MAPREDUCE-4607. Race condition in ReduceTask completion can result in Task being incorrectly failed. Contributed by Bikas Saha.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1383422 13f79535-47bb-0310-9956-ffa450edef68

hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/main/java/org/apache/hadoop/mapreduce/v2/app/job/impl/TaskAttemptImpl.java

@@ -71,6 +71,7 @@
 import org.apache.hadoop.mapreduce.v2.api.records.TaskAttemptReport;
 import org.apache.hadoop.mapreduce.v2.api.records.TaskAttemptState;
 import org.apache.hadoop.mapreduce.v2.api.records.TaskId;
+import org.apache.hadoop.mapreduce.v2.api.records.TaskState;
 import org.apache.hadoop.mapreduce.v2.api.records.TaskType;
 import org.apache.hadoop.mapreduce.v2.app.AppContext;
 import org.apache.hadoop.mapreduce.v2.app.TaskAttemptListener;
@@ -86,6 +87,7 @@
 import org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptEventType;
 import org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptKillEvent;
 import org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptStatusUpdateEvent;
+import org.apache.hadoop.mapreduce.v2.app.job.event.TaskEvent;
 import org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptStatusUpdateEvent.TaskAttemptStatus;
 import org.apache.hadoop.mapreduce.v2.app.job.event.TaskEventType;
 import org.apache.hadoop.mapreduce.v2.app.job.event.TaskTAttemptEvent;
@@ -120,6 +122,7 @@
 import org.apache.hadoop.yarn.factories.RecordFactory;
 import org.apache.hadoop.yarn.factory.providers.RecordFactoryProvider;
 import org.apache.hadoop.yarn.state.InvalidStateTransitonException;
+import org.apache.hadoop.yarn.state.MultipleArcTransition;
 import org.apache.hadoop.yarn.state.SingleArcTransition;
 import org.apache.hadoop.yarn.state.StateMachine;
 import org.apache.hadoop.yarn.state.StateMachineFactory;
@@ -128,6 +131,8 @@
 import org.apache.hadoop.yarn.util.ConverterUtils;
 import org.apache.hadoop.yarn.util.RackResolver;
 
+import com.google.common.base.Preconditions;
+
 /**
  * Implementation of TaskAttempt interface.
  */
@@ -404,10 +409,10 @@ TaskAttemptEventType.TA_CONTAINER_CLEANED, new TaskCleanupTransition())
          TaskAttemptState.FAILED,
          TaskAttemptEventType.TA_TOO_MANY_FETCH_FAILURE,
          new TooManyFetchFailureTransition())
-     .addTransition(
+      .addTransition(TaskAttemptState.SUCCEEDED,
-         TaskAttemptState.SUCCEEDED, TaskAttemptState.KILLED,
+          EnumSet.of(TaskAttemptState.SUCCEEDED, TaskAttemptState.KILLED),
-         TaskAttemptEventType.TA_KILL,
+          TaskAttemptEventType.TA_KILL, 
-         new KilledAfterSuccessTransition())
+          new KilledAfterSuccessTransition())
      .addTransition(
          TaskAttemptState.SUCCEEDED, TaskAttemptState.SUCCEEDED,
          TaskAttemptEventType.TA_DIAGNOSTICS_UPDATE,
@@ -1483,6 +1488,9 @@ private static class TooManyFetchFailureTransition implements
     @SuppressWarnings("unchecked")
     @Override
     public void transition(TaskAttemptImpl taskAttempt, TaskAttemptEvent event) {
+      // too many fetch failure can only happen for map tasks
+      Preconditions
+          .checkArgument(taskAttempt.getID().getTaskId().getTaskType() == TaskType.MAP);
       //add to diagnostic
       taskAttempt.addDiagnosticInfo("Too Many fetch failures.Failing the attempt");
       //set the finish time
@@ -1506,15 +1514,30 @@ public void transition(TaskAttemptImpl taskAttempt, TaskAttemptEvent event) {
   }
   
   private static class KilledAfterSuccessTransition implements
-      SingleArcTransition<TaskAttemptImpl, TaskAttemptEvent> {
+      MultipleArcTransition<TaskAttemptImpl, TaskAttemptEvent, TaskAttemptState> {
 
