[SPARK-32037][CORE] Rename blacklisting feature

core/src/main/java/org/apache/spark/SparkFirehoseListener.java

@@ -124,34 +124,67 @@ public final void onExecutorBlacklisted(SparkListenerExecutorBlacklisted executo
     onEvent(executorBlacklisted);
   }
 
+  @Override
+  public final void onExecutorExcluded(SparkListenerExecutorExcluded executorExcluded) {
+    onEvent(executorExcluded);
+  }
+
   @Override
   public void onExecutorBlacklistedForStage(
       SparkListenerExecutorBlacklistedForStage executorBlacklistedForStage) {
     onEvent(executorBlacklistedForStage);
   }
 
+  @Override
+  public void onExecutorExcludedForStage(
+      SparkListenerExecutorExcludedForStage executorExcludedForStage) {
+    onEvent(executorExcludedForStage);
+  }
+
   @Override
   public void onNodeBlacklistedForStage(
       SparkListenerNodeBlacklistedForStage nodeBlacklistedForStage) {
     onEvent(nodeBlacklistedForStage);
   }
 
+  @Override
+  public void onNodeExcludedForStage(
+      SparkListenerNodeExcludedForStage nodeExcludedForStage) {
+    onEvent(nodeExcludedForStage);
+  }
+
   @Override
   public final void onExecutorUnblacklisted(
       SparkListenerExecutorUnblacklisted executorUnblacklisted) {
     onEvent(executorUnblacklisted);
   }
 
+  @Override
+  public final void onExecutorUnexcluded(
+      SparkListenerExecutorUnexcluded executorUnexcluded) {
+    onEvent(executorUnexcluded);
+  }
+
   @Override
   public final void onNodeBlacklisted(SparkListenerNodeBlacklisted nodeBlacklisted) {
     onEvent(nodeBlacklisted);
   }
 
+  @Override
+  public final void onNodeExcluded(SparkListenerNodeExcluded nodeExcluded) {
+    onEvent(nodeExcluded);
+  }
+
   @Override
   public final void onNodeUnblacklisted(SparkListenerNodeUnblacklisted nodeUnblacklisted) {
     onEvent(nodeUnblacklisted);
   }
 
+  @Override
+  public final void onNodeUnexcluded(SparkListenerNodeUnexcluded nodeUnexcluded) {
+    onEvent(nodeUnexcluded);
+  }
+
   @Override
   public void onBlockUpdated(SparkListenerBlockUpdated blockUpdated) {
     onEvent(blockUpdated);

core/src/main/resources/org/apache/spark/ui/static/executorspage-template.html

@@ -56,8 +56,8 @@ <h4 class="title-table">Summary</h4>
         </th>
         <th>
           <span data-toggle="tooltip" data-placement="top"
-                title="Number of executors blacklisted by the scheduler due to task failures.">
-            Blacklisted</span>
+                title="Number of executors excluded by the scheduler due to task failures.">
+            Excluded</span>
         </th>
           </tr>
         </thead>

core/src/main/resources/org/apache/spark/ui/static/executorspage.js

@@ -26,15 +26,15 @@ function getThreadDumpEnabled() {
 }
 
 function formatStatus(status, type, row) {
-    if (row.isBlacklisted) {
-        return "Blacklisted";
+    if (row.isExcluded) {
+        return "Excluded";
     }
 
     if (status) {
-        if (row.blacklistedInStages.length == 0) {
+        if (row.excludedInStages.length == 0) {
             return "Active"
         }
-        return "Active (Blacklisted in Stages: [" + row.blacklistedInStages.join(", ") + "])";
+        return "Active (Excluded in Stages: [" + row.excludedInStages.join(", ") + "])";
     }
     return "Dead"
 }
@@ -168,7 +168,7 @@ $(document).ready(function () {
             var allTotalInputBytes = 0;
             var allTotalShuffleRead = 0;
             var allTotalShuffleWrite = 0;
-            var allTotalBlacklisted = 0;
+            var allTotalExcluded = 0;
 
             var activeExecCnt = 0;
             var activeRDDBlocks = 0;
@@ -190,7 +190,7 @@ $(document).ready(function () {
             var activeTotalInputBytes = 0;
             var activeTotalShuffleRead = 0;
             var activeTotalShuffleWrite = 0;
-            var activeTotalBlacklisted = 0;
+            var activeTotalExcluded = 0;
 
