, attemptIdToStr,
       sparkUser, startTime.toString, endTimeToStr, durToStr,
       durationStr, sparkRuntime.toString, sparkVersion, pluginEnabled.toString,
-      totalCoreSeconds.toString,
-      f"$maxTaskInputBytesRead%.0f",
-      maxColumnarExchangeToStr,
-      stageIdsToStr(scanStagesWithGpuOom),
-      stageIdsToStr(shuffleStagesWithOom))
+      totalCoreSeconds.toString)
   }
 
   override def convertToCSVSeq(): Array[String] = {
@@ -528,11 +512,52 @@ case class AppInfoProfileResults(
       StringUtils.reformatCSVString(sparkRuntime.toString),
       StringUtils.reformatCSVString(sparkVersion),
       pluginEnabled.toString,
-      totalCoreSeconds.toString,
-      f"$maxTaskInputBytesRead%.0f",
-      maxColumnarExchangeToStr,
-      StringUtils.reformatCSVString(stageIdsToStr(scanStagesWithGpuOom)),
-      StringUtils.reformatCSVString(stageIdsToStr(shuffleStagesWithOom)))
+      totalCoreSeconds.toString)
+  }
+}
+
+case class AppTuningMetricsProfileResult(
+    appId: String,
+    maxTaskInputBytesRead: String,
+    maxColumnarExchangeDataSizeBytes: String,
+    scanStagesWithGpuOom: String,
+    shuffleStagesWithOom: String) extends ProfileResult {
+  override def outputHeaders: Array[String] = {
+    OutHeaderRegistry.outputHeaders("AppTuningMetricsProfileResult")
+  }
+
+  override def convertToSeq(): Array[String] = {
+    Array(appId, maxTaskInputBytesRead, maxColumnarExchangeDataSizeBytes,
+      scanStagesWithGpuOom, shuffleStagesWithOom)
+  }
+
+  override def convertToCSVSeq(): Array[String] = {
+    Array(StringUtils.reformatCSVString(appId),
+      maxTaskInputBytesRead,
+      maxColumnarExchangeDataSizeBytes,
+      StringUtils.reformatCSVString(scanStagesWithGpuOom),
+      StringUtils.reformatCSVString(shuffleStagesWithOom))
+  }
+}
+
+object AppTuningMetricsProfileResult {
+  private def stageIdsToStr(stageIds: Set[Long]): String = {
+    if (stageIds.isEmpty) "" else stageIds.toSeq.sorted.mkString(",")
+  }
+
+  def apply(
+      appId: String,
+      maxTaskInputBytesRead: Double,
+      maxColumnarExchangeDataSizeBytes: Option[Long],
+      scanStagesWithGpuOom: Set[Long],
+      shuffleStagesWithOom: Set[Long]): AppTuningMetricsProfileResult = {
+    AppTuningMetricsProfileResult(
+      appId = appId,
+      maxTaskInputBytesRead = f"$maxTaskInputBytesRead%.0f",
+      maxColumnarExchangeDataSizeBytes =
+        maxColumnarExchangeDataSizeBytes.map(_.toString).getOrElse(""),
+      scanStagesWithGpuOom = stageIdsToStr(scanStagesWithGpuOom),
+      shuffleStagesWithOom = stageIdsToStr(shuffleStagesWithOom))
   }
 }
 
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
index 3b7bd878e..c794b9703 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
@@ -305,7 +305,7 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
     val failedTasks = healthCheck.getFailedTasks
     val failedStages = healthCheck.getFailedStages
 
-    // Compute AutoTuner inputs to enrich application_information.csv
+    // Compute AutoTuner inputs for app_tuning_metrics.csv
     val singleApp = analyzedApps.head
     val pluginEnabled = singleApp.gpuMode
     val maxTaskInput = analysis.maxTaskInputSizes.headOption
@@ -318,9 +318,12 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
       pluginEnabled, singleApp.sparkProperties.get("spark.master"),
       failedStages, failedTasks)
 
-    val appInfo = collect.getExtendedAppInfo(
-      maxTaskInput, maxColumnarExchange, scanOomStages, shuffleOomStages)
-    logDebug(s"Time to collect Profiling Info [${appInfo.head.appId}]: ${endTime - startTime}.")
+    val appInfo = collect.getAppInfo
+    val appId = appInfo.headOption.flatMap(_.appId).getOrElse("")
+    val tuningMetrics = Seq(AppTuningMetricsProfileResult(
+      appId, maxTaskInput, maxColumnarExchange, scanOomStages, shuffleOomStages))
+
+    logDebug(s"Time to collect Profiling Info [$appId]: ${endTime - startTime}.")
     val appInfoSummary = ApplicationSummaryInfo(
       appInfo = appInfo,
       dsInfo = collect.getDataSourceInfo(sqlMetrics),
@@ -351,7 +354,8 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
       sqlCleanedAlignedIds = sqlIdAlign,
       sparkRapidsBuildInfo = collect.getSparkRapidsInfo,
       writeOpsInfo = collect.getWriteOperationInfo,
-      sqlPlanInfo = collect.getSQLPlanInfoTruncated)
+      sqlPlanInfo = collect.getSQLPlanInfoTruncated,
+      appTuningMetrics = tuningMetrics)
     (appInfoSummary,
      DiagnosticSummaryInfo(analysis.stageDiagnostics, collect.getIODiagnosticMetrics))
   }
@@ -424,6 +428,7 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
     // writeOps are generated in only CSV format
     profileOutputWriter.writeCSVTable(ProfWriteOpsView.getLabel, app.writeOpsInfo)
     profileOutputWriter.writeCSVTable(TASK_SHUFFLE_SKEW, app.skewInfo)
+    profileOutputWriter.writeCSVTable(APP_TUNING_METRICS, app.appTuningMetrics)
     profileOutputWriter.writeText("\n### C. Health Check###\n")
     profileOutputWriter.writeCSVTable(ProfFailedTaskView.getLabel, app.failedTasks)
     profileOutputWriter.writeTable(ProfFailedStageView.getLabel, app.failedStages)
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
index cf9b0e2bb..9e7c48408 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
@@ -69,8 +69,9 @@ object OutHeaderRegistry {
       Array("sqlID", "nodeID", "nodeName", "nodeDescription", "reason"),
     "AppInfoProfileResults" ->
       Array("appName", "appId", "attemptId", "sparkUser", "startTime", "endTime", "duration",
-        "durationStr", "sparkRuntime", "sparkVersion", "pluginEnabled", "totalCoreSeconds",
-        "maxTaskInputBytesRead", "maxColumnarExchangeDataSizeBytes",
+        "durationStr", "sparkRuntime", "sparkVersion", "pluginEnabled", "totalCoreSeconds"),
+    "AppTuningMetricsProfileResult" ->
+      Array("appId", "maxTaskInputBytesRead", "maxColumnarExchangeDataSizeBytes",
         "scanStagesWithGpuOom", "shuffleStagesWithOom"),
     "AppLogPathProfileResults" ->
       Array("appName", "appId", "eventLogPath"),
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
index a9e146ed6..3f2a4cd22 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
@@ -17,7 +17,7 @@
 package com.nvidia.spark.rapids.tool.views
 
 import com.nvidia.spark.rapids.tool.analysis.{AggRawMetricsResult, AppSQLPlanAnalyzer, QualSparkMetricsAggregator}
-import com.nvidia.spark.rapids.tool.profiling.{DataSourceProfileResult, ProfileOutputWriter, ProfileResult, SQLAccumProfileResults}
+import com.nvidia.spark.rapids.tool.profiling.{AppTuningMetricsProfileResult, DataSourceProfileResult, ProfileOutputWriter, ProfileResult, SQLAccumProfileResults}
 
 import org.apache.spark.internal.Logging
 import org.apache.spark.sql.rapids.tool.qualification.QualificationAppInfo
@@ -72,18 +72,15 @@ object QualRawReportGenerator extends Logging {
       new ProfileOutputWriter(metricsDirectory, "profile", 10000000, outputCSV = true)
     try {
       // Compute aggregate metrics early so maxTaskInputBytesRead is available for
-      // application_information.csv enrichment
+      // app_tuning_metrics.csv
       val aggRawMetrics = QualSparkMetricsAggregator
         .getAggRawMetrics(app, sqlAnalyzer = Some(sqlPlanAnalyzer))
       val maxTaskInput = aggRawMetrics.maxTaskInputSizes.headOption
         .map(_.maxTaskInputBytesRead).getOrElse(0.0)
 
       pWriter.writeText("### A. Information Collected ###")
-      // Extend application info with maxTaskInputBytesRead from aggregate metrics.
-      // OOM and ColumnarExchange columns are empty for qualification (CPU event logs).
-      val extendedAppInfo = QualInformationView.getRawView(Seq(app)).map(
-        _.copy(maxTaskInputBytesRead = maxTaskInput))
-      pWriter.writeTable(QualInformationView.getLabel, extendedAppInfo)
+      pWriter.writeTable(
+        QualInformationView.getLabel, QualInformationView.getRawView(Seq(app)))
       pWriter.writeTable(QualLogPathView.getLabel, QualLogPathView.getRawView(Seq(app)))
       val sqlPlanMetricsResults = generateSQLProcessingView(pWriter, sqlPlanAnalyzer)
       pWriter.writeJsonL(
@@ -110,6 +107,10 @@ object QualRawReportGenerator extends Logging {
       constructLabelsMaps(aggRawMetrics).foreach { case (label, metrics) =>
           pWriter.writeCSVTable(label, metrics)
       }
+      // Write tuning metrics (GPU-only fields are empty for qualification)
+      val tuningMetrics = Seq(AppTuningMetricsProfileResult(
+        app.appId, maxTaskInput, Option.empty[Long], Set.empty[Long], Set.empty[Long]))
+      pWriter.writeCSVTable(APP_TUNING_METRICS, tuningMetrics)
       pWriter.writeText("\n### C. Health Check###\n")
       pWriter.writeCSVTable(QualFailedTaskView.getLabel, QualFailedTaskView.getRawView(Seq(app)))
       pWriter.writeTable(
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala
index bab4b8a2b..60a5f28cf 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala
@@ -29,6 +29,7 @@ package object views {
   val SQL_DUR_LABEL = "SQL Duration and Executor CPU Time Percent"
   val SQL_MAX_INPUT_SIZE = "SQL Max Task Input Size"
   val STAGE_DIAGNOSTICS_LABEL = "Stage Level Diagnostic Metrics"
+  val APP_TUNING_METRICS = "App Tuning Metrics"
   val CLUSTER_INFORMATION_LABEL = "Cluster Information"
 
   val AGG_DESCRIPTION = Map(
diff --git a/core/src/test/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationInfoSuite.scala b/core/src/test/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationInfoSuite.scala
index 8af32d73c..60e2572cb 100644
--- a/core/src/test/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationInfoSuite.scala
+++ b/core/src/test/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationInfoSuite.scala
@@ -901,22 +901,22 @@ class ApplicationInfoSuite extends AnyFunSuite with Logging {
       ("failures without diagnostic views",
         s"$logDir/tasks_executors_fail_compressed_eventlog.zstd",
         "application_1603128018386_7846",
-        20,
+        21,
         false),
       ("failures with diagnostic views",
         s"$logDir/tasks_executors_fail_compressed_eventlog.zstd",
         "application_1603128018386_7846",
-        22,
+        23,
         true),
       ("gpu without diagnostic views",
         s"$qualLogDir/udf_dataset_eventlog",
         "local-1651188809790",
-        16,
+        17,
         false),
       ("gpu with diagnostic views",
         s"$qualLogDir/udf_dataset_eventlog",
         "local-1651188809790",
-        18,
+        19,
         true)
     )
 

From 7531a6bd2bdedac40a1adbd98bac2313668f2c85 Mon Sep 17 00:00:00 2001
From: Partho Sarthi <psarthi@nvidia.com>
Date: Tue, 21 Apr 2026 22:32:36 -0700
Subject: [PATCH 05/16] Rename output file to application_tuning_metrics.csv,
 default values to 0

- Rename file from app_tuning_metrics.csv to application_tuning_metrics.csv
  (constant APP_TUNING_METRICS -> APPLICATION_TUNING_METRICS).
- Change maxTaskInputBytesRead and maxColumnarExchangeDataSizeBytes in
  AppTuningMetricsProfileResult to Long with default 0, so missing values
  render as "0" symmetrically instead of mixing empty strings and numbers.
- Add application_tuning_metrics.csv to the per-app output tree comment
  in coreRawMetricsReport.yaml.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Signed-off-by: Partho Sarthi <psarthi@nvidia.com>
---
 .../configs/reports/coreRawMetricsReport.yaml          |  3 ++-
 .../rapids/tool/profiling/ProfileClassWarehouse.scala  |  9 ++++-----
 .../nvidia/spark/rapids/tool/profiling/Profiler.scala  |  9 +++++----
 .../rapids/tool/views/QualRawReportGenerator.scala     | 10 +++++-----
 .../com/nvidia/spark/rapids/tool/views/package.scala   |  2 +-
 5 files changed, 17 insertions(+), 16 deletions(-)

diff --git a/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml b/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
index 16d35afed..d6ebb0765 100644
--- a/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
+++ b/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
@@ -18,6 +18,7 @@
 # <application-id>/
 #    ├── application_information.csv
 #    ├── application_log_path_mapping.csv
+#    ├── application_tuning_metrics.csv
 #    ├── cluster_information.json
 #    ├── data_source_information.csv
 #    ├── executor_information.csv
@@ -1430,7 +1431,7 @@ reportDefinitions:
       - label: coreRawAppTuningMetricsCSV
         description: >-
           Per-app metrics used by the AutoTuner for partition sizing and cluster sizing.
-        fileName: app_tuning_metrics.csv
+        fileName: application_tuning_metrics.csv
         scope: per-app
         columns:
           - name: appId
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
index 7079220af..6bfe64554 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
@@ -547,15 +547,14 @@ object AppTuningMetricsProfileResult {
 
   def apply(
       appId: String,
-      maxTaskInputBytesRead: Double,
-      maxColumnarExchangeDataSizeBytes: Option[Long],
+      maxTaskInputBytesRead: Long,
+      maxColumnarExchangeDataSizeBytes: Long,
       scanStagesWithGpuOom: Set[Long],
       shuffleStagesWithOom: Set[Long]): AppTuningMetricsProfileResult = {
     AppTuningMetricsProfileResult(
       appId = appId,
-      maxTaskInputBytesRead = f"$maxTaskInputBytesRead%.0f",
-      maxColumnarExchangeDataSizeBytes =
-        maxColumnarExchangeDataSizeBytes.map(_.toString).getOrElse(""),
+      maxTaskInputBytesRead = maxTaskInputBytesRead.toString,
+      maxColumnarExchangeDataSizeBytes = maxColumnarExchangeDataSizeBytes.toString,
       scanStagesWithGpuOom = stageIdsToStr(scanStagesWithGpuOom),
       shuffleStagesWithOom = stageIdsToStr(shuffleStagesWithOom))
   }
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
index c794b9703..734cb2393 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
@@ -305,11 +305,11 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
     val failedTasks = healthCheck.getFailedTasks
     val failedStages = healthCheck.getFailedStages
 
-    // Compute AutoTuner inputs for app_tuning_metrics.csv
+    // Compute AutoTuner inputs for application_tuning_metrics.csv
     val singleApp = analyzedApps.head
     val pluginEnabled = singleApp.gpuMode
     val maxTaskInput = analysis.maxTaskInputSizes.headOption
-      .map(_.maxTaskInputBytesRead).getOrElse(0.0)
+      .map(_.maxTaskInputBytesRead.toLong).getOrElse(0L)
     val maxColumnarExchange =
       SingleAppSummaryInfoProvider.computeMaxColumnarExchangeDataSizeBytes(sqlMetrics)
     val scanOomStages = SingleAppSummaryInfoProvider.computeScanStagesWithGpuOom(
@@ -321,7 +321,8 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
     val appInfo = collect.getAppInfo
     val appId = appInfo.headOption.flatMap(_.appId).getOrElse("")
     val tuningMetrics = Seq(AppTuningMetricsProfileResult(
-      appId, maxTaskInput, maxColumnarExchange, scanOomStages, shuffleOomStages))
+      appId, maxTaskInput, maxColumnarExchange.getOrElse(0L),
+      scanOomStages, shuffleOomStages))
 
     logDebug(s"Time to collect Profiling Info [$appId]: ${endTime - startTime}.")
     val appInfoSummary = ApplicationSummaryInfo(
@@ -428,7 +429,7 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
     // writeOps are generated in only CSV format
     profileOutputWriter.writeCSVTable(ProfWriteOpsView.getLabel, app.writeOpsInfo)
     profileOutputWriter.writeCSVTable(TASK_SHUFFLE_SKEW, app.skewInfo)
-    profileOutputWriter.writeCSVTable(APP_TUNING_METRICS, app.appTuningMetrics)
+    profileOutputWriter.writeCSVTable(APPLICATION_TUNING_METRICS, app.appTuningMetrics)
     profileOutputWriter.writeText("\n### C. Health Check###\n")
     profileOutputWriter.writeCSVTable(ProfFailedTaskView.getLabel, app.failedTasks)
     profileOutputWriter.writeTable(ProfFailedStageView.getLabel, app.failedStages)
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
index 3f2a4cd22..97a8ca40b 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
@@ -72,11 +72,11 @@ object QualRawReportGenerator extends Logging {
       new ProfileOutputWriter(metricsDirectory, "profile", 10000000, outputCSV = true)
     try {
       // Compute aggregate metrics early so maxTaskInputBytesRead is available for
-      // app_tuning_metrics.csv
+      // application_tuning_metrics.csv
       val aggRawMetrics = QualSparkMetricsAggregator
         .getAggRawMetrics(app, sqlAnalyzer = Some(sqlPlanAnalyzer))
       val maxTaskInput = aggRawMetrics.maxTaskInputSizes.headOption
-        .map(_.maxTaskInputBytesRead).getOrElse(0.0)
+        .map(_.maxTaskInputBytesRead.toLong).getOrElse(0L)
 
       pWriter.writeText("### A. Information Collected ###")
       pWriter.writeTable(
@@ -107,10 +107,10 @@ object QualRawReportGenerator extends Logging {
       constructLabelsMaps(aggRawMetrics).foreach { case (label, metrics) =>
           pWriter.writeCSVTable(label, metrics)
       }
-      // Write tuning metrics (GPU-only fields are empty for qualification)
+      // Write tuning metrics (GPU-only fields default to 0/empty for qualification)
       val tuningMetrics = Seq(AppTuningMetricsProfileResult(
-        app.appId, maxTaskInput, Option.empty[Long], Set.empty[Long], Set.empty[Long]))
-      pWriter.writeCSVTable(APP_TUNING_METRICS, tuningMetrics)
+        app.appId, maxTaskInput, 0L, Set.empty[Long], Set.empty[Long]))
+      pWriter.writeCSVTable(APPLICATION_TUNING_METRICS, tuningMetrics)
       pWriter.writeText("\n### C. Health Check###\n")
       pWriter.writeCSVTable(QualFailedTaskView.getLabel, QualFailedTaskView.getRawView(Seq(app)))
       pWriter.writeTable(
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala
index 60a5f28cf..9a0a22725 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala
@@ -29,7 +29,7 @@ package object views {
   val SQL_DUR_LABEL = "SQL Duration and Executor CPU Time Percent"
   val SQL_MAX_INPUT_SIZE = "SQL Max Task Input Size"
   val STAGE_DIAGNOSTICS_LABEL = "Stage Level Diagnostic Metrics"
-  val APP_TUNING_METRICS = "App Tuning Metrics"
+  val APPLICATION_TUNING_METRICS = "Application Tuning Metrics"
   val CLUSTER_INFORMATION_LABEL = "Cluster Information"
 
   val AGG_DESCRIPTION = Map(

From e41bd1708ef9cb7b4a298271fcd5e58875fa8971 Mon Sep 17 00:00:00 2001
From: Partho Sarthi <psarthi@nvidia.com>
Date: Wed, 22 Apr 2026 02:44:04 -0700
Subject: [PATCH 06/16] Remove redundant comments in
 ApplicationSummaryInfo.scala and QualRawReportGenerator.scala to enhance code
 clarity and maintainability.