     @SuppressWarnings("unchecked")
     @Override
-    public void transition(TaskAttemptImpl taskAttempt, 
+    public TaskAttemptState transition(TaskAttemptImpl taskAttempt, 
         TaskAttemptEvent event) {
-      TaskAttemptKillEvent msgEvent = (TaskAttemptKillEvent) event;
+      if(taskAttempt.getID().getTaskId().getTaskType() == TaskType.REDUCE) {
-      //add to diagnostic
+        // after a reduce task has succeeded, its outputs are in safe in HDFS.
-      taskAttempt.addDiagnosticInfo(msgEvent.getMessage());
+        // logically such a task should not be killed. we only come here when
+        // there is a race condition in the event queue. E.g. some logic sends
+        // a kill request to this attempt when the successful completion event
+        // for this task is already in the event queue. so the kill event will
+        // get executed immediately after the attempt is marked successful and 
+        // result in this transition being exercised.
+        // ignore this for reduce tasks
+        LOG.info("Ignoring killed event for successful reduce task attempt" +
+                  taskAttempt.getID().toString());
+        return TaskAttemptState.SUCCEEDED;
+      }
+      if(event instanceof TaskAttemptKillEvent) {
+        TaskAttemptKillEvent msgEvent = (TaskAttemptKillEvent) event;
+        //add to diagnostic
+        taskAttempt.addDiagnosticInfo(msgEvent.getMessage());
+      }
 
       // not setting a finish time since it was set on success
       assert (taskAttempt.getFinishTime() != 0);
@@ -1528,6 +1551,7 @@ public void transition(TaskAttemptImpl taskAttempt,
           .getTaskId().getJobId(), tauce));
       taskAttempt.eventHandler.handle(new TaskTAttemptEvent(
           taskAttempt.attemptId, TaskEventType.T_ATTEMPT_KILLED));
+      return TaskAttemptState.KILLED;
     }
   }
 

hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/main/java/org/apache/hadoop/mapreduce/v2/app/job/impl/TaskImpl.java

@@ -191,12 +191,12 @@ TaskEventType.T_ADD_SPEC_ATTEMPT, new RedundantScheduleTransition())
             TaskEventType.T_ADD_SPEC_ATTEMPT))
 
     // Transitions from SUCCEEDED state
-    .addTransition(TaskState.SUCCEEDED, //only possible for map tasks
+    .addTransition(TaskState.SUCCEEDED,
         EnumSet.of(TaskState.SCHEDULED, TaskState.SUCCEEDED, TaskState.FAILED),
-        TaskEventType.T_ATTEMPT_FAILED, new MapRetroactiveFailureTransition())
+        TaskEventType.T_ATTEMPT_FAILED, new RetroactiveFailureTransition())
-    .addTransition(TaskState.SUCCEEDED, //only possible for map tasks
+    .addTransition(TaskState.SUCCEEDED,
         EnumSet.of(TaskState.SCHEDULED, TaskState.SUCCEEDED),
-        TaskEventType.T_ATTEMPT_KILLED, new MapRetroactiveKilledTransition())
+        TaskEventType.T_ATTEMPT_KILLED, new RetroactiveKilledTransition())
     // Ignore-able transitions.
     .addTransition(
         TaskState.SUCCEEDED, TaskState.SUCCEEDED,
@@ -897,7 +897,7 @@ protected TaskState getDefaultState(Task task) {
     }
   }
 
-  private static class MapRetroactiveFailureTransition
+  private static class RetroactiveFailureTransition
       extends AttemptFailedTransition {
 
     @Override
@@ -911,8 +911,8 @@ public TaskState transition(TaskImpl task, TaskEvent event) {
           return TaskState.SUCCEEDED;
         }
       }
-      
+
-      //verify that this occurs only for map task
+      // a successful REDUCE task should not be overridden
       //TODO: consider moving it to MapTaskImpl
       if (!TaskType.MAP.equals(task.getType())) {
         LOG.error("Unexpected event for REDUCE task " + event.getType());
@@ -938,42 +938,46 @@ protected TaskState getDefaultState(Task task) {
     }
   }
 
-  private static class MapRetroactiveKilledTransition implements
+  private static class RetroactiveKilledTransition implements
     MultipleArcTransition<TaskImpl, TaskEvent, TaskState> {
 