             var deadExecCnt = 0;
             var deadRDDBlocks = 0;
@@ -212,7 +212,7 @@ $(document).ready(function () {
             var deadTotalInputBytes = 0;
             var deadTotalShuffleRead = 0;
             var deadTotalShuffleWrite = 0;
-            var deadTotalBlacklisted = 0;
+            var deadTotalExcluded = 0;
 
             response.forEach(function (exec) {
                 var memoryMetrics = {
@@ -246,7 +246,7 @@ $(document).ready(function () {
                 allTotalInputBytes += exec.totalInputBytes;
                 allTotalShuffleRead += exec.totalShuffleRead;
                 allTotalShuffleWrite += exec.totalShuffleWrite;
-                allTotalBlacklisted += exec.isBlacklisted ? 1 : 0;
+                allTotalExcluded += exec.isExcluded ? 1 : 0;
                 if (exec.isActive) {
                     activeExecCnt += 1;
                     activeRDDBlocks += exec.rddBlocks;
@@ -268,7 +268,7 @@ $(document).ready(function () {
                     activeTotalInputBytes += exec.totalInputBytes;
                     activeTotalShuffleRead += exec.totalShuffleRead;
                     activeTotalShuffleWrite += exec.totalShuffleWrite;
-                    activeTotalBlacklisted += exec.isBlacklisted ? 1 : 0;
+                    activeTotalExcluded += exec.isExcluded ? 1 : 0;
                 } else {
                     deadExecCnt += 1;
                     deadRDDBlocks += exec.rddBlocks;
@@ -290,7 +290,7 @@ $(document).ready(function () {
                     deadTotalInputBytes += exec.totalInputBytes;
                     deadTotalShuffleRead += exec.totalShuffleRead;
                     deadTotalShuffleWrite += exec.totalShuffleWrite;
-                    deadTotalBlacklisted += exec.isBlacklisted ? 1 : 0;
+                    deadTotalExcluded += exec.isExcluded ? 1 : 0; // todo - TEST BACKWARDS compatibility history?
                 }
             });
 
@@ -315,7 +315,7 @@ $(document).ready(function () {
                 "allTotalInputBytes": allTotalInputBytes,
                 "allTotalShuffleRead": allTotalShuffleRead,
                 "allTotalShuffleWrite": allTotalShuffleWrite,
-                "allTotalBlacklisted": allTotalBlacklisted
+                "allTotalExcluded": allTotalExcluded
             };
             var activeSummary = {
                 "execCnt": ( "Active(" + activeExecCnt + ")"),
@@ -338,7 +338,7 @@ $(document).ready(function () {
                 "allTotalInputBytes": activeTotalInputBytes,
                 "allTotalShuffleRead": activeTotalShuffleRead,
                 "allTotalShuffleWrite": activeTotalShuffleWrite,
-                "allTotalBlacklisted": activeTotalBlacklisted
+                "allTotalExcluded": activeTotalExcluded
             };
             var deadSummary = {
                 "execCnt": ( "Dead(" + deadExecCnt + ")" ),
@@ -361,7 +361,7 @@ $(document).ready(function () {
                 "allTotalInputBytes": deadTotalInputBytes,
                 "allTotalShuffleRead": deadTotalShuffleRead,
                 "allTotalShuffleWrite": deadTotalShuffleWrite,
-                "allTotalBlacklisted": deadTotalBlacklisted
+                "allTotalExcluded": deadTotalExcluded
             };
 
             var data = {executors: response, "execSummary": [activeSummary, deadSummary, totalSummary]};
@@ -547,7 +547,7 @@ $(document).ready(function () {
                         {data: 'allTotalInputBytes', render: formatBytes},
                         {data: 'allTotalShuffleRead', render: formatBytes},
                         {data: 'allTotalShuffleWrite', render: formatBytes},
-                        {data: 'allTotalBlacklisted'}
+                        {data: 'allTotalExcluded'}
                     ],
                     "paging": false,
                     "searching": false,

core/src/main/resources/org/apache/spark/ui/static/stagepage.js

@@ -433,7 +433,7 @@ $(document).ready(function () {
                         {data : "failedTasks"},
                         {data : "killedTasks"},
                         {data : "succeededTasks"},
-                        {data : "isBlacklistedForStage"},
+                        {data : "isExcludedForStage"},
                         {
                             data : function (row, type) {
                                 return row.inputRecords != 0 ? formatBytes(row.inputBytes, type) + " / " + row.inputRecords : "";