---
 .../spark/rapids/tool/profiling/ApplicationSummaryInfo.scala   | 3 ---
 .../spark/rapids/tool/views/QualRawReportGenerator.scala       | 1 -
 2 files changed, 4 deletions(-)

diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
index 3c50ee477..5965e28c0 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
@@ -255,7 +255,6 @@ class SingleAppSummaryInfoProvider(
 object SingleAppSummaryInfoProvider {
   /**
    * Computes the set of scan stage IDs that had GPU OOM failures.
-   * This static method enables computing the result before ApplicationSummaryInfo is assembled.
    */
   def computeScanStagesWithGpuOom(
       pluginEnabled: Boolean,
@@ -293,7 +292,6 @@ object SingleAppSummaryInfoProvider {
    * Computes the set of shuffle stage IDs that had container OOM failures (YARN only).
    * Detects ExecutorLostFailure with exit code 137 (SIGKILL from container memory enforcement).
    * See: https://github.com/NVIDIA/spark-rapids-tools/issues/1566
-   * This static method enables computing the result before ApplicationSummaryInfo is assembled.
    */
   def computeShuffleStagesWithOom(
       pluginEnabled: Boolean,
@@ -341,7 +339,6 @@ object SingleAppSummaryInfoProvider {
 
   /**
    * Computes the maximum data size from ColumnarExchange metrics.
-   * This static method enables computing the result before ApplicationSummaryInfo is assembled.
    */
   def computeMaxColumnarExchangeDataSizeBytes(
       sqlMetrics: Seq[SQLAccumProfileResults]): Option[Long] = {
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
index 97a8ca40b..ac0cb2afa 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
@@ -103,7 +103,6 @@ object QualRawReportGenerator extends Logging {
         SystemQualPropertiesView.getRawView(Seq(app)),
         Some(SystemQualPropertiesView.getDescription))
       pWriter.writeText("\n### B. Analysis ###\n")
-      // Reuse already-computed aggRawMetrics instead of computing again
       constructLabelsMaps(aggRawMetrics).foreach { case (label, metrics) =>
           pWriter.writeCSVTable(label, metrics)
       }

From 9652fc3494951d52fa4dec06551644cb9ce31e7b Mon Sep 17 00:00:00 2001
From: Partho Sarthi <psarthi@nvidia.com>
Date: Wed, 22 Apr 2026 02:53:27 -0700
Subject: [PATCH 07/16] Update copyright year in package.scala to 2024-2026

---
 .../main/scala/com/nvidia/spark/rapids/tool/views/package.scala | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala
index 9a0a22725..33a812eac 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2024-2025, NVIDIA CORPORATION.
+ * Copyright (c) 2024-2026, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.

From 30ccfc3bb3c3aab41a579acb9294721937b64dd2 Mon Sep 17 00:00:00 2001
From: Partho Sarthi <psarthi@nvidia.com>
Date: Wed, 22 Apr 2026 08:27:24 -0700
Subject: [PATCH 08/16] Revert --csv default to opt-in

Keep the original Scallop default (false) to avoid changing CLI behavior.
Qualification always writes CSVs via ProfileOutputWriter(outputCSV = true)
so application_tuning_metrics.csv is still produced there unconditionally.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Signed-off-by: Partho Sarthi <psarthi@nvidia.com>
---
 .../com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala    | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala
index 076a50dd3..84349f148 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala
@@ -89,7 +89,7 @@ Usage: java -cp rapids-4-spark-tools_2.12-<version>.jar:$SPARK_HOME/jars/*
       descr = "Number of thread to use for parallel processing. The default is the " +
         "number of cores on host divided by 4.")
   val csv: ScallopOption[Boolean] =
-    opt[Boolean](required = false, default = Some(true),
+    opt[Boolean](required = false,
       descr = "Output each table to a CSV file as well creating the summary text file.")
   val timeout: ScallopOption[Long] =
     opt[Long](required = false,

From baa8a80e009fc6e56544ebfedc2ab711a788eff3 Mon Sep 17 00:00:00 2001
From: Partho Sarthi <psarthi@nvidia.com>
Date: Wed, 22 Apr 2026 08:30:13 -0700
Subject: [PATCH 09/16] Restore ProfileArgs.scala to origin/dev (no copyright
 bump)

File is unchanged functionally vs dev; skip the auto-copyrighter hook
to keep the copyright year at 2025 since this PR does not modify the file.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Signed-off-by: Partho Sarthi <psarthi@nvidia.com>
---
 .../com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala    | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala
index 84349f148..81a253ebf 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileArgs.scala
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2021-2026, NVIDIA CORPORATION.
+ * Copyright (c) 2021-2025, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.

From aec9fd21694ecbde408295e6003716a79f1297df Mon Sep 17 00:00:00 2001
From: Partho Sarthi <psarthi@nvidia.com>
Date: Wed, 22 Apr 2026 11:02:18 -0700
Subject: [PATCH 10/16] Restructure tuning metrics as vertical
 tuning_signals.csv

- Rename output file application_tuning_metrics.csv -> tuning_signals.csv
  and switch to vertical metricName,value layout (matches spark_properties.csv).
  appId is dropped since the file lives in a per-app directory.
- Replace AppTuningMetricsProfileResult with a row-shaped TuningSignalProfileResult
  plus a builder that emits one row per metric. Simplifies adding new signals.
- Rename trait method shuffleStagesWithOom -> gpuShuffleStagesWithContainerOom
  to make the GPU-only gating explicit and distinguish container-level
  (YARN SIGKILL) OOM from device-level GPU OOM.
- Drop Gpu prefix on the static compute helper
  (computeShuffleStagesWithContainerOom) since it lives in the
  profiling-only SingleAppSummaryInfoProvider companion.
- Update YAML schema, OutHeaderRegistry, Profiler/QualRawReportGenerator
  call sites, AutoTuner consumer, and test mocks accordingly.
- ToolsAPI auto-discovers the new coreRawTuningSignalsCSV label via the
  YAML; no Python changes needed.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Signed-off-by: Partho Sarthi <psarthi@nvidia.com>
---
 .../configs/reports/coreRawMetricsReport.yaml | 34 +++++--------
 .../profiling/ApplicationSummaryInfo.scala    | 10 ++--
 .../profiling/ProfileClassWarehouse.scala     | 48 ++++++++-----------
 .../rapids/tool/profiling/Profiler.scala      | 17 +++----
 .../spark/rapids/tool/tuning/AutoTuner.scala  |  2 +-
 .../rapids/tool/views/OutHeaderRegistry.scala |  5 +-
 .../tool/views/QualRawReportGenerator.scala   | 10 ++--
 .../spark/rapids/tool/views/package.scala     |  2 +-
 .../tool/tuning/BaseAutoTunerSuite.scala      |  8 ++--
 .../tool/tuning/ProfilingAutoTunerSuite.scala |  4 +-
 .../tuning/ProfilingAutoTunerSuiteV2.scala    |  2 +-
 11 files changed, 62 insertions(+), 80 deletions(-)

diff --git a/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml b/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
index d6ebb0765..4d324ed23 100644
--- a/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
+++ b/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
@@ -18,7 +18,6 @@
 # <application-id>/
 #    ├── application_information.csv
 #    ├── application_log_path_mapping.csv
-#    ├── application_tuning_metrics.csv
 #    ├── cluster_information.json
 #    ├── data_source_information.csv
 #    ├── executor_information.csv
@@ -41,6 +40,7 @@
 #    ├── stage_level_aggregated_task_metrics.csv
 #    ├── stage_level_all_metrics.csv
 #    ├── system_properties.csv
+#    ├── tuning_signals.csv
 #    ├── unsupported_sql_plan.csv
 #    ├── wholestagecodegen_mapping.csv
 #    ├── write_operations.csv
@@ -1427,30 +1427,22 @@ reportDefinitions:
         fileName: profile.log
         fileFormat: TXT
         scope: per-app
-      # AppTuningMetricsProfileResult
-      - label: coreRawAppTuningMetricsCSV
+      # TuningSignalProfileResult
+      - label: coreRawTuningSignalsCSV
         description: >-
-          Per-app metrics used by the AutoTuner for partition sizing and cluster sizing.
-        fileName: application_tuning_metrics.csv
+          Per-app diagnostic signals used as AutoTuner inputs (partition sizing, cluster
+          sizing). Vertical key-value layout; each emitted metric is one row. Current
+          rows: maxTaskInputBytesRead, maxColumnarExchangeDataSizeBytes (profiling only),
+          scanStagesWithGpuOom (profiling only), gpuShuffleStagesWithContainerOom (profiling only).
+        fileName: tuning_signals.csv
         scope: per-app
         columns:
-          - name: appId
-            dataType: String
-            description: >-
-              Application ID
-          - name: maxTaskInputBytesRead
-            dataType: Long
-            description: >-
-              Maximum per-task input bytes read across all SQLs
-          - name: maxColumnarExchangeDataSizeBytes
-            dataType: Long
-            description: >-
-              Maximum data size from ColumnarExchange metrics (profiling only)
-          - name: scanStagesWithGpuOom
+          - name: metricName
             dataType: String
             description: >-
-              Comma-separated stage IDs of scan stages with GPU OOM (profiling only)
-          - name: shuffleStagesWithOom
+              Name of the tuning signal
+          - name: value
             dataType: String
             description: >-
-              Comma-separated stage IDs of shuffle stages with OOM (profiling only)
+              Signal value. Numeric signals are stored as their integer string form;
+              stage-ID signals are comma-separated lists (empty string when none).
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
index 5965e28c0..eba1ff24b 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
@@ -55,7 +55,7 @@ case class ApplicationSummaryInfo(
     sparkRapidsBuildInfo: Seq[SparkRapidsBuildInfoEvent],
     writeOpsInfo: Seq[WriteOpProfileResult],
     sqlPlanInfo: Seq[SQLPlanInfoProfileResult],
-    appTuningMetrics: Seq[AppTuningMetricsProfileResult] = Seq.empty)
+    tuningSignals: Seq[TuningSignalProfileResult] = Seq.empty)
 
 trait AppInfoPropertyGetter {
   // returns all the properties (i.e., spark)
@@ -92,7 +92,7 @@ trait AppInfoReadMetrics {
 
 trait AppInfoGpuOomCheck {
   def scanStagesWithGpuOom: Set[Long] = Set.empty
-  def shuffleStagesWithOom: Set[Long] = Set.empty
+  def gpuShuffleStagesWithContainerOom: Set[Long] = Set.empty
 }
 
 trait AppInfoColumnarExchangeMetrics {
@@ -236,8 +236,8 @@ class SingleAppSummaryInfoProvider(
       app.failedTasks, app.stageMetrics, appInfo)
   }
 
-  override def shuffleStagesWithOom: Set[Long] = {
-    SingleAppSummaryInfoProvider.computeShuffleStagesWithOom(
+  override def gpuShuffleStagesWithContainerOom: Set[Long] = {
+    SingleAppSummaryInfoProvider.computeShuffleStagesWithContainerOom(
       app.appInfo.exists(_.pluginEnabled),
       getSparkProperty("spark.master"),
       app.failedStages, app.failedTasks)
@@ -293,7 +293,7 @@ object SingleAppSummaryInfoProvider {
    * Detects ExecutorLostFailure with exit code 137 (SIGKILL from container memory enforcement).
    * See: https://github.com/NVIDIA/spark-rapids-tools/issues/1566
    */
-  def computeShuffleStagesWithOom(
+  def computeShuffleStagesWithContainerOom(
       pluginEnabled: Boolean,
       sparkMasterStr: Option[String],
       failedStages: Seq[FailedStagesProfileResults],
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
index 6bfe64554..4af4676f0 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
@@ -516,48 +516,38 @@ case class AppInfoProfileResults(
   }
 }
 
-case class AppTuningMetricsProfileResult(
-    appId: String,
-    maxTaskInputBytesRead: String,
-    maxColumnarExchangeDataSizeBytes: String,
-    scanStagesWithGpuOom: String,
-    shuffleStagesWithOom: String) extends ProfileResult {
+case class TuningSignalProfileResult(
+    metricName: String,
+    value: String) extends ProfileResult {
   override def outputHeaders: Array[String] = {
-    OutHeaderRegistry.outputHeaders("AppTuningMetricsProfileResult")
+    OutHeaderRegistry.outputHeaders("TuningSignalProfileResult")
   }
 
-  override def convertToSeq(): Array[String] = {
-    Array(appId, maxTaskInputBytesRead, maxColumnarExchangeDataSizeBytes,
-      scanStagesWithGpuOom, shuffleStagesWithOom)
-  }
+  override def convertToSeq(): Array[String] = Array(metricName, value)
 
-  override def convertToCSVSeq(): Array[String] = {
-    Array(StringUtils.reformatCSVString(appId),
-      maxTaskInputBytesRead,
-      maxColumnarExchangeDataSizeBytes,
-      StringUtils.reformatCSVString(scanStagesWithGpuOom),
-      StringUtils.reformatCSVString(shuffleStagesWithOom))
-  }
+  override def convertToCSVSeq(): Array[String] = Array(
+    StringUtils.reformatCSVString(metricName),
+    StringUtils.reformatCSVString(value))
 }
 
-object AppTuningMetricsProfileResult {
+object TuningSignalProfileResult {
   private def stageIdsToStr(stageIds: Set[Long]): String = {
     if (stageIds.isEmpty) "" else stageIds.toSeq.sorted.mkString(",")
   }
 
-  def apply(
-      appId: String,
+  def build(
       maxTaskInputBytesRead: Long,
       maxColumnarExchangeDataSizeBytes: Long,
       scanStagesWithGpuOom: Set[Long],
-      shuffleStagesWithOom: Set[Long]): AppTuningMetricsProfileResult = {
-    AppTuningMetricsProfileResult(
-      appId = appId,
-      maxTaskInputBytesRead = maxTaskInputBytesRead.toString,
-      maxColumnarExchangeDataSizeBytes = maxColumnarExchangeDataSizeBytes.toString,
-      scanStagesWithGpuOom = stageIdsToStr(scanStagesWithGpuOom),
-      shuffleStagesWithOom = stageIdsToStr(shuffleStagesWithOom))
-  }
+      gpuShuffleStagesWithContainerOom: Set[Long]): Seq[TuningSignalProfileResult] = Seq(
+    TuningSignalProfileResult("maxTaskInputBytesRead",
+      maxTaskInputBytesRead.toString),
+    TuningSignalProfileResult("maxColumnarExchangeDataSizeBytes",
+      maxColumnarExchangeDataSizeBytes.toString),
+    TuningSignalProfileResult("scanStagesWithGpuOom",
+      stageIdsToStr(scanStagesWithGpuOom)),
+    TuningSignalProfileResult("gpuShuffleStagesWithContainerOom",
+      stageIdsToStr(gpuShuffleStagesWithContainerOom)))
 }
 
 case class AppLogPathProfileResults(
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
index 734cb2393..1863cd60b 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
@@ -314,15 +314,16 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
       SingleAppSummaryInfoProvider.computeMaxColumnarExchangeDataSizeBytes(sqlMetrics)
     val scanOomStages = SingleAppSummaryInfoProvider.computeScanStagesWithGpuOom(
       pluginEnabled, failedTasks, stageMetrics, singleApp)
-    val shuffleOomStages = SingleAppSummaryInfoProvider.computeShuffleStagesWithOom(
-      pluginEnabled, singleApp.sparkProperties.get("spark.master"),
-      failedStages, failedTasks)
+    val gpuShuffleContainerOomStages =
+      SingleAppSummaryInfoProvider.computeShuffleStagesWithContainerOom(
+        pluginEnabled, singleApp.sparkProperties.get("spark.master"),
+        failedStages, failedTasks)
 
     val appInfo = collect.getAppInfo
     val appId = appInfo.headOption.flatMap(_.appId).getOrElse("")
-    val tuningMetrics = Seq(AppTuningMetricsProfileResult(
-      appId, maxTaskInput, maxColumnarExchange.getOrElse(0L),
-      scanOomStages, shuffleOomStages))
+    val tuningSignals = TuningSignalProfileResult.build(
+      maxTaskInput, maxColumnarExchange.getOrElse(0L),
+      scanOomStages, gpuShuffleContainerOomStages)
 
     logDebug(s"Time to collect Profiling Info [$appId]: ${endTime - startTime}.")
     val appInfoSummary = ApplicationSummaryInfo(
@@ -356,7 +357,7 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
       sparkRapidsBuildInfo = collect.getSparkRapidsInfo,
       writeOpsInfo = collect.getWriteOperationInfo,
       sqlPlanInfo = collect.getSQLPlanInfoTruncated,
-      appTuningMetrics = tuningMetrics)
+      tuningSignals = tuningSignals)
     (appInfoSummary,
      DiagnosticSummaryInfo(analysis.stageDiagnostics, collect.getIODiagnosticMetrics))
   }
@@ -429,7 +430,7 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
     // writeOps are generated in only CSV format
     profileOutputWriter.writeCSVTable(ProfWriteOpsView.getLabel, app.writeOpsInfo)
     profileOutputWriter.writeCSVTable(TASK_SHUFFLE_SKEW, app.skewInfo)
-    profileOutputWriter.writeCSVTable(APPLICATION_TUNING_METRICS, app.appTuningMetrics)
+    profileOutputWriter.writeCSVTable(TUNING_SIGNALS, app.tuningSignals)
     profileOutputWriter.writeText("\n### C. Health Check###\n")
     profileOutputWriter.writeCSVTable(ProfFailedTaskView.getLabel, app.failedTasks)
     profileOutputWriter.writeTable(ProfFailedStageView.getLabel, app.failedStages)
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/tuning/AutoTuner.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/tuning/AutoTuner.scala
index 3996cfd34..9cf961620 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/tuning/AutoTuner.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/tuning/AutoTuner.scala
@@ -2203,7 +2203,7 @@ class ProfilingAutoTuner(
    */
   override def recommendShufflePartitionsInternal(): Int = {
     val calculatedValue = super.recommendShufflePartitionsInternal()
-    if (appInfoProvider.shuffleStagesWithOom.nonEmpty) {
+    if (appInfoProvider.gpuShuffleStagesWithContainerOom.nonEmpty) {
       // Shuffle Stages with Task OOM detected. We may want to increase shuffle partitions.
       val recShufflePartitions = shufflePartitionValue *
         configProvider.getEntry("SHUFFLE_PARTITION_MULTIPLIER").getDefault.toInt
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
index 9e7c48408..560c8731e 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
@@ -70,9 +70,8 @@ object OutHeaderRegistry {
     "AppInfoProfileResults" ->
       Array("appName", "appId", "attemptId", "sparkUser", "startTime", "endTime", "duration",
         "durationStr", "sparkRuntime", "sparkVersion", "pluginEnabled", "totalCoreSeconds"),
-    "AppTuningMetricsProfileResult" ->
-      Array("appId", "maxTaskInputBytesRead", "maxColumnarExchangeDataSizeBytes",
-        "scanStagesWithGpuOom", "shuffleStagesWithOom"),
+    "TuningSignalProfileResult" ->
+      Array("metricName", "value"),
     "AppLogPathProfileResults" ->
       Array("appName", "appId", "eventLogPath"),
     "FailedTaskProfileResults" ->
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
index ac0cb2afa..c6f9a0cc5 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
@@ -17,7 +17,7 @@
 package com.nvidia.spark.rapids.tool.views
 
 import com.nvidia.spark.rapids.tool.analysis.{AggRawMetricsResult, AppSQLPlanAnalyzer, QualSparkMetricsAggregator}
-import com.nvidia.spark.rapids.tool.profiling.{AppTuningMetricsProfileResult, DataSourceProfileResult, ProfileOutputWriter, ProfileResult, SQLAccumProfileResults}
+import com.nvidia.spark.rapids.tool.profiling.{DataSourceProfileResult, ProfileOutputWriter, ProfileResult, SQLAccumProfileResults, TuningSignalProfileResult}
 