     @Override
     public TaskState transition(TaskImpl task, TaskEvent event) {
-      // verify that this occurs only for map task
+      TaskAttemptId attemptId = null;
+      if (event instanceof TaskTAttemptEvent) {
+        TaskTAttemptEvent castEvent = (TaskTAttemptEvent) event;
+        attemptId = castEvent.getTaskAttemptID(); 
+        if (task.getState() == TaskState.SUCCEEDED &&
+            !attemptId.equals(task.successfulAttempt)) {
+          // don't allow a different task attempt to override a previous
+          // succeeded state
+          return TaskState.SUCCEEDED;
+        }
+      }
+
+      // a successful REDUCE task should not be overridden
       // TODO: consider moving it to MapTaskImpl
       if (!TaskType.MAP.equals(task.getType())) {
         LOG.error("Unexpected event for REDUCE task " + event.getType());
         task.internalError(event.getType());
       }
 
-      TaskTAttemptEvent attemptEvent = (TaskTAttemptEvent) event;
+      // successful attempt is now killed. reschedule
-      TaskAttemptId attemptId = attemptEvent.getTaskAttemptID();
+      // tell the job about the rescheduling
-      if(task.successfulAttempt == attemptId) {
+      unSucceed(task);
-        // successful attempt is now killed. reschedule
+      task.handleTaskAttemptCompletion(attemptId,
-        // tell the job about the rescheduling
+          TaskAttemptCompletionEventStatus.KILLED);
-        unSucceed(task);
+      task.eventHandler.handle(new JobMapTaskRescheduledEvent(task.taskId));
-        task.handleTaskAttemptCompletion(
+      // typically we are here because this map task was run on a bad node and
-            attemptId, 
+      // we want to reschedule it on a different node.
-            TaskAttemptCompletionEventStatus.KILLED);
+      // Depending on whether there are previous failed attempts or not this
-        task.eventHandler.handle(new JobMapTaskRescheduledEvent(task.taskId));
+      // can SCHEDULE or RESCHEDULE the container allocate request. If this
-        // typically we are here because this map task was run on a bad node and 
+      // SCHEDULE's then the dataLocal hosts of this taskAttempt will be used
-        // we want to reschedule it on a different node.
+      // from the map splitInfo. So the bad node might be sent as a location
-        // Depending on whether there are previous failed attempts or not this 
+      // to the RM. But the RM would ignore that just like it would ignore
-        // can SCHEDULE or RESCHEDULE the container allocate request. If this
+      // currently pending container requests affinitized to bad nodes.
-        // SCHEDULE's then the dataLocal hosts of this taskAttempt will be used
+      task.addAndScheduleAttempt();
-        // from the map splitInfo. So the bad node might be sent as a location 
+      return TaskState.SCHEDULED;
-        // to the RM. But the RM would ignore that just like it would ignore 
-        // currently pending container requests affinitized to bad nodes.
-        task.addAndScheduleAttempt();
-        return TaskState.SCHEDULED;
-      } else {
-        // nothing to do
-        return TaskState.SUCCEEDED;
-      }
     }
   }
 

hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/test/java/org/apache/hadoop/mapreduce/v2/app/TestMRApp.java

@@ -180,7 +180,7 @@ public void testCompletedMapsForReduceSlowstart() throws Exception {
   @Test
   public void testUpdatedNodes() throws Exception {
     int runCount = 0;
-    MRApp app = new MRAppWithHistory(2, 1, false, this.getClass().getName(),
+    MRApp app = new MRAppWithHistory(2, 2, false, this.getClass().getName(),
         true, ++runCount);
     Configuration conf = new Configuration();
     // after half of the map completion, reduce will start
@@ -189,7 +189,7 @@ public void testUpdatedNodes() throws Exception {
     conf.setBoolean(MRJobConfig.JOB_UBERTASK_ENABLE, false);
     Job job = app.submit(conf);
     app.waitForState(job, JobState.RUNNING);
-    Assert.assertEquals("Num tasks not correct", 3, job.getTasks().size());
+    Assert.assertEquals("Num tasks not correct", 4, job.getTasks().size());
     Iterator<Task> it = job.getTasks().values().iterator();
     Task mapTask1 = it.next();
     Task mapTask2 = it.next();
@@ -272,18 +272,19 @@ public void testUpdatedNodes() throws Exception {
 