core/src/main/resources/org/apache/spark/ui/static/stagespage-template.html

@@ -50,8 +50,8 @@ <h4 class="title-table">Aggregated Metrics by Executor</h4>
                 <th>Succeeded Tasks</th>
                 <th>
           <span data-toggle="tooltip" data-placement="top"
-                title="Shows if this executor has been blacklisted by the scheduler due to task failures.">
-            Blacklisted</span>
+                title="Shows if this executor has been excluded by the scheduler due to task failures.">
+            Excluded</span>
                 </th>
                 <th><span id="executor-summary-input">Input Size / Records</span></th>
                 <th><span id="executor-summary-output">Output Size / Records</span></th>

core/src/main/scala/org/apache/spark/ExecutorAllocationManager.scala

@@ -312,8 +312,8 @@ private[spark] class ExecutorAllocationManager(
 
     if (unschedulableTaskSets > 0) {
       // Request additional executors to account for task sets having tasks that are unschedulable
-      // due to blacklisting when the active executor count has already reached the max needed
-      // which we would normally get.
+      // due to executors excluded for failures when the active executor count has already reached
+      // the max needed which we would normally get.
       val maxNeededForUnschedulables = math.ceil(unschedulableTaskSets * executorAllocationRatio /
         tasksPerExecutor).toInt
       math.max(maxNeededWithSpeculationLocalityOffset,
@@ -659,10 +659,10 @@ private[spark] class ExecutorAllocationManager(
     private val resourceProfileIdToStageAttempt =
       new mutable.HashMap[Int, mutable.Set[StageAttempt]]
 
-    // Keep track of unschedulable task sets due to blacklisting. This is a Set of StageAttempt's
-    // because we'll only take the last unschedulable task in a taskset although there can be more.
-    // This is done in order to avoid costly loops in the scheduling.
-    // Check TaskSetManager#getCompletelyBlacklistedTaskIfAny for more details.
+    // Keep track of unschedulable task sets because of executor/node exclusions from too many task
+    // failures. This is a Set of StageAttempt's because we'll only take the last unschedulable task
+    // in a taskset although there can be more. This is done in order to avoid costly loops in the
+    // scheduling. Check TaskSetManager#getCompletelyExcludedTaskIfAny for more details.
     private val unschedulableTaskSets = new mutable.HashSet[StageAttempt]
 
     // stageAttempt to tuple (the number of task with locality preferences, a map where each pair

core/src/main/scala/org/apache/spark/SparkConf.scala

@@ -603,7 +603,7 @@ private[spark] object SparkConf extends Logging {
           "are no longer accepted. To specify the equivalent now, one may use '64k'."),
       DeprecatedConfig("spark.rpc", "2.0", "Not used anymore."),
       DeprecatedConfig("spark.scheduler.executorTaskBlacklistTime", "2.1.0",
-        "Please use the new blacklisting options, spark.blacklist.*"),
+        "Please use the new excludedOnFailure options, spark.excludeOnFailure.*"),
       DeprecatedConfig("spark.yarn.am.port", "2.0.0", "Not used anymore"),
       DeprecatedConfig("spark.executor.port", "2.0.0", "Not used anymore"),
       DeprecatedConfig("spark.shuffle.service.index.cache.entries", "2.3.0",
@@ -612,7 +612,29 @@ private[spark] object SparkConf extends Logging {
       DeprecatedConfig("spark.yarn.credentials.file.retention.days", "2.4.0", "Not used anymore."),
       DeprecatedConfig("spark.yarn.services", "3.0.0", "Feature no longer available."),
       DeprecatedConfig("spark.executor.plugins", "3.0.0",
-        "Feature replaced with new plugin API. See Monitoring documentation.")
+        "Feature replaced with new plugin API. See Monitoring documentation."),
+      DeprecatedConfig("spark.blacklist.enabled", "3.1.0",
+        "Please use spark.excludeOnFailure.enabled"),
+      DeprecatedConfig("spark.blacklist.task.maxTaskAttemptsPerExecutor", "3.1.0",
+        "Please use spark.excludeOnFailure.task.maxTaskAttemptsPerExecutor"),
+      DeprecatedConfig("spark.blacklist.task.maxTaskAttemptsPerNode", "3.1.0",
+        "Please use spark.excludeOnFailure.task.maxTaskAttemptsPerNode"),
+      DeprecatedConfig("spark.blacklist.application.maxFailedTasksPerExecutor", "3.1.0",
+        "Please use spark.excludeOnFailure.application.maxFailedTasksPerExecutor"),
+      DeprecatedConfig("spark.blacklist.stage.maxFailedTasksPerExecutor", "3.1.0",
+        "Please use spark.excludeOnFailure.stage.maxFailedTasksPerExecutor"),
+      DeprecatedConfig("spark.blacklist.application.maxFailedExecutorsPerNode", "3.1.0",
+        "Please use spark.excludeOnFailure.application.maxFailedExecutorsPerNode"),
+      DeprecatedConfig("spark.blacklist.stage.maxFailedExecutorsPerNode", "3.1.0",
+        "Please use spark.excludeOnFailure.stage.maxFailedExecutorsPerNode"),
+      DeprecatedConfig("spark.blacklist.timeout", "3.1.0",
+        "Please use spark.excludeOnFailure.timeout"),
+      DeprecatedConfig("spark.scheduler.executorTaskBlacklistTime", "3.1.0",
+        "Please use spark.scheduler.executorTaskExcludeOnFailureTime"),
+      DeprecatedConfig("spark.blacklist.application.fetchFailure.enabled", "3.1.0",
+        "Please use spark.excludeOnFailure.application.fetchFailure.enabled"),
+      DeprecatedConfig("spark.scheduler.blacklist.unschedulableTaskSetTimeout", "3.1.0",
+        "Please use spark.scheduler.excludeOnFailure.unschedulableTaskSetTimeout")
     )
 