 import org.apache.spark.internal.Logging
 import org.apache.spark.sql.rapids.tool.qualification.QualificationAppInfo
@@ -106,10 +106,10 @@ object QualRawReportGenerator extends Logging {
       constructLabelsMaps(aggRawMetrics).foreach { case (label, metrics) =>
           pWriter.writeCSVTable(label, metrics)
       }
-      // Write tuning metrics (GPU-only fields default to 0/empty for qualification)
-      val tuningMetrics = Seq(AppTuningMetricsProfileResult(
-        app.appId, maxTaskInput, 0L, Set.empty[Long], Set.empty[Long]))
-      pWriter.writeCSVTable(APPLICATION_TUNING_METRICS, tuningMetrics)
+      // GPU-only signals default to 0/empty for qualification (CPU event logs)
+      val tuningSignals = TuningSignalProfileResult.build(
+        maxTaskInput, 0L, Set.empty[Long], Set.empty[Long])
+      pWriter.writeCSVTable(TUNING_SIGNALS, tuningSignals)
       pWriter.writeText("\n### C. Health Check###\n")
       pWriter.writeCSVTable(QualFailedTaskView.getLabel, QualFailedTaskView.getRawView(Seq(app)))
       pWriter.writeTable(
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala
index 33a812eac..972bc6fd5 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala
@@ -29,7 +29,7 @@ package object views {
   val SQL_DUR_LABEL = "SQL Duration and Executor CPU Time Percent"
   val SQL_MAX_INPUT_SIZE = "SQL Max Task Input Size"
   val STAGE_DIAGNOSTICS_LABEL = "Stage Level Diagnostic Metrics"
-  val APPLICATION_TUNING_METRICS = "Application Tuning Metrics"
+  val TUNING_SIGNALS = "Tuning Signals"
   val CLUSTER_INFORMATION_LABEL = "Cluster Information"
 
   val AGG_DESCRIPTION = Map(
diff --git a/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/BaseAutoTunerSuite.scala b/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/BaseAutoTunerSuite.scala
index ceffc08c7..d76ca0b38 100644
--- a/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/BaseAutoTunerSuite.scala
+++ b/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/BaseAutoTunerSuite.scala
@@ -50,7 +50,7 @@ class AppInfoProviderMockTest(val maxInput: Double,
     val shuffleStagesWithPosSpilling: Set[Long],
     val shuffleSkewStages: Set[Long],
     val scanStagesWithGpuOomSet: Set[Long],
-    val shuffleStagesWithOomSet: Set[Long],
+    val gpuShuffleStagesWithContainerOomSet: Set[Long],
     val maxColumnarExchangeDataSizeBytes: Option[Long] = None)
     extends BaseProfilingAppSummaryInfoProvider {
   override def isAppInfoAvailable = true
@@ -70,7 +70,7 @@ class AppInfoProviderMockTest(val maxInput: Double,
   override def getShuffleStagesWithPosSpilling: Set[Long] = shuffleStagesWithPosSpilling
   override def getShuffleSkewStages: Set[Long] = shuffleSkewStages
   override def scanStagesWithGpuOom: Set[Long] = scanStagesWithGpuOomSet
-  override def shuffleStagesWithOom: Set[Long] = shuffleStagesWithOomSet
+  override def gpuShuffleStagesWithContainerOom: Set[Long] = gpuShuffleStagesWithContainerOomSet
   override def getMaxColumnarExchangeDataSizeBytes: Option[Long] = maxColumnarExchangeDataSizeBytes
 
   /**
@@ -136,12 +136,12 @@ abstract class BaseAutoTunerSuite extends AnyFunSuite with BeforeAndAfterEach
       shuffleStagesWithPosSpilling: Set[Long] = Set(),
       shuffleSkewStages: Set[Long] = Set(),
       scanStagesWithGpuOom: Set[Long] = Set(),
-      shuffleStagesWithOom: Set[Long] = Set(),
+      gpuShuffleStagesWithContainerOom: Set[Long] = Set(),
       maxColumnarExchangeDataSizeBytes: Option[Long] = None): AppInfoProviderMockTest = {
     new AppInfoProviderMockTest(maxInput, spilledMetrics, jvmGCFractions, propsFromLog,
       sparkVersion, rapidsJars, distinctLocationPct, redundantReadSize, meanInput, meanShuffleRead,
       shuffleStagesWithPosSpilling, shuffleSkewStages, scanStagesWithGpuOom,
-      shuffleStagesWithOom, maxColumnarExchangeDataSizeBytes)
+      gpuShuffleStagesWithContainerOom, maxColumnarExchangeDataSizeBytes)
   }
 
   /**
diff --git a/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/ProfilingAutoTunerSuite.scala b/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/ProfilingAutoTunerSuite.scala
index 19ccb32cf..24583ea60 100644
--- a/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/ProfilingAutoTunerSuite.scala
+++ b/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/ProfilingAutoTunerSuite.scala
@@ -3292,7 +3292,7 @@ class ProfilingAutoTunerSuite extends ProfilingAutoTunerSuiteBase {
         "spark.plugins" -> "com.nvidia.spark.SQLPlugin",
         "spark.rapids.sql.concurrentGpuTasks" -> "4")
     val infoProvider = getMockInfoProvider(0, Seq(0), Seq(0.0),
-      logEventsProps, Some(testSparkVersion), shuffleStagesWithOom = Set(1L))
+      logEventsProps, Some(testSparkVersion), gpuShuffleStagesWithContainerOom = Set(1L))
     val platform = PlatformFactory.createInstance(PlatformNames.DATAPROC)
 
     // Configure cluster info using Platform's existing method
@@ -3380,7 +3380,7 @@ class ProfilingAutoTunerSuite extends ProfilingAutoTunerSuiteBase {
         "spark.plugins" -> "com.nvidia.spark.SQLPlugin",
         "spark.rapids.sql.concurrentGpuTasks" -> "4")
     val infoProvider = getMockInfoProvider(0, Seq(0), Seq(0.0),
-      logEventsProps, Some(testSparkVersion), shuffleStagesWithOom = Set(1L),
+      logEventsProps, Some(testSparkVersion), gpuShuffleStagesWithContainerOom = Set(1L),
       meanInput = 50000, meanShuffleRead = 80000)
     val platform = PlatformFactory.createInstance(PlatformNames.DATAPROC)
 
diff --git a/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/ProfilingAutoTunerSuiteV2.scala b/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/ProfilingAutoTunerSuiteV2.scala
index 03205bc24..612ad4a62 100644
--- a/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/ProfilingAutoTunerSuiteV2.scala
+++ b/core/src/test/scala/com/nvidia/spark/rapids/tool/tuning/ProfilingAutoTunerSuiteV2.scala
@@ -1693,7 +1693,7 @@ class ProfilingAutoTunerSuiteV2 extends ProfilingAutoTunerSuiteBase {
       shuffleStagesWithPosSpilling = Set.empty,
       shuffleSkewStages = Set.empty,
       scanStagesWithGpuOom = Set.empty,
-      shuffleStagesWithOom = Set.empty,
+      gpuShuffleStagesWithContainerOom = Set.empty,
       maxColumnarExchangeDataSizeBytes = Some(1000L * 1024 * 1024 * 1024) // 1000GB
     )
     val platform = PlatformFactory.createInstance(PlatformNames.DATAPROC, Some(targetClusterInfo))

From 59b1167b9067c154159714cf2440bae8eef99528 Mon Sep 17 00:00:00 2001
From: Partho Sarthi <psarthi@nvidia.com>
Date: Wed, 22 Apr 2026 11:18:38 -0700
Subject: [PATCH 11/16] Rename tuning_signals column metricName -> name

Shorter header matches the spark_properties.csv (propertyName, propertyValue)
pattern but with 'name' since the containing CSV already conveys the
'metric/signal' context via its filename.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Signed-off-by: Partho Sarthi <psarthi@nvidia.com>
---
 .../resources/configs/reports/coreRawMetricsReport.yaml     | 2 +-
 .../spark/rapids/tool/profiling/ProfileClassWarehouse.scala | 6 +++---
 .../nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala  | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml b/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
index 4d324ed23..6cfb7d023 100644
--- a/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
+++ b/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
@@ -1437,7 +1437,7 @@ reportDefinitions:
         fileName: tuning_signals.csv
         scope: per-app
         columns:
-          - name: metricName
+          - name: name
             dataType: String
             description: >-
               Name of the tuning signal
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
index 4af4676f0..d3a382a7c 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
@@ -517,16 +517,16 @@ case class AppInfoProfileResults(
 }
 
 case class TuningSignalProfileResult(
-    metricName: String,
+    name: String,
     value: String) extends ProfileResult {
   override def outputHeaders: Array[String] = {
     OutHeaderRegistry.outputHeaders("TuningSignalProfileResult")
   }
 
-  override def convertToSeq(): Array[String] = Array(metricName, value)
+  override def convertToSeq(): Array[String] = Array(name, value)
 
   override def convertToCSVSeq(): Array[String] = Array(
-    StringUtils.reformatCSVString(metricName),
+    StringUtils.reformatCSVString(name),
     StringUtils.reformatCSVString(value))
 }
 
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
index 560c8731e..297665444 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
@@ -71,7 +71,7 @@ object OutHeaderRegistry {
       Array("appName", "appId", "attemptId", "sparkUser", "startTime", "endTime", "duration",
         "durationStr", "sparkRuntime", "sparkVersion", "pluginEnabled", "totalCoreSeconds"),
     "TuningSignalProfileResult" ->
-      Array("metricName", "value"),
+      Array("name", "value"),
     "AppLogPathProfileResults" ->
       Array("appName", "appId", "eventLogPath"),
     "FailedTaskProfileResults" ->

From 8ddd520f68a25b0c252ad44aad252bfd3b34e527 Mon Sep 17 00:00:00 2001
From: Partho Sarthi <psarthi@nvidia.com>
Date: Wed, 22 Apr 2026 16:09:21 -0700
Subject: [PATCH 12/16] Relocate size signals to aggregated CSVs; keep OOM-only
 tuning_signals.csv
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- Add input_bytesRead_max column to stage/SQL/job aggregated task metrics
  CSVs alongside existing _max columns (duration_max, peakExecutionMemory_max,
  resultSize_max). Tracked in TaskMetricsAccumRec and flows through
  StageAggAccum / SQLAggAccum / JobAggAccum to the three result case classes.
- Drop maxTaskInputBytesRead and maxColumnarExchangeDataSizeBytes rows from
  tuning_signals.csv. The file now contains only the two GPU OOM signals.
- Remove SQLMaxTaskInputSizes case class, maxTaskInputSizeBytesPerSQL()
  helper, and the maxTaskInputSizes field from AggRawMetricsResult and
  ApplicationSummaryInfo — consumers (getMaxInput on both providers) now
  derive the value from sqlTaskAggMetrics/sqlAggs via inputBytesReadMax.
- AutoTuner's AQE ColumnarExchange adjustment is unchanged — it reads
  from in-memory sqlMetrics via SingleAppSummaryInfoProvider, not from CSV.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Signed-off-by: Partho Sarthi <psarthi@nvidia.com>
---
 .../configs/reports/coreRawMetricsReport.yaml | 18 +++++++++---
 .../tool/analysis/AggRawMetricsResult.scala   |  4 +--
 .../analysis/AppSparkMetricsAggTrait.scala    |  2 --
 .../analysis/AppSparkMetricsAnalyzer.scala    | 29 ++-----------------
 .../analysis/util/TaskMetricsAccumRec.scala   |  4 +++
 .../profiling/ApplicationSummaryInfo.scala    |  5 ++--
 .../profiling/ProfileClassWarehouse.scala     | 20 +++++--------
 .../rapids/tool/profiling/Profiler.scala      |  9 +-----
 .../tuning/QualAppSummaryInfoProvider.scala   |  4 +--
 .../rapids/tool/views/OutHeaderRegistry.scala |  3 ++
 .../tool/views/QualRawReportGenerator.scala   |  9 ++----
 .../rapids/tool/views/RawMetricProfView.scala |  4 +--
 12 files changed, 41 insertions(+), 70 deletions(-)

diff --git a/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml b/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
index 6cfb7d023..4fadf8a3f 100644
--- a/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
+++ b/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
@@ -373,6 +373,10 @@ reportDefinitions:
             dataType: Long
             description: >-
               TBD
+          - name: input_bytesRead_max
+            dataType: Long
+            description: >-
+              Maximum per-task input bytes read within the aggregation unit
           - name: input_recordsRead_sum
             dataType: Long
             description: >-
@@ -570,6 +574,10 @@ reportDefinitions:
             dataType: Long
             description: >-
               TBD
+          - name: input_bytesRead_max
+            dataType: Long
+            description: >-
+              Maximum per-task input bytes read within the aggregation unit
           - name: input_recordsRead_sum
             dataType: Long
             description: >-
@@ -783,6 +791,10 @@ reportDefinitions:
             dataType: Long
             description: >-
               TBD
+          - name: input_bytesRead_max
+            dataType: Long
+            description: >-
+              Maximum per-task input bytes read within the aggregation unit
           - name: input_recordsRead_sum
             dataType: Long
             description: >-
@@ -1430,10 +1442,8 @@ reportDefinitions:
       # TuningSignalProfileResult
       - label: coreRawTuningSignalsCSV
         description: >-
-          Per-app diagnostic signals used as AutoTuner inputs (partition sizing, cluster
-          sizing). Vertical key-value layout; each emitted metric is one row. Current
-          rows: maxTaskInputBytesRead, maxColumnarExchangeDataSizeBytes (profiling only),
-          scanStagesWithGpuOom (profiling only), gpuShuffleStagesWithContainerOom (profiling only).
+          Per-app GPU-OOM diagnostic signals used as AutoTuner inputs. Vertical key-value
+          layout; each emitted metric is one row.
         fileName: tuning_signals.csv
         scope: per-app
         columns:
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AggRawMetricsResult.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AggRawMetricsResult.scala
index 2584d702f..8a5f2d7ac 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AggRawMetricsResult.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AggRawMetricsResult.scala
@@ -16,7 +16,7 @@
 
 package com.nvidia.spark.rapids.tool.analysis
 
-import com.nvidia.spark.rapids.tool.profiling.{IOAnalysisProfileResult, JobAggTaskMetricsProfileResult, ShuffleSkewProfileResult, SQLDurationExecutorTimeProfileResult, SQLMaxTaskInputSizes, SQLTaskAggMetricsProfileResult, StageAggTaskMetricsProfileResult, StageDiagnosticResult}
+import com.nvidia.spark.rapids.tool.profiling.{IOAnalysisProfileResult, JobAggTaskMetricsProfileResult, ShuffleSkewProfileResult, SQLDurationExecutorTimeProfileResult, SQLTaskAggMetricsProfileResult, StageAggTaskMetricsProfileResult, StageDiagnosticResult}
 
 /**
  * The result of the aggregation of the raw metrics. It contains the aggregated metrics for an
@@ -31,7 +31,6 @@ import com.nvidia.spark.rapids.tool.profiling.{IOAnalysisProfileResult, JobAggTa
  * @param sqlAggs           the aggregated Spark metrics for SQLs
  * @param ioAggs            lists the SQLs along their IO metrics
  * @param sqlDurAggs        the aggregated duration and CPU time for SQLs
- * @param maxTaskInputSizes a sequence of SQLMaxTaskInputSizes that contains the maximum input size
  * @param stageDiagnostics  the stage level Spark metrics for diagnostic purposes
  */
 case class AggRawMetricsResult(
@@ -41,5 +40,4 @@ case class AggRawMetricsResult(
     sqlAggs: Seq[SQLTaskAggMetricsProfileResult],
     ioAggs: Seq[IOAnalysisProfileResult],
     sqlDurAggs: Seq[SQLDurationExecutorTimeProfileResult],
-    maxTaskInputSizes: Seq[SQLMaxTaskInputSizes],
     stageDiagnostics: Seq[StageDiagnosticResult])
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AppSparkMetricsAggTrait.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AppSparkMetricsAggTrait.scala
index cec8cc9b8..a8f39a168 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AppSparkMetricsAggTrait.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AppSparkMetricsAggTrait.scala
@@ -45,7 +45,6 @@ trait AppSparkMetricsAggTrait extends AppIndexMapperTrait {
       sqlMetricsAgg,
       analysisObj.aggregateIOMetricsBySql(sqlMetricsAgg),
       analysisObj.aggregateDurationAndCPUTimeBySql(index),
-      Seq(analysisObj.maxTaskInputSizeBytesPerSQL(index)),
       analysisObj.aggregateDiagnosticMetricsByStage(index, sqlAnalyzer))
   }
 
@@ -67,7 +66,6 @@ trait AppSparkMetricsAggTrait extends AppIndexMapperTrait {
         agg1.sqlAggs ++ agg2.sqlAggs,
         agg1.ioAggs ++ agg2.ioAggs,
         agg1.sqlDurAggs ++ agg2.sqlDurAggs,
-        agg1.maxTaskInputSizes ++ agg2.maxTaskInputSizes,
         agg1.stageDiagnostics ++ agg2.stageDiagnostics)
     }
   }
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AppSparkMetricsAnalyzer.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AppSparkMetricsAnalyzer.scala
index ee7332cb6..1a4c3f12e 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AppSparkMetricsAnalyzer.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AppSparkMetricsAnalyzer.scala
@@ -106,6 +106,7 @@ class AppSparkMetricsAnalyzer(app: AppBase) extends AppAnalysisBase(app) {
             perJobRec.executorDeserializeTimeSum,
             perJobRec.executorRunTimeSum,
             perJobRec.inputBytesReadSum,
+            perJobRec.inputBytesReadMax,
             perJobRec.inputRecordsReadSum,
             perJobRec.jvmGCTimeSum,
             perJobRec.memoryBytesSpilledSum,
@@ -203,6 +204,7 @@ class AppSparkMetricsAnalyzer(app: AppBase) extends AppAnalysisBase(app) {
             preSqlRec.executorDeserializeTimeSum,
             preSqlRec.executorRunTimeSum,
             preSqlRec.inputBytesReadSum,
+            preSqlRec.inputBytesReadMax,
             preSqlRec.inputBytesReadAvg,
             preSqlRec.inputRecordsReadSum,
             preSqlRec.jvmGCTimeSum,
@@ -252,32 +254,6 @@ class AppSparkMetricsAnalyzer(app: AppBase) extends AppAnalysisBase(app) {
     }.toSeq
   }
 