     // rerun
     // in rerun the 1st map will be recovered from previous run
-    app = new MRAppWithHistory(2, 1, false, this.getClass().getName(), false,
+    app = new MRAppWithHistory(2, 2, false, this.getClass().getName(), false,
         ++runCount);
     conf = new Configuration();
     conf.setBoolean(MRJobConfig.MR_AM_JOB_RECOVERY_ENABLE, true);
     conf.setBoolean(MRJobConfig.JOB_UBERTASK_ENABLE, false);
     job = app.submit(conf);
     app.waitForState(job, JobState.RUNNING);
-    Assert.assertEquals("No of tasks not correct", 3, job.getTasks().size());
+    Assert.assertEquals("No of tasks not correct", 4, job.getTasks().size());
     it = job.getTasks().values().iterator();
     mapTask1 = it.next();
     mapTask2 = it.next();
-    Task reduceTask = it.next();
+    Task reduceTask1 = it.next();
+    Task reduceTask2 = it.next();
 
     // map 1 will be recovered, no need to send done
     app.waitForState(mapTask1, TaskState.SUCCEEDED);
@@ -306,19 +307,36 @@ public void testUpdatedNodes() throws Exception {
     Assert.assertEquals("Expecting 1 more completion events for success", 3,
         events.length);
 
-    app.waitForState(reduceTask, TaskState.RUNNING);
+    app.waitForState(reduceTask1, TaskState.RUNNING);
-    TaskAttempt task3Attempt = reduceTask.getAttempts().values().iterator()
+    app.waitForState(reduceTask2, TaskState.RUNNING);
+
+    TaskAttempt task3Attempt = reduceTask1.getAttempts().values().iterator()
         .next();
     app.getContext()
         .getEventHandler()
         .handle(
             new TaskAttemptEvent(task3Attempt.getID(),
                 TaskAttemptEventType.TA_DONE));
-    app.waitForState(reduceTask, TaskState.SUCCEEDED);
+    app.waitForState(reduceTask1, TaskState.SUCCEEDED);
+    app.getContext()
+    .getEventHandler()
+    .handle(
+        new TaskAttemptEvent(task3Attempt.getID(),
+            TaskAttemptEventType.TA_KILL));
+    app.waitForState(reduceTask1, TaskState.SUCCEEDED);
+    
+    TaskAttempt task4Attempt = reduceTask2.getAttempts().values().iterator()
+        .next();
+    app.getContext()
+        .getEventHandler()
+        .handle(
+            new TaskAttemptEvent(task4Attempt.getID(),
+                TaskAttemptEventType.TA_DONE));
+    app.waitForState(reduceTask2, TaskState.SUCCEEDED);    
 
     events = job.getTaskAttemptCompletionEvents(0, 100);
-    Assert.assertEquals("Expecting 1 more completion events for success", 4,
+    Assert.assertEquals("Expecting 2 more completion events for reduce success",
-        events.length);
+        5, events.length);
 
     // job succeeds
     app.waitForState(job, JobState.SUCCEEDED);

hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/test/java/org/apache/hadoop/mapreduce/v2/app/job/impl/TestTaskImpl.java

@@ -84,7 +84,6 @@ public class TestTaskImpl {
   private ApplicationId appId;
   private TaskSplitMetaInfo taskSplitMetaInfo;  
   private String[] dataLocations = new String[0]; 
-  private final TaskType taskType = TaskType.MAP;
   private AppContext appContext;
   
   private int startCount = 0;
@@ -97,6 +96,7 @@ public class TestTaskImpl {
   private class MockTaskImpl extends TaskImpl {
         
     private int taskAttemptCounter = 0;
+    TaskType taskType;
 
     public MockTaskImpl(JobId jobId, int partition,
         EventHandler eventHandler, Path remoteJobConfFile, JobConf conf,
@@ -104,11 +104,12 @@ public MockTaskImpl(JobId jobId, int partition,
         Token<JobTokenIdentifier> jobToken,
         Credentials credentials, Clock clock,
         Map<TaskId, TaskInfo> completedTasksFromPreviousRun, int startCount,
-        MRAppMetrics metrics, AppContext appContext) {
+        MRAppMetrics metrics, AppContext appContext, TaskType taskType) {
       super(jobId, taskType , partition, eventHandler,
           remoteJobConfFile, conf, taskAttemptListener, committer, 
           jobToken, credentials, clock,
           completedTasksFromPreviousRun, startCount, metrics, appContext);
+      this.taskType = taskType;
     }
 