     Map(configs.map { cfg => (cfg.key -> cfg) } : _*)

core/src/main/scala/org/apache/spark/TaskEndReason.scala

@@ -98,10 +98,11 @@ case class FetchFailed(
   /**
    * Fetch failures lead to a different failure handling path: (1) we don't abort the stage after
    * 4 task failures, instead we immediately go back to the stage which generated the map output,
-   * and regenerate the missing data.  (2) we don't count fetch failures for blacklisting, since
-   * presumably its not the fault of the executor where the task ran, but the executor which
-   * stored the data. This is especially important because we might rack up a bunch of
-   * fetch-failures in rapid succession, on all nodes of the cluster, due to one bad node.
+   * and regenerate the missing data. (2) we don't count fetch failures from executors excluded
+   * due to too many task failures, since presumably its not the fault of the executor where
+   * the task ran, but the executor which stored the data. This is especially important because
+   * we might rack up a bunch of fetch-failures in rapid succession, on all nodes of the cluster,
+   * due to one bad node.
    */
   override def countTowardsTaskFailures: Boolean = false
 }

core/src/main/scala/org/apache/spark/deploy/history/BasicEventFilterBuilder.scala

@@ -160,6 +160,8 @@ private[spark] class BasicEventFilter(
     case e: SparkListenerExecutorRemoved => liveExecutors.contains(e.executorId)
     case e: SparkListenerExecutorBlacklisted => liveExecutors.contains(e.executorId)
     case e: SparkListenerExecutorUnblacklisted => liveExecutors.contains(e.executorId)
+    case e: SparkListenerExecutorExcluded => liveExecutors.contains(e.executorId)
+    case e: SparkListenerExecutorUnexcluded => liveExecutors.contains(e.executorId)
     case e: SparkListenerStageExecutorMetrics => liveExecutors.contains(e.execId)
     case e: SparkListenerBlockManagerAdded => acceptBlockManagerEvent(e.blockManagerId)
     case e: SparkListenerBlockManagerRemoved => acceptBlockManagerEvent(e.blockManagerId)

core/src/main/scala/org/apache/spark/deploy/history/HistoryAppStatusStore.scala

@@ -73,7 +73,7 @@ private[spark] class HistoryAppStatusStore(
       source.totalShuffleWrite, source.isBlacklisted, source.maxMemory, source.addTime,
       source.removeTime, source.removeReason, newExecutorLogs, source.memoryMetrics,
       source.blacklistedInStages, source.peakMemoryMetrics, source.attributes, source.resources,
-      source.resourceProfileId)
+      source.resourceProfileId, source.isExcluded, source.excludedInStages)
   }
 
 }