-  /**
-   * Find the maximum task input size
-   * @param index App index  (used by the profiler tool)
-   * @return a single SQLMaxTaskInputSizes record that contains the maximum value. If none, it will
-   *         be 0L
-   */
-  def maxTaskInputSizeBytesPerSQL(index: Int): SQLMaxTaskInputSizes = {
-    // TODO: We should keep maxInputSize as a field in the stageAggregate to avoid doing an
-    //       extra path on the tasks
-    val maxOfSqls = app.sqlIdToStages.map { case (_, stageIds) =>
-      // TODO: Should we only consider successful tasks?
-      val tasksInSQL = app.taskManager.getTasksByStageIds(stageIds)
-      if (tasksInSQL.isEmpty) {
-        0L
-      } else {
-        tasksInSQL.map(_.input_bytesRead).max
-      }
-    }
-    val maxVal = if (maxOfSqls.nonEmpty) {
-      maxOfSqls.max
-    } else {
-      0L
-    }
-    SQLMaxTaskInputSizes(app.appId, maxVal)
-  }
-
   /**
    * Aggregates the duration and CPU time (milliseconds) by SQL
    * @param index App index  (used by the profiler tool)
@@ -398,6 +374,7 @@ class AppSparkMetricsAnalyzer(app: AppBase) extends AppAnalysisBase(app) {
         perStageRec.executorDeserializeTimeSum,
         perStageRec.executorRunTimeSum,
         perStageRec.inputBytesReadSum,
+        perStageRec.inputBytesReadMax,
         perStageRec.inputRecordsReadSum,
         perStageRec.jvmGCTimeSum,
         perStageRec.memoryBytesSpilledSum,
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/util/TaskMetricsAccumRec.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/util/TaskMetricsAccumRec.scala
index b5d98b9ac..79174e3b0 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/util/TaskMetricsAccumRec.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/util/TaskMetricsAccumRec.scala
@@ -38,6 +38,7 @@ class TaskMetricsAccumRec {
   var executorDeserializeTimeSum: Long = 0
   var executorRunTimeSum: Long = 0
   var inputBytesReadSum: Long = 0
+  var inputBytesReadMax: Long = Long.MinValue
   var inputRecordsReadSum: Long = 0
   var jvmGCTimeSum: Long = 0
   var memoryBytesSpilledSum: Long = 0
@@ -70,6 +71,7 @@ class TaskMetricsAccumRec {
   def resetFields(): Unit = {
     durationMax = 0
     durationMin = 0
+    inputBytesReadMax = 0
     peakExecutionMemoryMax = 0
     resultSizeMax = 0
   }
@@ -102,6 +104,7 @@ class TaskMetricsAccumRec {
     swWriteTimeSum += rec.sw_writeTime
     // Max fields
     durationMax = math.max(durationMax, rec.duration)
+    inputBytesReadMax = math.max(inputBytesReadMax, rec.input_bytesRead)
     peakExecutionMemoryMax = math.max(peakExecutionMemoryMax, rec.peakExecutionMemory)
     resultSizeMax = math.max(resultSizeMax, rec.resultSize)
     // Min Fields
@@ -136,6 +139,7 @@ class TaskMetricsAccumRec {
     swWriteTimeSum += rec.swWriteTimeSum
     // Max
     durationMax = math.max(durationMax, rec.durationMax)
+    inputBytesReadMax = math.max(inputBytesReadMax, rec.inputBytesReadMax)
     peakExecutionMemoryMax = math.max(peakExecutionMemoryMax, rec.peakExecutionMemoryMax)
     resultSizeMax = math.max(resultSizeMax, rec.resultSizeMax)
     // Min
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
index eba1ff24b..c8db4cced 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
@@ -47,7 +47,6 @@ case class ApplicationSummaryInfo(
     sparkProps: Seq[RapidsPropertyProfileResult],
     sqlStageInfo: Seq[SQLStageInfoProfileResult],
     wholeStage: Seq[WholeStageCodeGenResults],
-    maxTaskInputBytesRead: Seq[SQLMaxTaskInputSizes],
     appLogPath: Seq[AppLogPathProfileResults],
     ioMetrics: Seq[IOAnalysisProfileResult],
     sysProps: Seq[RapidsPropertyProfileResult],
@@ -187,8 +186,8 @@ class SingleAppSummaryInfoProvider(
   }
 
   override def getMaxInput: Double = {
-    if (app.maxTaskInputBytesRead.nonEmpty) {
-      app.maxTaskInputBytesRead.head.maxTaskInputBytesRead
+    if (app.sqlTaskAggMetrics.nonEmpty) {
+      app.sqlTaskAggMetrics.map(_.inputBytesReadMax).max.toDouble
     } else {
       0.0
     }
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
index d3a382a7c..bd4f74677 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
@@ -536,14 +536,8 @@ object TuningSignalProfileResult {
   }
 
   def build(
-      maxTaskInputBytesRead: Long,
-      maxColumnarExchangeDataSizeBytes: Long,
       scanStagesWithGpuOom: Set[Long],
       gpuShuffleStagesWithContainerOom: Set[Long]): Seq[TuningSignalProfileResult] = Seq(
-    TuningSignalProfileResult("maxTaskInputBytesRead",
-      maxTaskInputBytesRead.toString),
-    TuningSignalProfileResult("maxColumnarExchangeDataSizeBytes",
-      maxColumnarExchangeDataSizeBytes.toString),
     TuningSignalProfileResult("scanStagesWithGpuOom",
       stageIdsToStr(scanStagesWithGpuOom)),
     TuningSignalProfileResult("gpuShuffleStagesWithContainerOom",
@@ -785,6 +779,7 @@ trait BaseJobStageAggTaskMetricsProfileResult extends ProfileResult {
   def executorDeserializeTimeSum: Long
   def executorRunTimeSum: Long
   def inputBytesReadSum: Long
+  def inputBytesReadMax: Long
   def inputRecordsReadSum: Long
   def jvmGCTimeSum: Long
   def memoryBytesSpilledSum: Long
@@ -825,6 +820,7 @@ trait BaseJobStageAggTaskMetricsProfileResult extends ProfileResult {
       executorDeserializeTimeSum.toString,
       executorRunTimeSum.toString,
       inputBytesReadSum.toString,
+      inputBytesReadMax.toString,
       inputRecordsReadSum.toString,
       jvmGCTimeSum.toString,
       memoryBytesSpilledSum.toString,
@@ -862,6 +858,7 @@ case class JobAggTaskMetricsProfileResult(
     executorDeserializeTimeSum: Long,
     executorRunTimeSum: Long,
     inputBytesReadSum: Long,
+    inputBytesReadMax: Long,
     inputRecordsReadSum: Long,
     jvmGCTimeSum: Long,
     memoryBytesSpilledSum: Long,
@@ -902,6 +899,7 @@ case class StageAggTaskMetricsProfileResult(
     executorDeserializeTimeSum: Long,
     executorRunTimeSum: Long,
     inputBytesReadSum: Long,
+    inputBytesReadMax: Long,
     inputRecordsReadSum: Long,
     jvmGCTimeSum: Long,
     memoryBytesSpilledSum: Long,
@@ -956,6 +954,7 @@ case class StageAggTaskMetricsProfileResult(
         other.executorDeserializeTimeSum,
       executorRunTimeSum = this.executorRunTimeSum + other.executorRunTimeSum,
       inputBytesReadSum = this.inputBytesReadSum + other.inputBytesReadSum,
+      inputBytesReadMax = Math.max(this.inputBytesReadMax, other.inputBytesReadMax),
       inputRecordsReadSum = this.inputRecordsReadSum + other.inputRecordsReadSum,
       jvmGCTimeSum = this.jvmGCTimeSum + other.jvmGCTimeSum,
       memoryBytesSpilledSum = this.memoryBytesSpilledSum + other.memoryBytesSpilledSum,
@@ -1118,12 +1117,6 @@ case class StageDiagnosticResult(
   }
 }
 
-case class SQLMaxTaskInputSizes(
-    appId: String,
-    // Not added to the output since it is used only by the AutoTuner
-    maxTaskInputBytesRead: Double
-)
-
 case class SQLTaskAggMetricsProfileResult(
     appId: String,
     sqlId: Long,
@@ -1141,6 +1134,7 @@ case class SQLTaskAggMetricsProfileResult(
     executorDeserializeTimeSum: Long,
     executorRunTimeSum: Long,
     inputBytesReadSum: Long,
+    inputBytesReadMax: Long,
     // Not added to the output since it is used only by the AutoTuner
     inputBytesReadAvg: Double,
     inputRecordsReadSum: Long,
@@ -1189,6 +1183,7 @@ case class SQLTaskAggMetricsProfileResult(
       executorDeserializeTimeSum.toString,
       executorRunTimeSum.toString,
       inputBytesReadSum.toString,
+      inputBytesReadMax.toString,
       inputRecordsReadSum.toString,
       jvmGCTimeSum.toString,
       memoryBytesSpilledSum.toString,
@@ -1225,6 +1220,7 @@ case class SQLTaskAggMetricsProfileResult(
       executorDeserializeTimeSum.toString,
       executorRunTimeSum.toString,
       inputBytesReadSum.toString,
+      inputBytesReadMax.toString,
       inputRecordsReadSum.toString,
       jvmGCTimeSum.toString,
       memoryBytesSpilledSum.toString,
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
index 1863cd60b..915232f45 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
@@ -293,7 +293,6 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
       }
     }
     val analysis = RawMetricProfilerView.getAggMetrics(analyzedApps)
-    val maxTaskInputInfo = analysis.maxTaskInputSizes
     val sqlIdAlign = if (outputAlignedSQLIds) {
       collect.getSQLCleanAndAligned
     } else {
@@ -305,13 +304,9 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
     val failedTasks = healthCheck.getFailedTasks
     val failedStages = healthCheck.getFailedStages
 
-    // Compute AutoTuner inputs for application_tuning_metrics.csv
+    // Compute GPU OOM signals for tuning_signals.csv
     val singleApp = analyzedApps.head
     val pluginEnabled = singleApp.gpuMode
-    val maxTaskInput = analysis.maxTaskInputSizes.headOption
-      .map(_.maxTaskInputBytesRead.toLong).getOrElse(0L)
-    val maxColumnarExchange =
-      SingleAppSummaryInfoProvider.computeMaxColumnarExchangeDataSizeBytes(sqlMetrics)
     val scanOomStages = SingleAppSummaryInfoProvider.computeScanStagesWithGpuOom(
       pluginEnabled, failedTasks, stageMetrics, singleApp)
     val gpuShuffleContainerOomStages =
@@ -322,7 +317,6 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
     val appInfo = collect.getAppInfo
     val appId = appInfo.headOption.flatMap(_.appId).getOrElse("")
     val tuningSignals = TuningSignalProfileResult.build(
-      maxTaskInput, maxColumnarExchange.getOrElse(0L),
       scanOomStages, gpuShuffleContainerOomStages)
 
     logDebug(s"Time to collect Profiling Info [$appId]: ${endTime - startTime}.")
@@ -349,7 +343,6 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
       sparkProps = collect.getSparkProperties,
       sqlStageInfo = collect.getSQLToStage,
       wholeStage = collect.getWholeStageCodeGenMapping,
-      maxTaskInputBytesRead = maxTaskInputInfo,
       appLogPath = collect.getAppLogPath,
       ioMetrics = analysis.ioAggs,
       sysProps = collect.getSystemProperties,
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/tuning/QualAppSummaryInfoProvider.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/tuning/QualAppSummaryInfoProvider.scala
index 8e681e70e..d6d478167 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/tuning/QualAppSummaryInfoProvider.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/tuning/QualAppSummaryInfoProvider.scala
@@ -98,8 +98,8 @@ class QualAppSummaryInfoProvider(
   }
 
   override def getMaxInput: Double = {
-    if (rawAggMetrics.maxTaskInputSizes.nonEmpty) {
-      rawAggMetrics.maxTaskInputSizes.head.maxTaskInputBytesRead
+    if (rawAggMetrics.sqlAggs.nonEmpty) {
+      rawAggMetrics.sqlAggs.map(_.inputBytesReadMax).max.toDouble
     } else {
       0.0
     }
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
index 297665444..f09b4eec6 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
@@ -126,6 +126,7 @@ object OutHeaderRegistry {
         "executorDeserializeTime_sum",
         "executorRunTime_sum",
         "input_bytesRead_sum",
+        "input_bytesRead_max",
         "input_recordsRead_sum",
         "jvmGCTime_sum",
         "memoryBytesSpilled_sum",
@@ -158,6 +159,7 @@ object OutHeaderRegistry {
         "executorDeserializeTime_sum",
         "executorRunTime_sum",
         "input_bytesRead_sum",
+        "input_bytesRead_max",
         "input_recordsRead_sum",
         "jvmGCTime_sum",
         "memoryBytesSpilled_sum",
@@ -233,6 +235,7 @@ object OutHeaderRegistry {
         "executorDeserializeTime_sum",
         "executorRunTime_sum",
         "input_bytesRead_sum",
+        "input_bytesRead_max",
         "input_recordsRead_sum",
         "jvmGCTime_sum",
         "memoryBytesSpilled_sum",
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
index c6f9a0cc5..e7002d198 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
@@ -37,7 +37,6 @@ object QualRawReportGenerator extends Logging {
       AggMetricsResultSorter.sortSqlAgg(aggRawResult.sqlAggs),
       AggMetricsResultSorter.sortIO(aggRawResult.ioAggs),
       AggMetricsResultSorter.sortSqlDurationAgg(aggRawResult.sqlDurAggs),
-      aggRawResult.maxTaskInputSizes,
       AggMetricsResultSorter.sortStageDiagnostics(aggRawResult.stageDiagnostics))
     Map(
       STAGE_AGG_LABEL -> sortedRes.stageAggs,
@@ -71,12 +70,8 @@ object QualRawReportGenerator extends Logging {
     val pWriter =
       new ProfileOutputWriter(metricsDirectory, "profile", 10000000, outputCSV = true)
     try {
-      // Compute aggregate metrics early so maxTaskInputBytesRead is available for
-      // application_tuning_metrics.csv
       val aggRawMetrics = QualSparkMetricsAggregator
         .getAggRawMetrics(app, sqlAnalyzer = Some(sqlPlanAnalyzer))
-      val maxTaskInput = aggRawMetrics.maxTaskInputSizes.headOption
-        .map(_.maxTaskInputBytesRead.toLong).getOrElse(0L)
 
       pWriter.writeText("### A. Information Collected ###")
       pWriter.writeTable(
@@ -106,9 +101,9 @@ object QualRawReportGenerator extends Logging {
       constructLabelsMaps(aggRawMetrics).foreach { case (label, metrics) =>
           pWriter.writeCSVTable(label, metrics)
       }
-      // GPU-only signals default to 0/empty for qualification (CPU event logs)
+      // GPU-only signals default to empty for qualification (CPU event logs)
       val tuningSignals = TuningSignalProfileResult.build(
-        maxTaskInput, 0L, Set.empty[Long], Set.empty[Long])
+        Set.empty[Long], Set.empty[Long])
       pWriter.writeCSVTable(TUNING_SIGNALS, tuningSignals)
       pWriter.writeText("\n### C. Health Check###\n")
       pWriter.writeCSVTable(QualFailedTaskView.getLabel, QualFailedTaskView.getRawView(Seq(app)))
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/RawMetricProfView.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/RawMetricProfView.scala
index 2d10c8f36..a55caaf9b 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/RawMetricProfView.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/RawMetricProfView.scala
@@ -17,7 +17,7 @@
 package com.nvidia.spark.rapids.tool.views
 
 import com.nvidia.spark.rapids.tool.analysis.ProfSparkMetricsAggregator
-import com.nvidia.spark.rapids.tool.profiling.{IOAnalysisProfileResult, JobAggTaskMetricsProfileResult, ShuffleSkewProfileResult, SQLDurationExecutorTimeProfileResult, SQLMaxTaskInputSizes, SQLTaskAggMetricsProfileResult, StageAggTaskMetricsProfileResult, StageDiagnosticResult}
+import com.nvidia.spark.rapids.tool.profiling.{IOAnalysisProfileResult, JobAggTaskMetricsProfileResult, ShuffleSkewProfileResult, SQLDurationExecutorTimeProfileResult, SQLTaskAggMetricsProfileResult, StageAggTaskMetricsProfileResult, StageDiagnosticResult}
 
 import org.apache.spark.sql.rapids.tool.profiling.ApplicationInfo
 
@@ -30,7 +30,6 @@ case class ProfilerAggregatedView(
     sqlAggs: Seq[SQLTaskAggMetricsProfileResult],
     ioAggs: Seq[IOAnalysisProfileResult],
     sqlDurAggs: Seq[SQLDurationExecutorTimeProfileResult],
-    maxTaskInputSizes: Seq[SQLMaxTaskInputSizes],
     stageDiagnostics: Seq[StageDiagnosticResult])
 
 object RawMetricProfilerView  {
@@ -43,7 +42,6 @@ object RawMetricProfilerView  {
       AggMetricsResultSorter.sortSqlAgg(aggMetricsResults.sqlAggs),
       AggMetricsResultSorter.sortIO(aggMetricsResults.ioAggs),
       AggMetricsResultSorter.sortSqlDurationAgg(aggMetricsResults.sqlDurAggs),
-      aggMetricsResults.maxTaskInputSizes,
       AggMetricsResultSorter.sortStageDiagnostics(aggMetricsResults.stageDiagnostics))
   }
 }

From cf9083c68b9469280f3b7f7d4bd56deaa0d8e0b0 Mon Sep 17 00:00:00 2001
From: Partho Sarthi <psarthi@nvidia.com>
Date: Wed, 22 Apr 2026 16:36:38 -0700
Subject: [PATCH 13/16] Regenerate AnalysisSuite golden CSVs for new
 input_bytesRead_max column

Updated the 9 expectation CSVs under ProfilingExpectations/ to include
the new inputBytesReadMax column produced by the stage/SQL/job task
metric aggregates. Values were regenerated by running AnalysisSuite
against the existing event logs and writing the actual DataFrames back
to the expectation files.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Signed-off-by: Partho Sarthi <psarthi@nvidia.com>
---
 ...on_db_13_3_job_metrics_agg_expectation.csv | 116 +++++++++---------
 ...on_db_13_3_sql_metrics_agg_expectation.csv |   4 +-
 ..._db_13_3_stage_metrics_agg_expectation.csv | 116 +++++++++---------
 ...in_eventlog_jobmetricsagg2_expectation.csv |   4 +-
 ...oin_eventlog_jobmetricsagg_expectation.csv |   4 +-
 ...in_eventlog_sqlmetricsagg2_expectation.csv |   4 +-
 ...oin_eventlog_sqlmetricsagg_expectation.csv |   4 +-
 ..._eventlog_stagemetricsagg2_expectation.csv |  10 +-
 ...n_eventlog_stagemetricsagg_expectation.csv |  10 +-
 9 files changed, 136 insertions(+), 136 deletions(-)

