hadoop-yarn-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From vinayakum...@apache.org
Subject svn commit: r1601869 [2/3] - in /hadoop/common/branches/HDFS-5442/hadoop-yarn-project: ./ hadoop-yarn/hadoop-yarn-api/src/main/java/org/apache/hadoop/yarn/api/records/ hadoop-yarn/hadoop-yarn-api/src/main/java/org/apache/hadoop/yarn/conf/ hadoop-yarn/h...
Date Wed, 11 Jun 2014 12:20:51 GMT
Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/ApplicationAttemptStateData.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/ApplicationAttemptStateData.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/ApplicationAttemptStateData.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/ApplicationAttemptStateData.java Wed Jun 11 12:20:48 2014
@@ -18,31 +18,73 @@
 
 package org.apache.hadoop.yarn.server.resourcemanager.recovery.records;
 
+import java.io.IOException;
 import java.nio.ByteBuffer;
 
 import org.apache.hadoop.classification.InterfaceAudience.Public;
 import org.apache.hadoop.classification.InterfaceStability.Unstable;
+import org.apache.hadoop.io.DataOutputBuffer;
+import org.apache.hadoop.security.Credentials;
 import org.apache.hadoop.yarn.api.records.ApplicationAttemptId;
 import org.apache.hadoop.yarn.api.records.Container;
 import org.apache.hadoop.yarn.api.records.FinalApplicationStatus;
+import org.apache.hadoop.yarn.proto.YarnServerResourceManagerServiceProtos.ApplicationAttemptStateDataProto;
+import org.apache.hadoop.yarn.server.resourcemanager.recovery.RMStateStore.ApplicationAttemptState;
 import org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptState;
+import org.apache.hadoop.yarn.util.Records;
 
 /*
  * Contains the state data that needs to be persisted for an ApplicationAttempt
  */
 @Public
 @Unstable
-public interface ApplicationAttemptStateData {
-  
+public abstract class ApplicationAttemptStateData {
+  public static ApplicationAttemptStateData newInstance(
+      ApplicationAttemptId attemptId, Container container,
+      ByteBuffer attemptTokens, long startTime, RMAppAttemptState finalState,
+      String finalTrackingUrl, String diagnostics,
+      FinalApplicationStatus amUnregisteredFinalStatus) {
+    ApplicationAttemptStateData attemptStateData =
+        Records.newRecord(ApplicationAttemptStateData.class);
+    attemptStateData.setAttemptId(attemptId);
+    attemptStateData.setMasterContainer(container);
+    attemptStateData.setAppAttemptTokens(attemptTokens);
+    attemptStateData.setState(finalState);
+    attemptStateData.setFinalTrackingUrl(finalTrackingUrl);
+    attemptStateData.setDiagnostics(diagnostics);
+    attemptStateData.setStartTime(startTime);
+    attemptStateData.setFinalApplicationStatus(amUnregisteredFinalStatus);
+    return attemptStateData;
+  }
+
+  public static ApplicationAttemptStateData newInstance(
+      ApplicationAttemptState attemptState) throws IOException {
+    Credentials credentials = attemptState.getAppAttemptCredentials();
+    ByteBuffer appAttemptTokens = null;
+    if (credentials != null) {
+      DataOutputBuffer dob = new DataOutputBuffer();
+      credentials.writeTokenStorageToStream(dob);
+      appAttemptTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength());
+    }
+    return newInstance(attemptState.getAttemptId(),
+        attemptState.getMasterContainer(), appAttemptTokens,
+        attemptState.getStartTime(), attemptState.getState(),
+        attemptState.getFinalTrackingUrl(),
+        attemptState.getDiagnostics(),
+        attemptState.getFinalApplicationStatus());
+  }
+
+  public abstract ApplicationAttemptStateDataProto getProto();
+
   /**
    * The ApplicationAttemptId for the application attempt
    * @return ApplicationAttemptId for the application attempt
    */
   @Public
   @Unstable
-  public ApplicationAttemptId getAttemptId();
+  public abstract ApplicationAttemptId getAttemptId();
   
-  public void setAttemptId(ApplicationAttemptId attemptId);
+  public abstract void setAttemptId(ApplicationAttemptId attemptId);
   