     @Override
@@ -120,7 +121,7 @@ public TaskType getType() {
     protected TaskAttemptImpl createAttempt() {
       MockTaskAttemptImpl attempt = new MockTaskAttemptImpl(getID(), ++taskAttemptCounter, 
           eventHandler, taskAttemptListener, remoteJobConfFile, partition,
-          conf, committer, jobToken, credentials, clock, appContext);
+          conf, committer, jobToken, credentials, clock, appContext, taskType);
       taskAttempts.add(attempt);
       return attempt;
     }
@@ -142,18 +143,20 @@ private class MockTaskAttemptImpl extends TaskAttemptImpl {
     private float progress = 0;
     private TaskAttemptState state = TaskAttemptState.NEW;
     private TaskAttemptId attemptId;
+    private TaskType taskType;
 
     public MockTaskAttemptImpl(TaskId taskId, int id, EventHandler eventHandler,
         TaskAttemptListener taskAttemptListener, Path jobFile, int partition,
         JobConf conf, OutputCommitter committer,
         Token<JobTokenIdentifier> jobToken,
         Credentials credentials, Clock clock,
-        AppContext appContext) {
+        AppContext appContext, TaskType taskType) {
       super(taskId, id, eventHandler, taskAttemptListener, jobFile, partition, conf,
           dataLocations, committer, jobToken, credentials, clock, appContext);
       attemptId = Records.newRecord(TaskAttemptId.class);
       attemptId.setId(id);
       attemptId.setTaskId(taskId);
+      this.taskType = taskType;
     }
 
     public TaskAttemptId getAttemptId() {
@@ -162,7 +165,7 @@ public TaskAttemptId getAttemptId() {
     
     @Override
     protected Task createRemoteTask() {
-      return new MockTask();
+      return new MockTask(taskType);
     }    
     
     public float getProgress() {
@@ -185,6 +188,11 @@ public TaskAttemptState getState() {
   
   private class MockTask extends Task {
 
+    private TaskType taskType;
+    MockTask(TaskType taskType) {
+      this.taskType = taskType;
+    }
+    
     @Override
     public void run(JobConf job, TaskUmbilicalProtocol umbilical)
         throws IOException, ClassNotFoundException, InterruptedException {
@@ -193,7 +201,7 @@ public void run(JobConf job, TaskUmbilicalProtocol umbilical)
 
     @Override
     public boolean isMapTask() {
-      return true;
+      return (taskType == TaskType.MAP);
     }    
     
   }
@@ -227,14 +235,15 @@ public void setup() {
     taskSplitMetaInfo = mock(TaskSplitMetaInfo.class);
     when(taskSplitMetaInfo.getLocations()).thenReturn(dataLocations); 
     
-    taskAttempts = new ArrayList<MockTaskAttemptImpl>();
+    taskAttempts = new ArrayList<MockTaskAttemptImpl>();    
-    
+  }
-    mockTask = new MockTaskImpl(jobId, partition, dispatcher.getEventHandler(),
+  
+  private MockTaskImpl createMockTask(TaskType taskType) {
+    return new MockTaskImpl(jobId, partition, dispatcher.getEventHandler(),
         remoteJobConfFile, conf, taskAttemptListener, committer, jobToken,
         credentials, clock,
         completedTasksFromPreviousRun, startCount,
-        metrics, appContext);        
+        metrics, appContext, taskType);
-    
   }
 