diff --git a/core/src/test/resources/ProfilingExpectations/nds_q88_photon_db_13_3_job_metrics_agg_expectation.csv b/core/src/test/resources/ProfilingExpectations/nds_q88_photon_db_13_3_job_metrics_agg_expectation.csv
index 26d62a885..a400bd298 100644
--- a/core/src/test/resources/ProfilingExpectations/nds_q88_photon_db_13_3_job_metrics_agg_expectation.csv
+++ b/core/src/test/resources/ProfilingExpectations/nds_q88_photon_db_13_3_job_metrics_agg_expectation.csv
@@ -1,58 +1,58 @@
-jobId,numTasks,Duration,diskBytesSpilled_sum,duration_sum,duration_max,duration_min,duration_avg,executorCPUTime_sum,executorDeserializeCPUTime_sum,executorDeserializeTime_sum,executorRunTime_sum,input_bytesRead_sum,input_recordsRead_sum,jvmGCTime_sum,memoryBytesSpilled_sum,output_bytesWritten_sum,output_recordsWritten_sum,peakExecutionMemory_max,resultSerializationTime_sum,resultSize_max,sr_fetchWaitTime_sum,sr_localBlocksFetched_sum,sr_localBytesRead_sum,sr_remoteBlocksFetched_sum,sr_remoteBytesRead_sum,sr_remoteBytesReadToDisk_sum,sr_totalBytesRead_sum,sw_bytesWritten_sum,sw_recordsWritten_sum,sw_writeTime_sum
-48,431,237976,0,371230,1032,333,861.3,343507,1365,1286,367785,190131859,8639936081,160,0,0,0,169667947,21,15767,8,3,15657,3,15657,0,31314,18964,431,16
-47,432,214657,0,376777,1133,499,872.2,346447,1388,1310,373271,14007280,8639936081,144,0,0,0,159530057,11,15577,4,3,15657,3,15657,0,31314,19008,432,20
-46,433,191440,0,457364,3323,391,1056.3,352977,1639,2509,451358,5242628040,8639936081,1912,0,0,0,250840493,9,16203,551,3,15657,3,15657,0,31314,19052,433,16
-49,1,186241,0,266,266,266,266.0,86,1,1,261,0,0,0,0,0,0,138414192,0,5344,10,209,9196,222,9768,0,18964,44,1,0
-45,433,166081,0,415849,1448,339,960.4,350015,1415,1375,412139,2276478144,8639936081,568,0,0,0,195992906,2,15780,7,3,15657,3,15657,0,31314,19052,433,34
-44,431,139667,0,398973,1403,365,925.7,354332,1420,1327,395265,1075691986,8639936081,328,0,0,0,188587155,0,15767,10,3,15657,3,15657,0,31314,18964,431,17
-50,1,122711,0,267,267,267,267.0,71,1,1,262,0,0,0,0,0,0,138414192,0,5343,58,219,9636,213,9372,0,19008,44,1,0
-43,432,114755,0,403652,1369,329,934.4,353529,1424,1326,399766,1395949742,8639936081,624,0,0,0,201771890,13,15767,14,3,15657,3,15657,0,31314,19008,432,16
-51,1,97958,0,386,386,386,386.0,60,1,1,381,0,0,0,0,0,0,138414192,0,5343,154,221,9724,210,9240,0,18964,44,1,0
-42,431,89634,0,616500,1899,589,1430.4,378287,1521,1515,612098,16461920726,8639936081,4132,0,0,0,216740322,23,15805,10,3,15657,3,15657,0,31314,18964,431,16
-52,1,71718,0,384,384,384,384.0,54,1,1,379,0,0,0,0,0,0,138414192,0,5343,170,223,9812,210,9240,0,19052,44,1,0
-41,431,51085,0,759623,2321,918,1762.5,395214,1706,2027,754015,26337468742,8639936081,7772,0,0,0,250648581,87,16157,170,3,15657,3,15657,0,31314,18964,431,19
-53,1,46297,0,136,136,136,136.0,57,1,1,131,0,0,0,0,0,0,138414192,0,5344,0,214,9416,219,9636,0,19052,44,1,0
-54,1,23051,0,340,340,340,340.0,36,1,1,334,0,0,0,0,0,0,138414192,0,5343,223,215,9460,217,9548,0,19008,44,1,0
-31,1,6979,0,6738,6738,6738,6738.0,5104,128,688,6035,349526,86400,53,0,0,0,155563380,1,10759,0,0,0,0,0,0,0,7239,1800,0
-34,1,6953,0,6725,6725,6725,6725.0,479,185,677,6036,349526,86400,53,0,0,0,155563380,0,9814,0,0,0,0,0,0,0,7239,1800,0
-33,1,6940,0,6729,6729,6729,6729.0,206,216,679,6035,349526,86400,53,0,0,0,155563380,1,9896,0,0,0,0,0,0,0,7239,1800,0
-35,1,6925,0,6729,6729,6729,6729.0,157,136,681,6035,12261,1350,53,0,0,0,155199546,1,9839,0,0,0,0,0,0,0,699,165,0
-38,1,6855,0,6743,6743,6743,6743.0,187,256,688,6035,349526,86400,53,0,0,0,155563380,1,9927,0,0,0,0,0,0,0,7239,1800,0
-0,1,6033,0,5699,5699,5699,5699.0,422,948,1114,4382,0,0,37,0,0,0,0,8,2794,0,0,0,0,0,0,0,0,0,0
-13,200,5707,0,87661,966,349,438.3,9924,528,951,84265,0,0,144,0,0,0,0,9,6258,0,0,0,0,0,0,0,0,0,0
-23,200,5479,0,84240,490,355,421.2,5394,292,214,82784,0,0,136,0,0,0,0,0,6214,0,0,0,0,0,0,0,0,0,0
-21,200,5271,0,80904,485,353,404.5,6100,304,220,79384,0,0,136,0,0,0,0,1,6302,0,0,0,0,0,0,0,0,0,0
-27,200,4728,0,70760,442,309,353.8,4145,287,209,69494,0,0,152,0,0,0,0,10,5788,0,0,0,0,0,0,0,0,0,0
-3,1,4708,0,4693,4693,4693,4693.0,280,701,804,3796,0,0,26,0,0,0,0,7,2834,0,0,0,0,0,0,0,0,0,0
-25,200,4603,0,70379,569,314,351.9,4200,294,216,69040,0,0,168,0,0,0,0,14,5708,0,0,0,0,0,0,0,0,0,0
-36,1,4556,0,4332,4332,4332,4332.0,3359,95,401,3907,30328,7200,39,0,0,0,155245068,1,10552,0,0,0,0,0,0,0,7719,1920,0
-29,200,4555,0,69682,423,310,348.4,3830,272,218,68521,0,0,168,0,0,0,0,9,5748,0,0,0,0,0,0,0,0,0,0
-32,1,4515,0,4334,4334,4334,4334.0,260,130,404,3907,349526,86400,39,0,0,0,155563380,1,9851,0,0,0,0,0,0,0,7239,1800,0
-39,1,4488,0,4322,4322,4322,4322.0,112,124,392,3907,349526,86400,39,0,0,0,155563380,1,9926,0,0,0,0,0,0,0,7239,1800,0
-37,1,4481,0,4334,4334,4334,4334.0,136,144,405,3907,349526,86400,39,0,0,0,155563380,1,9851,0,0,0,0,0,0,0,7239,1800,0
-40,1,4476,0,4327,4327,4327,4327.0,98,147,394,3907,349526,86400,39,0,0,0,155563380,1,9895,0,0,0,0,0,0,0,7239,1800,0
-56,1,1055,0,1022,1022,1022,1022.0,758,77,95,901,0,0,0,0,0,0,134218344,6,10091,5,218,9592,220,9680,0,19272,0,0,0
-19,200,803,0,11895,145,38,59.5,1050,321,252,10017,0,0,56,0,0,0,0,22,2739,0,0,0,0,0,0,0,0,0,0
-26,1,316,0,312,312,312,312.0,2,1,1,306,0,0,0,0,0,0,0,0,3777,0,0,0,0,0,0,0,0,0,0
-2,1,280,0,267,267,267,267.0,6,4,4,124,0,0,0,0,0,0,0,1,3342,0,0,0,0,0,0,0,0,0,0
-11,1,264,0,254,254,254,254.0,5,3,3,241,0,0,0,0,0,0,0,0,2913,0,0,0,0,0,0,0,0,0,0
-7,1,240,0,227,227,227,227.0,5,4,4,213,0,0,114,0,0,0,0,0,2206,0,0,0,0,0,0,0,0,0,0
-1,1,209,0,173,173,173,173.0,28,5,6,152,0,0,0,0,0,0,0,0,2506,0,0,0,0,0,0,0,0,0,0
-5,1,179,0,165,165,165,165.0,4,4,4,151,0,0,0,0,0,0,0,0,2475,0,0,0,0,0,0,0,0,0,0
-14,1,151,0,143,143,143,143.0,3,1,1,132,0,0,0,0,0,0,0,1,3120,0,0,0,0,0,0,0,0,0,0
-4,1,147,0,139,139,139,139.0,22,5,6,121,0,0,0,0,0,0,0,0,2334,0,0,0,0,0,0,0,0,0,0
-20,1,141,0,137,137,137,137.0,1,1,1,130,0,0,0,0,0,0,0,0,2170,0,0,0,0,0,0,0,0,0,0
-28,1,140,0,136,136,136,136.0,2,1,1,130,0,0,0,0,0,0,0,0,3784,0,0,0,0,0,0,0,0,0,0
-18,1,129,0,124,124,124,124.0,2,1,1,116,0,0,0,0,0,0,0,0,2501,0,0,0,0,0,0,0,0,0,0
-16,1,125,0,117,117,117,117.0,2,1,1,108,0,0,0,0,0,0,0,0,2758,0,0,0,0,0,0,0,0,0,0
-6,1,123,0,113,113,113,113.0,4,3,3,100,0,0,0,0,0,0,0,0,2208,0,0,0,0,0,0,0,0,0,0
-10,1,120,0,110,110,110,110.0,6,3,3,98,0,0,0,0,0,0,0,0,3565,0,0,0,0,0,0,0,0,0,0
-9,1,114,0,104,104,104,104.0,5,3,3,90,0,0,0,0,0,0,0,1,3514,0,0,0,0,0,0,0,0,0,0
-12,1,105,0,85,85,85,85.0,4,3,3,72,0,0,0,0,0,0,0,0,3369,0,0,0,0,0,0,0,0,0,0
-17,1,103,0,97,97,97,97.0,2,2,2,89,0,0,0,0,0,0,0,0,3003,0,0,0,0,0,0,0,0,0,0
-8,1,102,0,95,95,95,95.0,4,3,3,82,0,0,0,0,0,0,0,0,3142,0,0,0,0,0,0,0,0,0,0
-30,1,73,0,67,67,67,67.0,2,1,1,62,0,0,0,0,0,0,0,0,3199,0,0,0,0,0,0,0,0,0,0
-24,1,72,0,59,59,59,59.0,2,1,1,51,0,0,0,0,0,0,0,0,3288,0,0,0,0,0,0,0,0,0,0
-22,1,70,0,65,65,65,65.0,2,1,1,59,0,0,0,0,0,0,0,0,3436,0,0,0,0,0,0,0,0,0,0
-55,1,65,0,54,54,54,54.0,27,1,1,49,0,0,0,0,0,0,138414192,0,5343,0,216,9504,215,9460,0,18964,44,1,0
-15,1,64,0,58,58,58,58.0,2,1,1,50,0,0,0,0,0,0,0,0,2306,0,0,0,0,0,0,0,0,0,0
+id,numTasks,duration,diskBytesSpilledSum,durationSum,durationMax,durationMin,durationAvg,executorCPUTimeSum,executorDeserializeCpuTimeSum,executorDeserializeTimeSum,executorRunTimeSum,inputBytesReadSum,inputBytesReadMax,inputRecordsReadSum,jvmGCTimeSum,memoryBytesSpilledSum,outputBytesWrittenSum,outputRecordsWrittenSum,peakExecutionMemoryMax,resultSerializationTimeSum,resultSizeMax,srFetchWaitTimeSum,srLocalBlocksFetchedSum,srcLocalBytesReadSum,srRemoteBlocksFetchSum,srRemoteBytesReadSum,srRemoteBytesReadToDiskSum,srTotalBytesReadSum,swBytesWrittenSum,swRecordsWrittenSum,swWriteTimeSum
+48,431,237976,0,371230,1032,333,861.3,343507,1365,1286,367785,190131859,18373883,8639936081,160,0,0,0,169667947,21,15767,8,3,15657,3,15657,0,31314,18964,431,16
+47,432,214657,0,376777,1133,499,872.2,346447,1388,1310,373271,14007280,8235993,8639936081,144,0,0,0,159530057,11,15577,4,3,15657,3,15657,0,31314,19008,432,20
+46,433,191440,0,457364,3323,391,1056.3,352977,1639,2509,451358,5242628040,101021117,8639936081,1912,0,0,0,250840493,9,16203,551,3,15657,3,15657,0,31314,19052,433,16
+49,1,186241,0,266,266,266,266.0,86,1,1,261,0,0,0,0,0,0,0,138414192,0,5344,10,209,9196,222,9768,0,18964,44,1,0
+45,433,166081,0,415849,1448,339,960.4,350015,1415,1375,412139,2276478144,44711594,8639936081,568,0,0,0,195992906,2,15780,7,3,15657,3,15657,0,31314,19052,433,34
+44,431,139667,0,398973,1403,365,925.7,354332,1420,1327,395265,1075691986,37555235,8639936081,328,0,0,0,188587155,0,15767,10,3,15657,3,15657,0,31314,18964,431,17
+50,1,122711,0,267,267,267,267.0,71,1,1,262,0,0,0,0,0,0,0,138414192,0,5343,58,219,9636,213,9372,0,19008,44,1,0
+43,432,114755,0,403652,1369,329,934.4,353529,1424,1326,399766,1395949742,50513106,8639936081,624,0,0,0,201771890,13,15767,14,3,15657,3,15657,0,31314,19008,432,16
+51,1,97958,0,386,386,386,386.0,60,1,1,381,0,0,0,0,0,0,0,138414192,0,5343,154,221,9724,210,9240,0,18964,44,1,0
+42,431,89634,0,616500,1899,589,1430.4,378287,1521,1515,612098,16461920726,65476865,8639936081,4132,0,0,0,216740322,23,15805,10,3,15657,3,15657,0,31314,18964,431,16
+52,1,71718,0,384,384,384,384.0,54,1,1,379,0,0,0,0,0,0,0,138414192,0,5343,170,223,9812,210,9240,0,19052,44,1,0
+41,431,51085,0,759623,2321,918,1762.5,395214,1706,2027,754015,26337468742,99616661,8639936081,7772,0,0,0,250648581,87,16157,170,3,15657,3,15657,0,31314,18964,431,19
+53,1,46297,0,136,136,136,136.0,57,1,1,131,0,0,0,0,0,0,0,138414192,0,5344,0,214,9416,219,9636,0,19052,44,1,0
+54,1,23051,0,340,340,340,340.0,36,1,1,334,0,0,0,0,0,0,0,138414192,0,5343,223,215,9460,217,9548,0,19008,44,1,0
+31,1,6979,0,6738,6738,6738,6738.0,5104,128,688,6035,349526,349526,86400,53,0,0,0,155563380,1,10759,0,0,0,0,0,0,0,7239,1800,0
+34,1,6953,0,6725,6725,6725,6725.0,479,185,677,6036,349526,349526,86400,53,0,0,0,155563380,0,9814,0,0,0,0,0,0,0,7239,1800,0
+33,1,6940,0,6729,6729,6729,6729.0,206,216,679,6035,349526,349526,86400,53,0,0,0,155563380,1,9896,0,0,0,0,0,0,0,7239,1800,0
+35,1,6925,0,6729,6729,6729,6729.0,157,136,681,6035,12261,12261,1350,53,0,0,0,155199546,1,9839,0,0,0,0,0,0,0,699,165,0
+38,1,6855,0,6743,6743,6743,6743.0,187,256,688,6035,349526,349526,86400,53,0,0,0,155563380,1,9927,0,0,0,0,0,0,0,7239,1800,0
+0,1,6033,0,5699,5699,5699,5699.0,422,948,1114,4382,0,0,0,37,0,0,0,0,8,2794,0,0,0,0,0,0,0,0,0,0
+13,200,5707,0,87661,966,349,438.3,9924,528,951,84265,0,0,0,144,0,0,0,0,9,6258,0,0,0,0,0,0,0,0,0,0
+23,200,5479,0,84240,490,355,421.2,5394,292,214,82784,0,0,0,136,0,0,0,0,0,6214,0,0,0,0,0,0,0,0,0,0
+21,200,5271,0,80904,485,353,404.5,6100,304,220,79384,0,0,0,136,0,0,0,0,1,6302,0,0,0,0,0,0,0,0,0,0
+27,200,4728,0,70760,442,309,353.8,4145,287,209,69494,0,0,0,152,0,0,0,0,10,5788,0,0,0,0,0,0,0,0,0,0
+3,1,4708,0,4693,4693,4693,4693.0,280,701,804,3796,0,0,0,26,0,0,0,0,7,2834,0,0,0,0,0,0,0,0,0,0
+25,200,4603,0,70379,569,314,351.9,4200,294,216,69040,0,0,0,168,0,0,0,0,14,5708,0,0,0,0,0,0,0,0,0,0
+36,1,4556,0,4332,4332,4332,4332.0,3359,95,401,3907,30328,30328,7200,39,0,0,0,155245068,1,10552,0,0,0,0,0,0,0,7719,1920,0
+29,200,4555,0,69682,423,310,348.4,3830,272,218,68521,0,0,0,168,0,0,0,0,9,5748,0,0,0,0,0,0,0,0,0,0
+32,1,4515,0,4334,4334,4334,4334.0,260,130,404,3907,349526,349526,86400,39,0,0,0,155563380,1,9851,0,0,0,0,0,0,0,7239,1800,0
+39,1,4488,0,4322,4322,4322,4322.0,112,124,392,3907,349526,349526,86400,39,0,0,0,155563380,1,9926,0,0,0,0,0,0,0,7239,1800,0
+37,1,4481,0,4334,4334,4334,4334.0,136,144,405,3907,349526,349526,86400,39,0,0,0,155563380,1,9851,0,0,0,0,0,0,0,7239,1800,0
+40,1,4476,0,4327,4327,4327,4327.0,98,147,394,3907,349526,349526,86400,39,0,0,0,155563380,1,9895,0,0,0,0,0,0,0,7239,1800,0
+56,1,1055,0,1022,1022,1022,1022.0,758,77,95,901,0,0,0,0,0,0,0,134218344,6,10091,5,218,9592,220,9680,0,19272,0,0,0
+19,200,803,0,11895,145,38,59.5,1050,321,252,10017,0,0,0,56,0,0,0,0,22,2739,0,0,0,0,0,0,0,0,0,0
+26,1,316,0,312,312,312,312.0,2,1,1,306,0,0,0,0,0,0,0,0,0,3777,0,0,0,0,0,0,0,0,0,0
+2,1,280,0,267,267,267,267.0,6,4,4,124,0,0,0,0,0,0,0,0,1,3342,0,0,0,0,0,0,0,0,0,0
+11,1,264,0,254,254,254,254.0,5,3,3,241,0,0,0,0,0,0,0,0,0,2913,0,0,0,0,0,0,0,0,0,0
+7,1,240,0,227,227,227,227.0,5,4,4,213,0,0,0,114,0,0,0,0,0,2206,0,0,0,0,0,0,0,0,0,0
+1,1,209,0,173,173,173,173.0,28,5,6,152,0,0,0,0,0,0,0,0,0,2506,0,0,0,0,0,0,0,0,0,0
+5,1,179,0,165,165,165,165.0,4,4,4,151,0,0,0,0,0,0,0,0,0,2475,0,0,0,0,0,0,0,0,0,0
+14,1,151,0,143,143,143,143.0,3,1,1,132,0,0,0,0,0,0,0,0,1,3120,0,0,0,0,0,0,0,0,0,0
+4,1,147,0,139,139,139,139.0,22,5,6,121,0,0,0,0,0,0,0,0,0,2334,0,0,0,0,0,0,0,0,0,0
+20,1,141,0,137,137,137,137.0,1,1,1,130,0,0,0,0,0,0,0,0,0,2170,0,0,0,0,0,0,0,0,0,0
+28,1,140,0,136,136,136,136.0,2,1,1,130,0,0,0,0,0,0,0,0,0,3784,0,0,0,0,0,0,0,0,0,0
+18,1,129,0,124,124,124,124.0,2,1,1,116,0,0,0,0,0,0,0,0,0,2501,0,0,0,0,0,0,0,0,0,0
+16,1,125,0,117,117,117,117.0,2,1,1,108,0,0,0,0,0,0,0,0,0,2758,0,0,0,0,0,0,0,0,0,0
+6,1,123,0,113,113,113,113.0,4,3,3,100,0,0,0,0,0,0,0,0,0,2208,0,0,0,0,0,0,0,0,0,0
+10,1,120,0,110,110,110,110.0,6,3,3,98,0,0,0,0,0,0,0,0,0,3565,0,0,0,0,0,0,0,0,0,0
+9,1,114,0,104,104,104,104.0,5,3,3,90,0,0,0,0,0,0,0,0,1,3514,0,0,0,0,0,0,0,0,0,0
+12,1,105,0,85,85,85,85.0,4,3,3,72,0,0,0,0,0,0,0,0,0,3369,0,0,0,0,0,0,0,0,0,0
+17,1,103,0,97,97,97,97.0,2,2,2,89,0,0,0,0,0,0,0,0,0,3003,0,0,0,0,0,0,0,0,0,0
+8,1,102,0,95,95,95,95.0,4,3,3,82,0,0,0,0,0,0,0,0,0,3142,0,0,0,0,0,0,0,0,0,0
+30,1,73,0,67,67,67,67.0,2,1,1,62,0,0,0,0,0,0,0,0,0,3199,0,0,0,0,0,0,0,0,0,0
+24,1,72,0,59,59,59,59.0,2,1,1,51,0,0,0,0,0,0,0,0,0,3288,0,0,0,0,0,0,0,0,0,0
+22,1,70,0,65,65,65,65.0,2,1,1,59,0,0,0,0,0,0,0,0,0,3436,0,0,0,0,0,0,0,0,0,0
+55,1,65,0,54,54,54,54.0,27,1,1,49,0,0,0,0,0,0,0,138414192,0,5343,0,216,9504,215,9460,0,18964,44,1,0
+15,1,64,0,58,58,58,58.0,2,1,1,50,0,0,0,0,0,0,0,0,0,2306,0,0,0,0,0,0,0,0,0,0
diff --git a/core/src/test/resources/ProfilingExpectations/nds_q88_photon_db_13_3_sql_metrics_agg_expectation.csv b/core/src/test/resources/ProfilingExpectations/nds_q88_photon_db_13_3_sql_metrics_agg_expectation.csv
index 36f5fe11d..26517710f 100644
--- a/core/src/test/resources/ProfilingExpectations/nds_q88_photon_db_13_3_sql_metrics_agg_expectation.csv
+++ b/core/src/test/resources/ProfilingExpectations/nds_q88_photon_db_13_3_sql_metrics_agg_expectation.csv
@@ -1,2 +1,2 @@
-appID,sqlID,description,numTasks,Duration,executorCPURatio,diskBytesSpilled_sum,duration_sum,duration_max,duration_min,duration_avg,executorCPUTime_sum,executorDeserializeCPUTime_sum,executorDeserializeTime_sum,executorRunTime_sum,input_bytesRead_sum,input_recordsRead_sum,jvmGCTime_sum,memoryBytesSpilled_sum,output_bytesWritten_sum,output_recordsWritten_sum,peakExecutionMemory_max,resultSerializationTime_sum,resultSize_max,sr_fetchWaitTime_sum,sr_localBlocksFetched_sum,sr_localBytesRead_sum,sr_remoteBlocksFetched_sum,sr_remoteBytesRead_sum,sr_remoteBytesReadToDisk_sum,sr_totalBytesRead_sum,sw_bytesWritten_sum,sw_recordsWritten_sum,sw_writeTime_sum
-"app-20240919162642-0000",26,"query88",3472,250542,75.58,0,3858136,6743,54,1111.2,2885555,13523,18186,3818106,52997115316,69120188398,16100,0,0,0,250840493,181,16203,1394,1759,201596,1750,201200,0,402796,218614,19946,154
+appId,sqlId,description,numTasks,duration,executorCpuRatio,diskBytesSpilledSum,durationSum,durationMax,durationMin,durationAvg,executorCPUTimeSum,executorDeserializeCpuTimeSum,executorDeserializeTimeSum,executorRunTimeSum,inputBytesReadSum,inputBytesReadMax,inputRecordsReadSum,jvmGCTimeSum,memoryBytesSpilledSum,outputBytesWrittenSum,outputRecordsWrittenSum,peakExecutionMemoryMax,resultSerializationTimeSum,resultSizeMax,srFetchWaitTimeSum,srLocalBlocksFetchedSum,srcLocalBytesReadSum,srRemoteBlocksFetchSum,srRemoteBytesReadSum,srRemoteBytesReadToDiskSum,srTotalBytesReadSum,swBytesWrittenSum,swRecordsWrittenSum,swWriteTimeSum
+app-20240919162642-0000,26,query88,3472,250542,75.58,0,3858136,6743,54,1111.2,2885555,13523,18186,3818106,52997115316,101021117,69120188398,16100,0,0,0,250840493,181,16203,1394,1759,201596,1750,201200,0,402796,218614,19946,154
diff --git a/core/src/test/resources/ProfilingExpectations/nds_q88_photon_db_13_3_stage_metrics_agg_expectation.csv b/core/src/test/resources/ProfilingExpectations/nds_q88_photon_db_13_3_stage_metrics_agg_expectation.csv
index 0b7fd2182..7ff8071d8 100644
--- a/core/src/test/resources/ProfilingExpectations/nds_q88_photon_db_13_3_stage_metrics_agg_expectation.csv
+++ b/core/src/test/resources/ProfilingExpectations/nds_q88_photon_db_13_3_stage_metrics_agg_expectation.csv
@@ -1,58 +1,58 @@
-stageId,numTasks,Duration,diskBytesSpilled_sum,duration_sum,duration_max,duration_min,duration_avg,executorCPUTime_sum,executorDeserializeCPUTime_sum,executorDeserializeTime_sum,executorRunTime_sum,input_bytesRead_sum,input_recordsRead_sum,jvmGCTime_sum,memoryBytesSpilled_sum,output_bytesWritten_sum,output_recordsWritten_sum,peakExecutionMemory_max,resultSerializationTime_sum,resultSize_max,sr_fetchWaitTime_sum,sr_localBlocksFetched_sum,sr_localBytesRead_sum,sr_remoteBlocksFetched_sum,sr_remoteBytesRead_sum,sr_remoteBytesReadToDisk_sum,sr_totalBytesRead_sum,sw_bytesWritten_sum,sw_recordsWritten_sum,sw_writeTime_sum
-58,431,237799,0,371230,1032,333,861.3,343507,1365,1286,367785,190131859,8639936081,160,0,0,0,169667947,21,15767,8,3,15657,3,15657,0,31314,18964,431,16
-54,432,214633,0,376777,1133,499,872.2,346447,1388,1310,373271,14007280,8639936081,144,0,0,0,159530057,11,15577,4,3,15657,3,15657,0,31314,19008,432,20
-44,433,191384,0,457364,3323,391,1056.3,352977,1639,2509,451358,5242628040,8639936081,1912,0,0,0,250840493,9,16203,551,3,15657,3,15657,0,31314,19052,433,16
-61,1,186240,0,266,266,266,266.0,86,1,1,261,0,0,0,0,0,0,138414192,0,5344,10,209,9196,222,9768,0,18964,44,1,0
-46,433,166015,0,415849,1448,339,960.4,350015,1415,1375,412139,2276478144,8639936081,568,0,0,0,195992906,2,15780,7,3,15657,3,15657,0,31314,19052,433,34
-50,431,139628,0,398973,1403,365,925.7,354332,1420,1327,395265,1075691986,8639936081,328,0,0,0,188587155,0,15767,10,3,15657,3,15657,0,31314,18964,431,17
-64,1,122708,0,267,267,267,267.0,71,1,1,262,0,0,0,0,0,0,138414192,0,5343,58,219,9636,213,9372,0,19008,44,1,0
-48,432,114722,0,403652,1369,329,934.4,353529,1424,1326,399766,1395949742,8639936081,624,0,0,0,201771890,13,15767,14,3,15657,3,15657,0,31314,19008,432,16
-67,1,97957,0,386,386,386,386.0,60,1,1,381,0,0,0,0,0,0,138414192,0,5343,154,221,9724,210,9240,0,18964,44,1,0
-56,431,89600,0,616500,1899,589,1430.4,378287,1521,1515,612098,16461920726,8639936081,4132,0,0,0,216740322,23,15805,10,3,15657,3,15657,0,31314,18964,431,16
-70,1,71716,0,384,384,384,384.0,54,1,1,379,0,0,0,0,0,0,138414192,0,5343,170,223,9812,210,9240,0,19052,44,1,0
-52,431,51060,0,759623,2321,918,1762.5,395214,1706,2027,754015,26337468742,8639936081,7772,0,0,0,250648581,87,16157,170,3,15657,3,15657,0,31314,18964,431,19
-73,1,46297,0,136,136,136,136.0,57,1,1,131,0,0,0,0,0,0,138414192,0,5344,0,214,9416,219,9636,0,19052,44,1,0
-76,1,23048,0,340,340,340,340.0,36,1,1,334,0,0,0,0,0,0,138414192,0,5343,223,215,9460,217,9548,0,19008,44,1,0
-31,1,6956,0,6738,6738,6738,6738.0,5104,128,688,6035,349526,86400,53,0,0,0,155563380,1,10759,0,0,0,0,0,0,0,7239,1800,0
-32,1,6945,0,6725,6725,6725,6725.0,479,185,677,6036,349526,86400,53,0,0,0,155563380,0,9814,0,0,0,0,0,0,0,7239,1800,0
-33,1,6930,0,6729,6729,6729,6729.0,206,216,679,6035,349526,86400,53,0,0,0,155563380,1,9896,0,0,0,0,0,0,0,7239,1800,0
-34,1,6907,0,6729,6729,6729,6729.0,157,136,681,6035,12261,1350,53,0,0,0,155199546,1,9839,0,0,0,0,0,0,0,699,165,0
-38,1,6842,0,6743,6743,6743,6743.0,187,256,688,6035,349526,86400,53,0,0,0,155563380,1,9927,0,0,0,0,0,0,0,7239,1800,0
-0,1,5904,0,5699,5699,5699,5699.0,422,948,1114,4382,0,0,37,0,0,0,0,8,2794,0,0,0,0,0,0,0,0,0,0
-13,200,5697,0,87661,966,349,438.3,9924,528,951,84265,0,0,144,0,0,0,0,9,6258,0,0,0,0,0,0,0,0,0,0
-23,200,5476,0,84240,490,355,421.2,5394,292,214,82784,0,0,136,0,0,0,0,0,6214,0,0,0,0,0,0,0,0,0,0
-21,200,5265,0,80904,485,353,404.5,6100,304,220,79384,0,0,136,0,0,0,0,1,6302,0,0,0,0,0,0,0,0,0,0
-27,200,4719,0,70760,442,309,353.8,4145,287,209,69494,0,0,152,0,0,0,0,10,5788,0,0,0,0,0,0,0,0,0,0
-3,1,4701,0,4693,4693,4693,4693.0,280,701,804,3796,0,0,26,0,0,0,0,7,2834,0,0,0,0,0,0,0,0,0,0
-25,200,4599,0,70379,569,314,351.9,4200,294,216,69040,0,0,168,0,0,0,0,14,5708,0,0,0,0,0,0,0,0,0,0
-29,200,4552,0,69682,423,310,348.4,3830,272,218,68521,0,0,168,0,0,0,0,9,5748,0,0,0,0,0,0,0,0,0,0
-35,1,4525,0,4332,4332,4332,4332.0,3359,95,401,3907,30328,7200,39,0,0,0,155245068,1,10552,0,0,0,0,0,0,0,7719,1920,0
-36,1,4509,0,4334,4334,4334,4334.0,260,130,404,3907,349526,86400,39,0,0,0,155563380,1,9851,0,0,0,0,0,0,0,7239,1800,0
-39,1,4474,0,4322,4322,4322,4322.0,112,124,392,3907,349526,86400,39,0,0,0,155563380,1,9926,0,0,0,0,0,0,0,7239,1800,0
-40,1,4469,0,4327,4327,4327,4327.0,98,147,394,3907,349526,86400,39,0,0,0,155563380,1,9895,0,0,0,0,0,0,0,7239,1800,0
-37,1,4464,0,4334,4334,4334,4334.0,136,144,405,3907,349526,86400,39,0,0,0,155563380,1,9851,0,0,0,0,0,0,0,7239,1800,0
-107,1,1052,0,1022,1022,1022,1022.0,758,77,95,901,0,0,0,0,0,0,134218344,6,10091,5,218,9592,220,9680,0,19272,0,0,0
-19,200,794,0,11895,145,38,59.5,1050,321,252,10017,0,0,56,0,0,0,0,22,2739,0,0,0,0,0,0,0,0,0,0
-26,1,315,0,312,312,312,312.0,2,1,1,306,0,0,0,0,0,0,0,0,3777,0,0,0,0,0,0,0,0,0,0
-2,1,276,0,267,267,267,267.0,6,4,4,124,0,0,0,0,0,0,0,1,3342,0,0,0,0,0,0,0,0,0,0
-11,1,260,0,254,254,254,254.0,5,3,3,241,0,0,0,0,0,0,0,0,2913,0,0,0,0,0,0,0,0,0,0
-7,1,238,0,227,227,227,227.0,5,4,4,213,0,0,114,0,0,0,0,0,2206,0,0,0,0,0,0,0,0,0,0
-1,1,202,0,173,173,173,173.0,28,5,6,152,0,0,0,0,0,0,0,0,2506,0,0,0,0,0,0,0,0,0,0
-5,1,177,0,165,165,165,165.0,4,4,4,151,0,0,0,0,0,0,0,0,2475,0,0,0,0,0,0,0,0,0,0
-14,1,148,0,143,143,143,143.0,3,1,1,132,0,0,0,0,0,0,0,1,3120,0,0,0,0,0,0,0,0,0,0
-4,1,145,0,139,139,139,139.0,22,5,6,121,0,0,0,0,0,0,0,0,2334,0,0,0,0,0,0,0,0,0,0
-20,1,140,0,137,137,137,137.0,1,1,1,130,0,0,0,0,0,0,0,0,2170,0,0,0,0,0,0,0,0,0,0
-28,1,139,0,136,136,136,136.0,2,1,1,130,0,0,0,0,0,0,0,0,3784,0,0,0,0,0,0,0,0,0,0
-18,1,127,0,124,124,124,124.0,2,1,1,116,0,0,0,0,0,0,0,0,2501,0,0,0,0,0,0,0,0,0,0
-16,1,122,0,117,117,117,117.0,2,1,1,108,0,0,0,0,0,0,0,0,2758,0,0,0,0,0,0,0,0,0,0
-6,1,120,0,113,113,113,113.0,4,3,3,100,0,0,0,0,0,0,0,0,2208,0,0,0,0,0,0,0,0,0,0
-10,1,118,0,110,110,110,110.0,6,3,3,98,0,0,0,0,0,0,0,0,3565,0,0,0,0,0,0,0,0,0,0
-9,1,110,0,104,104,104,104.0,5,3,3,90,0,0,0,0,0,0,0,1,3514,0,0,0,0,0,0,0,0,0,0
-17,1,102,0,97,97,97,97.0,2,2,2,89,0,0,0,0,0,0,0,0,3003,0,0,0,0,0,0,0,0,0,0
-8,1,100,0,95,95,95,95.0,4,3,3,82,0,0,0,0,0,0,0,0,3142,0,0,0,0,0,0,0,0,0,0
-12,1,98,0,85,85,85,85.0,4,3,3,72,0,0,0,0,0,0,0,0,3369,0,0,0,0,0,0,0,0,0,0
-30,1,71,0,67,67,67,67.0,2,1,1,62,0,0,0,0,0,0,0,0,3199,0,0,0,0,0,0,0,0,0,0
-22,1,68,0,65,65,65,65.0,2,1,1,59,0,0,0,0,0,0,0,0,3436,0,0,0,0,0,0,0,0,0,0
-24,1,68,0,59,59,59,59.0,2,1,1,51,0,0,0,0,0,0,0,0,3288,0,0,0,0,0,0,0,0,0,0
-81,1,63,0,54,54,54,54.0,27,1,1,49,0,0,0,0,0,0,138414192,0,5343,0,216,9504,215,9460,0,18964,44,1,0
-15,1,62,0,58,58,58,58.0,2,1,1,50,0,0,0,0,0,0,0,0,2306,0,0,0,0,0,0,0,0,0,0
+id,numTasks,duration,diskBytesSpilledSum,durationSum,durationMax,durationMin,durationAvg,executorCPUTimeSum,executorDeserializeCpuTimeSum,executorDeserializeTimeSum,executorRunTimeSum,inputBytesReadSum,inputBytesReadMax,inputRecordsReadSum,jvmGCTimeSum,memoryBytesSpilledSum,outputBytesWrittenSum,outputRecordsWrittenSum,peakExecutionMemoryMax,resultSerializationTimeSum,resultSizeMax,srFetchWaitTimeSum,srLocalBlocksFetchedSum,srcLocalBytesReadSum,srRemoteBlocksFetchSum,srRemoteBytesReadSum,srRemoteBytesReadToDiskSum,srTotalBytesReadSum,swBytesWrittenSum,swRecordsWrittenSum,swWriteTimeSum
+58,431,237799,0,371230,1032,333,861.3,343507,1365,1286,367785,190131859,18373883,8639936081,160,0,0,0,169667947,21,15767,8,3,15657,3,15657,0,31314,18964,431,16
+54,432,214633,0,376777,1133,499,872.2,346447,1388,1310,373271,14007280,8235993,8639936081,144,0,0,0,159530057,11,15577,4,3,15657,3,15657,0,31314,19008,432,20
+44,433,191384,0,457364,3323,391,1056.3,352977,1639,2509,451358,5242628040,101021117,8639936081,1912,0,0,0,250840493,9,16203,551,3,15657,3,15657,0,31314,19052,433,16
+61,1,186240,0,266,266,266,266.0,86,1,1,261,0,0,0,0,0,0,0,138414192,0,5344,10,209,9196,222,9768,0,18964,44,1,0
+46,433,166015,0,415849,1448,339,960.4,350015,1415,1375,412139,2276478144,44711594,8639936081,568,0,0,0,195992906,2,15780,7,3,15657,3,15657,0,31314,19052,433,34
+50,431,139628,0,398973,1403,365,925.7,354332,1420,1327,395265,1075691986,37555235,8639936081,328,0,0,0,188587155,0,15767,10,3,15657,3,15657,0,31314,18964,431,17
+64,1,122708,0,267,267,267,267.0,71,1,1,262,0,0,0,0,0,0,0,138414192,0,5343,58,219,9636,213,9372,0,19008,44,1,0
+48,432,114722,0,403652,1369,329,934.4,353529,1424,1326,399766,1395949742,50513106,8639936081,624,0,0,0,201771890,13,15767,14,3,15657,3,15657,0,31314,19008,432,16
+67,1,97957,0,386,386,386,386.0,60,1,1,381,0,0,0,0,0,0,0,138414192,0,5343,154,221,9724,210,9240,0,18964,44,1,0
+56,431,89600,0,616500,1899,589,1430.4,378287,1521,1515,612098,16461920726,65476865,8639936081,4132,0,0,0,216740322,23,15805,10,3,15657,3,15657,0,31314,18964,431,16
+70,1,71716,0,384,384,384,384.0,54,1,1,379,0,0,0,0,0,0,0,138414192,0,5343,170,223,9812,210,9240,0,19052,44,1,0
+52,431,51060,0,759623,2321,918,1762.5,395214,1706,2027,754015,26337468742,99616661,8639936081,7772,0,0,0,250648581,87,16157,170,3,15657,3,15657,0,31314,18964,431,19
+73,1,46297,0,136,136,136,136.0,57,1,1,131,0,0,0,0,0,0,0,138414192,0,5344,0,214,9416,219,9636,0,19052,44,1,0
+76,1,23048,0,340,340,340,340.0,36,1,1,334,0,0,0,0,0,0,0,138414192,0,5343,223,215,9460,217,9548,0,19008,44,1,0
+31,1,6956,0,6738,6738,6738,6738.0,5104,128,688,6035,349526,349526,86400,53,0,0,0,155563380,1,10759,0,0,0,0,0,0,0,7239,1800,0
+32,1,6945,0,6725,6725,6725,6725.0,479,185,677,6036,349526,349526,86400,53,0,0,0,155563380,0,9814,0,0,0,0,0,0,0,7239,1800,0
+33,1,6930,0,6729,6729,6729,6729.0,206,216,679,6035,349526,349526,86400,53,0,0,0,155563380,1,9896,0,0,0,0,0,0,0,7239,1800,0
+34,1,6907,0,6729,6729,6729,6729.0,157,136,681,6035,12261,12261,1350,53,0,0,0,155199546,1,9839,0,0,0,0,0,0,0,699,165,0
+38,1,6842,0,6743,6743,6743,6743.0,187,256,688,6035,349526,349526,86400,53,0,0,0,155563380,1,9927,0,0,0,0,0,0,0,7239,1800,0
+0,1,5904,0,5699,5699,5699,5699.0,422,948,1114,4382,0,0,0,37,0,0,0,0,8,2794,0,0,0,0,0,0,0,0,0,0
+13,200,5697,0,87661,966,349,438.3,9924,528,951,84265,0,0,0,144,0,0,0,0,9,6258,0,0,0,0,0,0,0,0,0,0
+23,200,5476,0,84240,490,355,421.2,5394,292,214,82784,0,0,0,136,0,0,0,0,0,6214,0,0,0,0,0,0,0,0,0,0
+21,200,5265,0,80904,485,353,404.5,6100,304,220,79384,0,0,0,136,0,0,0,0,1,6302,0,0,0,0,0,0,0,0,0,0
+27,200,4719,0,70760,442,309,353.8,4145,287,209,69494,0,0,0,152,0,0,0,0,10,5788,0,0,0,0,0,0,0,0,0,0
+3,1,4701,0,4693,4693,4693,4693.0,280,701,804,3796,0,0,0,26,0,0,0,0,7,2834,0,0,0,0,0,0,0,0,0,0
+25,200,4599,0,70379,569,314,351.9,4200,294,216,69040,0,0,0,168,0,0,0,0,14,5708,0,0,0,0,0,0,0,0,0,0
+29,200,4552,0,69682,423,310,348.4,3830,272,218,68521,0,0,0,168,0,0,0,0,9,5748,0,0,0,0,0,0,0,0,0,0
+35,1,4525,0,4332,4332,4332,4332.0,3359,95,401,3907,30328,30328,7200,39,0,0,0,155245068,1,10552,0,0,0,0,0,0,0,7719,1920,0
+36,1,4509,0,4334,4334,4334,4334.0,260,130,404,3907,349526,349526,86400,39,0,0,0,155563380,1,9851,0,0,0,0,0,0,0,7239,1800,0
+39,1,4474,0,4322,4322,4322,4322.0,112,124,392,3907,349526,349526,86400,39,0,0,0,155563380,1,9926,0,0,0,0,0,0,0,7239,1800,0
+40,1,4469,0,4327,4327,4327,4327.0,98,147,394,3907,349526,349526,86400,39,0,0,0,155563380,1,9895,0,0,0,0,0,0,0,7239,1800,0
+37,1,4464,0,4334,4334,4334,4334.0,136,144,405,3907,349526,349526,86400,39,0,0,0,155563380,1,9851,0,0,0,0,0,0,0,7239,1800,0
+107,1,1052,0,1022,1022,1022,1022.0,758,77,95,901,0,0,0,0,0,0,0,134218344,6,10091,5,218,9592,220,9680,0,19272,0,0,0
+19,200,794,0,11895,145,38,59.5,1050,321,252,10017,0,0,0,56,0,0,0,0,22,2739,0,0,0,0,0,0,0,0,0,0
+26,1,315,0,312,312,312,312.0,2,1,1,306,0,0,0,0,0,0,0,0,0,3777,0,0,0,0,0,0,0,0,0,0
+2,1,276,0,267,267,267,267.0,6,4,4,124,0,0,0,0,0,0,0,0,1,3342,0,0,0,0,0,0,0,0,0,0
+11,1,260,0,254,254,254,254.0,5,3,3,241,0,0,0,0,0,0,0,0,0,2913,0,0,0,0,0,0,0,0,0,0
+7,1,238,0,227,227,227,227.0,5,4,4,213,0,0,0,114,0,0,0,0,0,2206,0,0,0,0,0,0,0,0,0,0
+1,1,202,0,173,173,173,173.0,28,5,6,152,0,0,0,0,0,0,0,0,0,2506,0,0,0,0,0,0,0,0,0,0
+5,1,177,0,165,165,165,165.0,4,4,4,151,0,0,0,0,0,0,0,0,0,2475,0,0,0,0,0,0,0,0,0,0
+14,1,148,0,143,143,143,143.0,3,1,1,132,0,0,0,0,0,0,0,0,1,3120,0,0,0,0,0,0,0,0,0,0
+4,1,145,0,139,139,139,139.0,22,5,6,121,0,0,0,0,0,0,0,0,0,2334,0,0,0,0,0,0,0,0,0,0
+20,1,140,0,137,137,137,137.0,1,1,1,130,0,0,0,0,0,0,0,0,0,2170,0,0,0,0,0,0,0,0,0,0
+28,1,139,0,136,136,136,136.0,2,1,1,130,0,0,0,0,0,0,0,0,0,3784,0,0,0,0,0,0,0,0,0,0
+18,1,127,0,124,124,124,124.0,2,1,1,116,0,0,0,0,0,0,0,0,0,2501,0,0,0,0,0,0,0,0,0,0
+16,1,122,0,117,117,117,117.0,2,1,1,108,0,0,0,0,0,0,0,0,0,2758,0,0,0,0,0,0,0,0,0,0
+6,1,120,0,113,113,113,113.0,4,3,3,100,0,0,0,0,0,0,0,0,0,2208,0,0,0,0,0,0,0,0,0,0
+10,1,118,0,110,110,110,110.0,6,3,3,98,0,0,0,0,0,0,0,0,0,3565,0,0,0,0,0,0,0,0,0,0
+9,1,110,0,104,104,104,104.0,5,3,3,90,0,0,0,0,0,0,0,0,1,3514,0,0,0,0,0,0,0,0,0,0
+17,1,102,0,97,97,97,97.0,2,2,2,89,0,0,0,0,0,0,0,0,0,3003,0,0,0,0,0,0,0,0,0,0
+8,1,100,0,95,95,95,95.0,4,3,3,82,0,0,0,0,0,0,0,0,0,3142,0,0,0,0,0,0,0,0,0,0
+12,1,98,0,85,85,85,85.0,4,3,3,72,0,0,0,0,0,0,0,0,0,3369,0,0,0,0,0,0,0,0,0,0
+30,1,71,0,67,67,67,67.0,2,1,1,62,0,0,0,0,0,0,0,0,0,3199,0,0,0,0,0,0,0,0,0,0
+22,1,68,0,65,65,65,65.0,2,1,1,59,0,0,0,0,0,0,0,0,0,3436,0,0,0,0,0,0,0,0,0,0
+24,1,68,0,59,59,59,59.0,2,1,1,51,0,0,0,0,0,0,0,0,0,3288,0,0,0,0,0,0,0,0,0,0
+81,1,63,0,54,54,54,54.0,27,1,1,49,0,0,0,0,0,0,0,138414192,0,5343,0,216,9504,215,9460,0,18964,44,1,0
+15,1,62,0,58,58,58,58.0,2,1,1,50,0,0,0,0,0,0,0,0,0,2306,0,0,0,0,0,0,0,0,0,0
diff --git a/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_jobmetricsagg2_expectation.csv b/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_jobmetricsagg2_expectation.csv
index 83eaa2b17..4c53a33c8 100644
--- a/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_jobmetricsagg2_expectation.csv
+++ b/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_jobmetricsagg2_expectation.csv
@@ -1,2 +1,2 @@
-jobId,numTasks,Duration,diskBytesSpilled_sum,duration_sum,duration_max,duration_min,duration_avg,executorCPUTime_sum,executorDeserializeCPUTime_sum,executorDeserializeTime_sum,executorRunTime_sum,input_bytesRead_sum,input_recordsRead_sum,jvmGCTime_sum,memoryBytesSpilled_sum,output_bytesWritten_sum,output_recordsWritten_sum,peakExecutionMemory_max,resultSerializationTime_sum,resultSize_max,sr_fetchWaitTime_sum,sr_localBlocksFetched_sum,sr_localBytesRead_sum,sr_remoteBlocksFetched_sum,sr_remoteBytesRead_sum,sr_remoteBytesReadToDisk_sum,sr_totalBytesRead_sum,sw_bytesWritten_sum,sw_recordsWritten_sum,sw_writeTime_sum
-0,213,2515,0,25761,1624,9,120.9,7151,3134,11178,13522,0,0,424,0,0,0,0,10,8075,0,2600,80279920,0,0,0,80279920,80279920,2600,901
+id,numTasks,duration,diskBytesSpilledSum,durationSum,durationMax,durationMin,durationAvg,executorCPUTimeSum,executorDeserializeCpuTimeSum,executorDeserializeTimeSum,executorRunTimeSum,inputBytesReadSum,inputBytesReadMax,inputRecordsReadSum,jvmGCTimeSum,memoryBytesSpilledSum,outputBytesWrittenSum,outputRecordsWrittenSum,peakExecutionMemoryMax,resultSerializationTimeSum,resultSizeMax,srFetchWaitTimeSum,srLocalBlocksFetchedSum,srcLocalBytesReadSum,srRemoteBlocksFetchSum,srRemoteBytesReadSum,srRemoteBytesReadToDiskSum,srTotalBytesReadSum,swBytesWrittenSum,swRecordsWrittenSum,swWriteTimeSum
+0,213,2515,0,25761,1624,9,120.9,7151,3134,11178,13522,0,0,0,424,0,0,0,0,10,8075,0,2600,80279920,0,0,0,80279920,80279920,2600,901
diff --git a/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_jobmetricsagg_expectation.csv b/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_jobmetricsagg_expectation.csv
index 956c547a8..15772ddfb 100644
--- a/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_jobmetricsagg_expectation.csv
+++ b/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_jobmetricsagg_expectation.csv
@@ -1,2 +1,2 @@
-jobId,numTasks,Duration,diskBytesSpilled_sum,duration_sum,duration_max,duration_min,duration_avg,executorCPUTime_sum,executorDeserializeCPUTime_sum,executorDeserializeTime_sum,executorRunTime_sum,input_bytesRead_sum,input_recordsRead_sum,jvmGCTime_sum,memoryBytesSpilled_sum,output_bytesWritten_sum,output_recordsWritten_sum,peakExecutionMemory_max,resultSerializationTime_sum,resultSize_max,sr_fetchWaitTime_sum,sr_localBlocksFetched_sum,sr_localBytesRead_sum,sr_remoteBlocksFetched_sum,sr_remoteBytesRead_sum,sr_remoteBytesReadToDisk_sum,sr_totalBytesRead_sum,sw_bytesWritten_sum,sw_recordsWritten_sum,sw_writeTime_sum
-0,213,2569,0,26735,1598,10,125.5,6608,3531,12095,13414,0,0,336,0,0,0,0,8,8075,0,2600,80279908,0,0,0,80279908,80279908,2600,1001
+id,numTasks,duration,diskBytesSpilledSum,durationSum,durationMax,durationMin,durationAvg,executorCPUTimeSum,executorDeserializeCpuTimeSum,executorDeserializeTimeSum,executorRunTimeSum,inputBytesReadSum,inputBytesReadMax,inputRecordsReadSum,jvmGCTimeSum,memoryBytesSpilledSum,outputBytesWrittenSum,outputRecordsWrittenSum,peakExecutionMemoryMax,resultSerializationTimeSum,resultSizeMax,srFetchWaitTimeSum,srLocalBlocksFetchedSum,srcLocalBytesReadSum,srRemoteBlocksFetchSum,srRemoteBytesReadSum,srRemoteBytesReadToDiskSum,srTotalBytesReadSum,swBytesWrittenSum,swRecordsWrittenSum,swWriteTimeSum
+0,213,2569,0,26735,1598,10,125.5,6608,3531,12095,13414,0,0,0,336,0,0,0,0,8,8075,0,2600,80279908,0,0,0,80279908,80279908,2600,1001
diff --git a/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_sqlmetricsagg2_expectation.csv b/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_sqlmetricsagg2_expectation.csv
index 0efee3efd..6085ba94c 100644
--- a/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_sqlmetricsagg2_expectation.csv
+++ b/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_sqlmetricsagg2_expectation.csv
@@ -1,2 +1,2 @@
-appID,sqlID,description,numTasks,Duration,executorCPURatio,diskBytesSpilled_sum,duration_sum,duration_max,duration_min,duration_avg,executorCPUTime_sum,executorDeserializeCPUTime_sum,executorDeserializeTime_sum,executorRunTime_sum,input_bytesRead_sum,input_recordsRead_sum,jvmGCTime_sum,memoryBytesSpilled_sum,output_bytesWritten_sum,output_recordsWritten_sum,peakExecutionMemory_max,resultSerializationTime_sum,resultSize_max,sr_fetchWaitTime_sum,sr_localBlocksFetched_sum,sr_localBytesRead_sum,sr_remoteBlocksFetched_sum,sr_remoteBytesRead_sum,sr_remoteBytesReadToDisk_sum,sr_totalBytesRead_sum,sw_bytesWritten_sum,sw_recordsWritten_sum,sw_writeTime_sum
-"local-1622821994212",0,"count at <console>:28",213,3041,52.88,0,25761,1624,9,120.9,7151,3134,11178,13522,0,0,424,0,0,0,0,10,8075,0,2600,80279920,0,0,0,80279920,80279920,2600,901
+appId,sqlId,description,numTasks,duration,executorCpuRatio,diskBytesSpilledSum,durationSum,durationMax,durationMin,durationAvg,executorCPUTimeSum,executorDeserializeCpuTimeSum,executorDeserializeTimeSum,executorRunTimeSum,inputBytesReadSum,inputBytesReadMax,inputRecordsReadSum,jvmGCTimeSum,memoryBytesSpilledSum,outputBytesWrittenSum,outputRecordsWrittenSum,peakExecutionMemoryMax,resultSerializationTimeSum,resultSizeMax,srFetchWaitTimeSum,srLocalBlocksFetchedSum,srcLocalBytesReadSum,srRemoteBlocksFetchSum,srRemoteBytesReadSum,srRemoteBytesReadToDiskSum,srTotalBytesReadSum,swBytesWrittenSum,swRecordsWrittenSum,swWriteTimeSum
+local-1622821994212,0,count at <console>:28,213,3041,52.88,0,25761,1624,9,120.9,7151,3134,11178,13522,0,0,0,424,0,0,0,0,10,8075,0,2600,80279920,0,0,0,80279920,80279920,2600,901
diff --git a/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_sqlmetricsagg_expectation.csv b/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_sqlmetricsagg_expectation.csv
index 8b52ac3cb..066b4c184 100644
--- a/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_sqlmetricsagg_expectation.csv
+++ b/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_sqlmetricsagg_expectation.csv
@@ -1,2 +1,2 @@
-appID,sqlID,description,numTasks,Duration,executorCPURatio,diskBytesSpilled_sum,duration_sum,duration_max,duration_min,duration_avg,executorCPUTime_sum,executorDeserializeCPUTime_sum,executorDeserializeTime_sum,executorRunTime_sum,input_bytesRead_sum,input_recordsRead_sum,jvmGCTime_sum,memoryBytesSpilled_sum,output_bytesWritten_sum,output_recordsWritten_sum,peakExecutionMemory_max,resultSerializationTime_sum,resultSize_max,sr_fetchWaitTime_sum,sr_localBlocksFetched_sum,sr_localBytesRead_sum,sr_remoteBlocksFetched_sum,sr_remoteBytesRead_sum,sr_remoteBytesReadToDisk_sum,sr_totalBytesRead_sum,sw_bytesWritten_sum,sw_recordsWritten_sum,sw_writeTime_sum
-"local-1622814619968",0,"count at <console>:28",213,3087,49.26,0,26735,1598,10,125.5,6608,3531,12095,13414,0,0,336,0,0,0,0,8,8075,0,2600,80279908,0,0,0,80279908,80279908,2600,1001
+appId,sqlId,description,numTasks,duration,executorCpuRatio,diskBytesSpilledSum,durationSum,durationMax,durationMin,durationAvg,executorCPUTimeSum,executorDeserializeCpuTimeSum,executorDeserializeTimeSum,executorRunTimeSum,inputBytesReadSum,inputBytesReadMax,inputRecordsReadSum,jvmGCTimeSum,memoryBytesSpilledSum,outputBytesWrittenSum,outputRecordsWrittenSum,peakExecutionMemoryMax,resultSerializationTimeSum,resultSizeMax,srFetchWaitTimeSum,srLocalBlocksFetchedSum,srcLocalBytesReadSum,srRemoteBlocksFetchSum,srRemoteBytesReadSum,srRemoteBytesReadToDiskSum,srTotalBytesReadSum,swBytesWrittenSum,swRecordsWrittenSum,swWriteTimeSum
+local-1622814619968,0,count at <console>:28,213,3087,49.26,0,26735,1598,10,125.5,6608,3531,12095,13414,0,0,0,336,0,0,0,0,8,8075,0,2600,80279908,0,0,0,80279908,80279908,2600,1001
diff --git a/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_stagemetricsagg2_expectation.csv b/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_stagemetricsagg2_expectation.csv
index 1dc0462ac..36e2b92cb 100644
--- a/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_stagemetricsagg2_expectation.csv
+++ b/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_stagemetricsagg2_expectation.csv
@@ -1,5 +1,5 @@
-stageId,numTasks,Duration,diskBytesSpilled_sum,duration_sum,duration_max,duration_min,duration_avg,executorCPUTime_sum,executorDeserializeCPUTime_sum,executorDeserializeTime_sum,executorRunTime_sum,input_bytesRead_sum,input_recordsRead_sum,jvmGCTime_sum,memoryBytesSpilled_sum,output_bytesWritten_sum,output_recordsWritten_sum,peakExecutionMemory_max,resultSerializationTime_sum,resultSize_max,sr_fetchWaitTime_sum,sr_localBlocksFetched_sum,sr_localBytesRead_sum,sr_remoteBlocksFetched_sum,sr_remoteBytesRead_sum,sr_remoteBytesReadToDisk_sum,sr_totalBytesRead_sum,sw_bytesWritten_sum,sw_recordsWritten_sum,sw_writeTime_sum
-0,6,1761,0,9455,1624,1540,1575.8,2917,1287,5056,4248,0,0,228,0,0,0,0,3,2951,0,0,0,0,0,0,0,40132263,1200,376
-1,6,1666,0,9274,1621,1528,1545.7,2570,1007,5016,4099,0,0,196,0,0,0,0,4,2951,0,0,0,0,0,0,0,40132257,1200,475
-2,200,592,0,6937,221,9,34.7,1619,802,1065,5125,0,0,0,0,0,0,0,3,7402,0,2400,80264520,0,0,0,80264520,15400,200,50
-3,1,101,0,95,95,95,95.0,45,38,41,50,0,0,0,0,0,0,0,0,8075,0,200,15400,0,0,0,15400,0,0,0
+id,numTasks,duration,diskBytesSpilledSum,durationSum,durationMax,durationMin,durationAvg,executorCPUTimeSum,executorDeserializeCpuTimeSum,executorDeserializeTimeSum,executorRunTimeSum,inputBytesReadSum,inputBytesReadMax,inputRecordsReadSum,jvmGCTimeSum,memoryBytesSpilledSum,outputBytesWrittenSum,outputRecordsWrittenSum,peakExecutionMemoryMax,resultSerializationTimeSum,resultSizeMax,srFetchWaitTimeSum,srLocalBlocksFetchedSum,srcLocalBytesReadSum,srRemoteBlocksFetchSum,srRemoteBytesReadSum,srRemoteBytesReadToDiskSum,srTotalBytesReadSum,swBytesWrittenSum,swRecordsWrittenSum,swWriteTimeSum
+0,6,1761,0,9455,1624,1540,1575.8,2917,1287,5056,4248,0,0,0,228,0,0,0,0,3,2951,0,0,0,0,0,0,0,40132263,1200,376
+1,6,1666,0,9274,1621,1528,1545.7,2570,1007,5016,4099,0,0,0,196,0,0,0,0,4,2951,0,0,0,0,0,0,0,40132257,1200,475
+2,200,592,0,6937,221,9,34.7,1619,802,1065,5125,0,0,0,0,0,0,0,0,3,7402,0,2400,80264520,0,0,0,80264520,15400,200,50
+3,1,101,0,95,95,95,95.0,45,38,41,50,0,0,0,0,0,0,0,0,0,8075,0,200,15400,0,0,0,15400,0,0,0
diff --git a/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_stagemetricsagg_expectation.csv b/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_stagemetricsagg_expectation.csv
index 1a937cf9e..93a4ed36f 100644
--- a/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_stagemetricsagg_expectation.csv
+++ b/core/src/test/resources/ProfilingExpectations/rapids_join_eventlog_stagemetricsagg_expectation.csv
@@ -1,5 +1,5 @@
-stageId,numTasks,Duration,diskBytesSpilled_sum,duration_sum,duration_max,duration_min,duration_avg,executorCPUTime_sum,executorDeserializeCPUTime_sum,executorDeserializeTime_sum,executorRunTime_sum,input_bytesRead_sum,input_recordsRead_sum,jvmGCTime_sum,memoryBytesSpilled_sum,output_bytesWritten_sum,output_recordsWritten_sum,peakExecutionMemory_max,resultSerializationTime_sum,resultSize_max,sr_fetchWaitTime_sum,sr_localBlocksFetched_sum,sr_localBytesRead_sum,sr_remoteBlocksFetched_sum,sr_remoteBytesRead_sum,sr_remoteBytesReadToDisk_sum,sr_totalBytesRead_sum,sw_bytesWritten_sum,sw_recordsWritten_sum,sw_writeTime_sum
-0,6,1743,0,9518,1598,1580,1586.3,2512,1393,5309,4043,0,0,168,0,0,0,0,3,2951,0,0,0,0,0,0,0,40132250,1200,400
-1,6,1631,0,9434,1582,1568,1572.3,2406,1067,5273,3998,0,0,168,0,0,0,0,5,2951,0,0,0,0,0,0,0,40132258,1200,508
-2,200,688,0,7705,237,10,38.5,1660,1034,1474,5337,0,0,0,0,0,0,0,0,7359,0,2400,80264508,0,0,0,80264508,15400,200,93
-3,1,83,0,78,78,78,78.0,30,37,39,36,0,0,0,0,0,0,0,0,8075,0,200,15400,0,0,0,15400,0,0,0
+id,numTasks,duration,diskBytesSpilledSum,durationSum,durationMax,durationMin,durationAvg,executorCPUTimeSum,executorDeserializeCpuTimeSum,executorDeserializeTimeSum,executorRunTimeSum,inputBytesReadSum,inputBytesReadMax,inputRecordsReadSum,jvmGCTimeSum,memoryBytesSpilledSum,outputBytesWrittenSum,outputRecordsWrittenSum,peakExecutionMemoryMax,resultSerializationTimeSum,resultSizeMax,srFetchWaitTimeSum,srLocalBlocksFetchedSum,srcLocalBytesReadSum,srRemoteBlocksFetchSum,srRemoteBytesReadSum,srRemoteBytesReadToDiskSum,srTotalBytesReadSum,swBytesWrittenSum,swRecordsWrittenSum,swWriteTimeSum
+0,6,1743,0,9518,1598,1580,1586.3,2512,1393,5309,4043,0,0,0,168,0,0,0,0,3,2951,0,0,0,0,0,0,0,40132250,1200,400
+1,6,1631,0,9434,1582,1568,1572.3,2406,1067,5273,3998,0,0,0,168,0,0,0,0,5,2951,0,0,0,0,0,0,0,40132258,1200,508
+2,200,688,0,7705,237,10,38.5,1660,1034,1474,5337,0,0,0,0,0,0,0,0,0,7359,0,2400,80264508,0,0,0,80264508,15400,200,93
+3,1,83,0,78,78,78,78.0,30,37,39,36,0,0,0,0,0,0,0,0,0,8075,0,200,15400,0,0,0,15400,0,0,0