   @After 
@@ -342,6 +351,7 @@ private void assertTaskSucceededState() {
   @Test
   public void testInit() {
     LOG.info("--- START: testInit ---");
+    mockTask = createMockTask(TaskType.MAP);        
     assertTaskNewState();
     assert(taskAttempts.size() == 0);
   }
@@ -352,6 +362,7 @@ public void testInit() {
    */
   public void testScheduleTask() {
     LOG.info("--- START: testScheduleTask ---");
+    mockTask = createMockTask(TaskType.MAP);        
     TaskId taskId = getNewTaskID();
     scheduleTaskAttempt(taskId);
   }
@@ -362,6 +373,7 @@ public void testScheduleTask() {
    */
   public void testKillScheduledTask() {
     LOG.info("--- START: testKillScheduledTask ---");
+    mockTask = createMockTask(TaskType.MAP);        
     TaskId taskId = getNewTaskID();
     scheduleTaskAttempt(taskId);
     killTask(taskId);
@@ -374,6 +386,7 @@ public void testKillScheduledTask() {
    */
   public void testKillScheduledTaskAttempt() {
     LOG.info("--- START: testKillScheduledTaskAttempt ---");
+    mockTask = createMockTask(TaskType.MAP);        
     TaskId taskId = getNewTaskID();
     scheduleTaskAttempt(taskId);
     killScheduledTaskAttempt(getLastAttempt().getAttemptId());
@@ -386,6 +399,7 @@ public void testKillScheduledTaskAttempt() {
    */
   public void testLaunchTaskAttempt() {
     LOG.info("--- START: testLaunchTaskAttempt ---");
+    mockTask = createMockTask(TaskType.MAP);        
     TaskId taskId = getNewTaskID();
     scheduleTaskAttempt(taskId);
     launchTaskAttempt(getLastAttempt().getAttemptId());
@@ -398,6 +412,7 @@ public void testLaunchTaskAttempt() {
    */
   public void testKillRunningTaskAttempt() {
     LOG.info("--- START: testKillRunningTaskAttempt ---");
+    mockTask = createMockTask(TaskType.MAP);        
     TaskId taskId = getNewTaskID();
     scheduleTaskAttempt(taskId);
     launchTaskAttempt(getLastAttempt().getAttemptId());
@@ -407,6 +422,7 @@ public void testKillRunningTaskAttempt() {
   @Test 
   public void testTaskProgress() {
     LOG.info("--- START: testTaskProgress ---");
+    mockTask = createMockTask(TaskType.MAP);        
         
     // launch task
     TaskId taskId = getNewTaskID();
@@ -444,6 +460,7 @@ public void testTaskProgress() {
   
   @Test
   public void testFailureDuringTaskAttemptCommit() {
+    mockTask = createMockTask(TaskType.MAP);        
     TaskId taskId = getNewTaskID();
     scheduleTaskAttempt(taskId);
     launchTaskAttempt(getLastAttempt().getAttemptId());
@@ -469,8 +486,7 @@ public void testFailureDuringTaskAttemptCommit() {
     assertTaskSucceededState();
   }
   
-  @Test
+  private void runSpeculativeTaskAttemptSucceedsEvenIfFirstFails(TaskEventType failEvent) {
-  public void testSpeculativeTaskAttemptSucceedsEvenIfFirstFails() {
     TaskId taskId = getNewTaskID();
     scheduleTaskAttempt(taskId);
     launchTaskAttempt(getLastAttempt().getAttemptId());
@@ -489,11 +505,34 @@ public void testSpeculativeTaskAttemptSucceedsEvenIfFirstFails() {
     
     // Now fail the first task attempt, after the second has succeeded
     mockTask.handle(new TaskTAttemptEvent(taskAttempts.get(0).getAttemptId(), 
-        TaskEventType.T_ATTEMPT_FAILED));
+        failEvent));
     
     // The task should still be in the succeeded state
     assertTaskSucceededState();
-    
+  }
+  
+  @Test
+  public void testMapSpeculativeTaskAttemptSucceedsEvenIfFirstFails() {
+    mockTask = createMockTask(TaskType.MAP);        
+    runSpeculativeTaskAttemptSucceedsEvenIfFirstFails(TaskEventType.T_ATTEMPT_FAILED);
+  }
+
+  @Test
+  public void testReduceSpeculativeTaskAttemptSucceedsEvenIfFirstFails() {
+    mockTask = createMockTask(TaskType.REDUCE);        
+    runSpeculativeTaskAttemptSucceedsEvenIfFirstFails(TaskEventType.T_ATTEMPT_FAILED);
+  }
+  
+  @Test
+  public void testMapSpeculativeTaskAttemptSucceedsEvenIfFirstIsKilled() {
+    mockTask = createMockTask(TaskType.MAP);        
+    runSpeculativeTaskAttemptSucceedsEvenIfFirstFails(TaskEventType.T_ATTEMPT_KILLED);
+  }
+
+  @Test
+  public void testReduceSpeculativeTaskAttemptSucceedsEvenIfFirstIsKilled() {
+    mockTask = createMockTask(TaskType.REDUCE);        
+    runSpeculativeTaskAttemptSucceedsEvenIfFirstFails(TaskEventType.T_ATTEMPT_KILLED);
   }
 
 }