From 42c5a19455c1e5722f9debdd5854e0b3bc6d737b Mon Sep 17 00:00:00 2001
From: Partho Sarthi <psarthi@nvidia.com>
Date: Wed, 22 Apr 2026 16:42:32 -0700
Subject: [PATCH 14/16] Bump copyright year to 2026 on modified Scala files

Files touched by the tuning-signals relocation change had outdated
copyright headers that the CI header check flagged as expired.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Signed-off-by: Partho Sarthi <psarthi@nvidia.com>
---
 .../nvidia/spark/rapids/tool/analysis/AggRawMetricsResult.scala | 2 +-
 .../spark/rapids/tool/analysis/AppSparkMetricsAggTrait.scala    | 2 +-
 .../spark/rapids/tool/analysis/AppSparkMetricsAnalyzer.scala    | 2 +-
 .../spark/rapids/tool/analysis/util/TaskMetricsAccumRec.scala   | 2 +-
 .../spark/rapids/tool/tuning/QualAppSummaryInfoProvider.scala   | 2 +-
 .../com/nvidia/spark/rapids/tool/views/RawMetricProfView.scala  | 2 +-
 6 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AggRawMetricsResult.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AggRawMetricsResult.scala
index 8a5f2d7ac..c460f3d5c 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AggRawMetricsResult.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AggRawMetricsResult.scala
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2024, NVIDIA CORPORATION.
+ * Copyright (c) 2024-2026, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AppSparkMetricsAggTrait.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AppSparkMetricsAggTrait.scala
index a8f39a168..67b899962 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AppSparkMetricsAggTrait.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AppSparkMetricsAggTrait.scala
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2024-2025, NVIDIA CORPORATION.
+ * Copyright (c) 2024-2026, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AppSparkMetricsAnalyzer.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AppSparkMetricsAnalyzer.scala
index 1a4c3f12e..c1b13463a 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AppSparkMetricsAnalyzer.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/AppSparkMetricsAnalyzer.scala
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2024-2025, NVIDIA CORPORATION.
+ * Copyright (c) 2024-2026, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/util/TaskMetricsAccumRec.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/util/TaskMetricsAccumRec.scala
index 79174e3b0..ad92f98a4 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/util/TaskMetricsAccumRec.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/analysis/util/TaskMetricsAccumRec.scala
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2024, NVIDIA CORPORATION.
+ * Copyright (c) 2024-2026, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/tuning/QualAppSummaryInfoProvider.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/tuning/QualAppSummaryInfoProvider.scala
index d6d478167..f69ffd298 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/tuning/QualAppSummaryInfoProvider.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/tuning/QualAppSummaryInfoProvider.scala
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2024-2025, NVIDIA CORPORATION.
+ * Copyright (c) 2024-2026, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/RawMetricProfView.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/RawMetricProfView.scala
index a55caaf9b..3ba76db7b 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/RawMetricProfView.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/RawMetricProfView.scala
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2024-2025, NVIDIA CORPORATION.
+ * Copyright (c) 2024-2026, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.

From 6aea65e173451f893c1a8819e8a1b22b8b84a1dd Mon Sep 17 00:00:00 2001
From: Partho Sarthi <psarthi@nvidia.com>
Date: Fri, 24 Apr 2026 14:35:10 -0700
Subject: [PATCH 15/16] Rename tuning_signals.csv ->
 app_level_recommendation_signals.csv (wide)

Replaces the 2-row vertical tuning_signals.csv with a single-row wide
app_level_recommendation_signals.csv. Addresses review feedback from
@hirakendu:
- Generic naming ("recommendation signals") so future qualx / tuneml
  features can co-locate their signals here.
- Explicit granularity in the filename so per-SQL / per-stage siblings
  can be added later without ambiguity.

Columns:
  appId
  numScanStagesWithGpuOom               (profiling only, 0 for qual)
  numGpuShuffleStagesWithContainerOom   (profiling only, 0 for qual)

Comma-separated stage-ID lists are replaced by simple counts, which are
aggregation-friendly and atomic. If downstream consumers need per-stage
detail, the stage IDs are still available via failed_tasks.csv /
failed_stages.csv cross-reference (the same data this file summarises).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Signed-off-by: Partho Sarthi <psarthi@nvidia.com>
---
 .../configs/reports/coreRawMetricsReport.yaml | 31 +++++++++-------
 .../profiling/ApplicationSummaryInfo.scala    |  2 +-
 .../profiling/ProfileClassWarehouse.scala     | 36 ++++++++++---------
 .../rapids/tool/profiling/Profiler.scala      |  9 ++---
 .../rapids/tool/views/OutHeaderRegistry.scala |  4 +--
 .../tool/views/QualRawReportGenerator.scala   | 10 +++---
 .../spark/rapids/tool/views/package.scala     |  2 +-
 7 files changed, 52 insertions(+), 42 deletions(-)

diff --git a/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml b/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
index 4fadf8a3f..404bc7ede 100644
--- a/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
+++ b/core/src/main/resources/configs/reports/coreRawMetricsReport.yaml
@@ -16,6 +16,7 @@
 # Note that all the output files in this list should be considered internal-use only.
 
 # <application-id>/
+#    ├── app_level_recommendation_signals.csv
 #    ├── application_information.csv
 #    ├── application_log_path_mapping.csv
 #    ├── cluster_information.json
@@ -40,7 +41,6 @@
 #    ├── stage_level_aggregated_task_metrics.csv
 #    ├── stage_level_all_metrics.csv
 #    ├── system_properties.csv
-#    ├── tuning_signals.csv
 #    ├── unsupported_sql_plan.csv
 #    ├── wholestagecodegen_mapping.csv
 #    ├── write_operations.csv
@@ -1439,20 +1439,27 @@ reportDefinitions:
         fileName: profile.log
         fileFormat: TXT
         scope: per-app
-      # TuningSignalProfileResult
-      - label: coreRawTuningSignalsCSV
+      # AppLevelRecommendationSignalsProfileResult
+      - label: coreRawAppLevelRecommendationSignalsCSV
         description: >-
-          Per-app GPU-OOM diagnostic signals used as AutoTuner inputs. Vertical key-value
-          layout; each emitted metric is one row.
-        fileName: tuning_signals.csv
+          Per-app derived signals that feed recommendation engines (AutoTuner, qualx,
+          etc.). Single row per app; wide layout, one column per signal.
+          GPU-only signals are 0 for qualification (CPU event logs).
+        fileName: app_level_recommendation_signals.csv
         scope: per-app
         columns:
-          - name: name
+          - name: appId
             dataType: String
             description: >-
-              Name of the tuning signal
-          - name: value
-            dataType: String
+              Application ID.
+          - name: numScanStagesWithGpuOom
+            dataType: Int
+            description: >-
+              Number of scan stages where failed tasks had GPU OOM errors
+              (GpuRetryOOM / GpuSplitAndRetryOOM / jni.GpuOOM). Profiling only.
+          - name: numGpuShuffleStagesWithContainerOom
+            dataType: Int
             description: >-
-              Signal value. Numeric signals are stored as their integer string form;
-              stage-ID signals are comma-separated lists (empty string when none).
+              Number of GPU shuffle stages (GpuShuffleExchangeExec) where YARN killed
+              the container due to container-level OOM (ExecutorLostFailure + exit 137).
+              Profiling only.
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
index c8db4cced..9e1d2ebc8 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ApplicationSummaryInfo.scala
@@ -54,7 +54,7 @@ case class ApplicationSummaryInfo(
     sparkRapidsBuildInfo: Seq[SparkRapidsBuildInfoEvent],
     writeOpsInfo: Seq[WriteOpProfileResult],
     sqlPlanInfo: Seq[SQLPlanInfoProfileResult],
-    tuningSignals: Seq[TuningSignalProfileResult] = Seq.empty)
+    appLevelRecommendationSignals: Seq[AppLevelRecommendationSignalsProfileResult] = Seq.empty)
 
 trait AppInfoPropertyGetter {
   // returns all the properties (i.e., spark)
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
index bd4f74677..7aa3cca55 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
@@ -516,32 +516,34 @@ case class AppInfoProfileResults(
   }
 }
 
-case class TuningSignalProfileResult(
-    name: String,
-    value: String) extends ProfileResult {
+case class AppLevelRecommendationSignalsProfileResult(
+    appId: String,
+    numScanStagesWithGpuOom: Int,
+    numGpuShuffleStagesWithContainerOom: Int) extends ProfileResult {
   override def outputHeaders: Array[String] = {
-    OutHeaderRegistry.outputHeaders("TuningSignalProfileResult")
+    OutHeaderRegistry.outputHeaders("AppLevelRecommendationSignalsProfileResult")
   }
 
-  override def convertToSeq(): Array[String] = Array(name, value)
+  override def convertToSeq(): Array[String] = Array(
+    appId,
+    numScanStagesWithGpuOom.toString,
+    numGpuShuffleStagesWithContainerOom.toString)
 
   override def convertToCSVSeq(): Array[String] = Array(
-    StringUtils.reformatCSVString(name),
-    StringUtils.reformatCSVString(value))
+    StringUtils.reformatCSVString(appId),
+    numScanStagesWithGpuOom.toString,
+    numGpuShuffleStagesWithContainerOom.toString)
 }
 
-object TuningSignalProfileResult {
-  private def stageIdsToStr(stageIds: Set[Long]): String = {
-    if (stageIds.isEmpty) "" else stageIds.toSeq.sorted.mkString(",")
-  }
-
+object AppLevelRecommendationSignalsProfileResult {
   def build(
+      appId: String,
       scanStagesWithGpuOom: Set[Long],
-      gpuShuffleStagesWithContainerOom: Set[Long]): Seq[TuningSignalProfileResult] = Seq(
-    TuningSignalProfileResult("scanStagesWithGpuOom",
-      stageIdsToStr(scanStagesWithGpuOom)),
-    TuningSignalProfileResult("gpuShuffleStagesWithContainerOom",
-      stageIdsToStr(gpuShuffleStagesWithContainerOom)))
+      gpuShuffleStagesWithContainerOom: Set[Long]): Seq[AppLevelRecommendationSignalsProfileResult] = Seq(
+    AppLevelRecommendationSignalsProfileResult(
+      appId,
+      scanStagesWithGpuOom.size,
+      gpuShuffleStagesWithContainerOom.size))
 }
 
 case class AppLogPathProfileResults(
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
index 915232f45..2b689397b 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/Profiler.scala
@@ -316,8 +316,8 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
 
     val appInfo = collect.getAppInfo
     val appId = appInfo.headOption.flatMap(_.appId).getOrElse("")
-    val tuningSignals = TuningSignalProfileResult.build(
-      scanOomStages, gpuShuffleContainerOomStages)
+    val appLevelRecommendationSignals = AppLevelRecommendationSignalsProfileResult.build(
+      appId, scanOomStages, gpuShuffleContainerOomStages)
 
     logDebug(s"Time to collect Profiling Info [$appId]: ${endTime - startTime}.")
     val appInfoSummary = ApplicationSummaryInfo(
@@ -350,7 +350,7 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
       sparkRapidsBuildInfo = collect.getSparkRapidsInfo,
       writeOpsInfo = collect.getWriteOperationInfo,
       sqlPlanInfo = collect.getSQLPlanInfoTruncated,
-      tuningSignals = tuningSignals)
+      appLevelRecommendationSignals = appLevelRecommendationSignals)
     (appInfoSummary,
      DiagnosticSummaryInfo(analysis.stageDiagnostics, collect.getIODiagnosticMetrics))
   }
@@ -423,7 +423,8 @@ class Profiler(hadoopConf: Configuration, appArgs: ProfileArgs, enablePB: Boolea
     // writeOps are generated in only CSV format
     profileOutputWriter.writeCSVTable(ProfWriteOpsView.getLabel, app.writeOpsInfo)
     profileOutputWriter.writeCSVTable(TASK_SHUFFLE_SKEW, app.skewInfo)
-    profileOutputWriter.writeCSVTable(TUNING_SIGNALS, app.tuningSignals)
+    profileOutputWriter.writeCSVTable(APP_LEVEL_RECOMMENDATION_SIGNALS,
+      app.appLevelRecommendationSignals)
     profileOutputWriter.writeText("\n### C. Health Check###\n")
     profileOutputWriter.writeCSVTable(ProfFailedTaskView.getLabel, app.failedTasks)
     profileOutputWriter.writeTable(ProfFailedStageView.getLabel, app.failedStages)
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
index f09b4eec6..ac61a5d3b 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/OutHeaderRegistry.scala
@@ -70,8 +70,8 @@ object OutHeaderRegistry {
     "AppInfoProfileResults" ->
       Array("appName", "appId", "attemptId", "sparkUser", "startTime", "endTime", "duration",
         "durationStr", "sparkRuntime", "sparkVersion", "pluginEnabled", "totalCoreSeconds"),
-    "TuningSignalProfileResult" ->
-      Array("name", "value"),
+    "AppLevelRecommendationSignalsProfileResult" ->
+      Array("appId", "numScanStagesWithGpuOom", "numGpuShuffleStagesWithContainerOom"),
     "AppLogPathProfileResults" ->
       Array("appName", "appId", "eventLogPath"),
     "FailedTaskProfileResults" ->
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
index e7002d198..c26a9b789 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/QualRawReportGenerator.scala
@@ -17,7 +17,7 @@
 package com.nvidia.spark.rapids.tool.views
 
 import com.nvidia.spark.rapids.tool.analysis.{AggRawMetricsResult, AppSQLPlanAnalyzer, QualSparkMetricsAggregator}
-import com.nvidia.spark.rapids.tool.profiling.{DataSourceProfileResult, ProfileOutputWriter, ProfileResult, SQLAccumProfileResults, TuningSignalProfileResult}
+import com.nvidia.spark.rapids.tool.profiling.{AppLevelRecommendationSignalsProfileResult, DataSourceProfileResult, ProfileOutputWriter, ProfileResult, SQLAccumProfileResults}
 
 import org.apache.spark.internal.Logging
 import org.apache.spark.sql.rapids.tool.qualification.QualificationAppInfo
@@ -101,10 +101,10 @@ object QualRawReportGenerator extends Logging {
       constructLabelsMaps(aggRawMetrics).foreach { case (label, metrics) =>
           pWriter.writeCSVTable(label, metrics)
       }
-      // GPU-only signals default to empty for qualification (CPU event logs)
-      val tuningSignals = TuningSignalProfileResult.build(
-        Set.empty[Long], Set.empty[Long])
-      pWriter.writeCSVTable(TUNING_SIGNALS, tuningSignals)
+      // GPU-only signals default to 0 for qualification (CPU event logs)
+      val appLevelRecommendationSignals = AppLevelRecommendationSignalsProfileResult.build(
+        app.appId, Set.empty[Long], Set.empty[Long])
+      pWriter.writeCSVTable(APP_LEVEL_RECOMMENDATION_SIGNALS, appLevelRecommendationSignals)
       pWriter.writeText("\n### C. Health Check###\n")
       pWriter.writeCSVTable(QualFailedTaskView.getLabel, QualFailedTaskView.getRawView(Seq(app)))
       pWriter.writeTable(
diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala
index 972bc6fd5..7251a4090 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/views/package.scala
@@ -29,7 +29,7 @@ package object views {
   val SQL_DUR_LABEL = "SQL Duration and Executor CPU Time Percent"
   val SQL_MAX_INPUT_SIZE = "SQL Max Task Input Size"
   val STAGE_DIAGNOSTICS_LABEL = "Stage Level Diagnostic Metrics"
-  val TUNING_SIGNALS = "Tuning Signals"
+  val APP_LEVEL_RECOMMENDATION_SIGNALS = "App Level Recommendation Signals"
   val CLUSTER_INFORMATION_LABEL = "Cluster Information"
 
   val AGG_DESCRIPTION = Map(

From 073601c066664a2afa11a94191e623d178c74eed Mon Sep 17 00:00:00 2001
From: Partho Sarthi <psarthi@nvidia.com>
Date: Fri, 24 Apr 2026 15:03:58 -0700
Subject: [PATCH 16/16] Fix scalastyle line-length violation in
 AppLevelRecommendationSignalsProfileResult.build

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Signed-off-by: Partho Sarthi <psarthi@nvidia.com>
---
 .../spark/rapids/tool/profiling/ProfileClassWarehouse.scala | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
index 7aa3cca55..b467db8c6 100644
--- a/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
+++ b/core/src/main/scala/com/nvidia/spark/rapids/tool/profiling/ProfileClassWarehouse.scala
@@ -539,11 +539,13 @@ object AppLevelRecommendationSignalsProfileResult {
   def build(
       appId: String,
       scanStagesWithGpuOom: Set[Long],
-      gpuShuffleStagesWithContainerOom: Set[Long]): Seq[AppLevelRecommendationSignalsProfileResult] = Seq(
-    AppLevelRecommendationSignalsProfileResult(
+      gpuShuffleStagesWithContainerOom: Set[Long])
+      : Seq[AppLevelRecommendationSignalsProfileResult] = {
+    Seq(AppLevelRecommendationSignalsProfileResult(
       appId,
       scanStagesWithGpuOom.size,
       gpuShuffleStagesWithContainerOom.size))
+  }
 }
 
 case class AppLogPathProfileResults(