   /*
    * The master container running the application attempt
@@ -50,9 +92,9 @@ public interface ApplicationAttemptState
    */
   @Public
   @Unstable
-  public Container getMasterContainer();
+  public abstract Container getMasterContainer();
   
-  public void setMasterContainer(Container container);
+  public abstract void setMasterContainer(Container container);
 
   /**
    * The application attempt tokens that belong to this attempt
@@ -60,17 +102,17 @@ public interface ApplicationAttemptState
    */
   @Public
   @Unstable
-  public ByteBuffer getAppAttemptTokens();
+  public abstract ByteBuffer getAppAttemptTokens();
 
-  public void setAppAttemptTokens(ByteBuffer attemptTokens);
+  public abstract void setAppAttemptTokens(ByteBuffer attemptTokens);
 
   /**
    * Get the final state of the application attempt.
    * @return the final state of the application attempt.
    */
-  public RMAppAttemptState getState();
+  public abstract RMAppAttemptState getState();
 
-  public void setState(RMAppAttemptState state);
+  public abstract void setState(RMAppAttemptState state);
 
   /**
    * Get the original not-proxied <em>final tracking url</em> for the
@@ -79,34 +121,34 @@ public interface ApplicationAttemptState
    * @return the original not-proxied <em>final tracking url</em> for the
    *         application
    */
-  public String getFinalTrackingUrl();
+  public abstract String getFinalTrackingUrl();
 
   /**
    * Set the final tracking Url of the AM.
    * @param url
    */
-  public void setFinalTrackingUrl(String url);
+  public abstract void setFinalTrackingUrl(String url);
   /**
    * Get the <em>diagnositic information</em> of the attempt 
    * @return <em>diagnositic information</em> of the attempt
    */
-  public String getDiagnostics();
+  public abstract String getDiagnostics();
 
-  public void setDiagnostics(String diagnostics);
+  public abstract void setDiagnostics(String diagnostics);
 
   /**
    * Get the <em>start time</em> of the application.
    * @return <em>start time</em> of the application
    */
-  public long getStartTime();
+  public abstract long getStartTime();
 
-  public void setStartTime(long startTime);
+  public abstract void setStartTime(long startTime);
 
   /**
    * Get the <em>final finish status</em> of the application.
    * @return <em>final finish status</em> of the application
    */
-  public FinalApplicationStatus getFinalApplicationStatus();
+  public abstract FinalApplicationStatus getFinalApplicationStatus();
 
-  public void setFinalApplicationStatus(FinalApplicationStatus finishState);
+  public abstract void setFinalApplicationStatus(FinalApplicationStatus finishState);
 }

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/ApplicationStateData.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/ApplicationStateData.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/ApplicationStateData.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/ApplicationStateData.java Wed Jun 11 12:20:48 2014
@@ -24,7 +24,10 @@ import org.apache.hadoop.classification.
 import org.apache.hadoop.classification.InterfaceStability.Unstable;
 import org.apache.hadoop.yarn.api.records.ApplicationId;
 import org.apache.hadoop.yarn.api.records.ApplicationSubmissionContext;
+import org.apache.hadoop.yarn.proto.YarnServerResourceManagerServiceProtos.ApplicationStateDataProto;
+import org.apache.hadoop.yarn.server.resourcemanager.recovery.RMStateStore.ApplicationState;
 import org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMAppState;
+import org.apache.hadoop.yarn.util.Records;
 
 /**
  * Contains all the state data that needs to be stored persistently 
@@ -32,19 +35,43 @@ import org.apache.hadoop.yarn.server.res
  */
 @Public
 @Unstable
-public interface ApplicationStateData {
-  
+public abstract class ApplicationStateData {
+  public static ApplicationStateData newInstance(long submitTime,
+      long startTime, String user,
+      ApplicationSubmissionContext submissionContext,
+      RMAppState state, String diagnostics, long finishTime) {
+    ApplicationStateData appState = Records.newRecord(ApplicationStateData.class);
+    appState.setSubmitTime(submitTime);
+    appState.setStartTime(startTime);
+    appState.setUser(user);
+    appState.setApplicationSubmissionContext(submissionContext);
+    appState.setState(state);
+    appState.setDiagnostics(diagnostics);
+    appState.setFinishTime(finishTime);
+    return appState;
+  }
+
+  public static ApplicationStateData newInstance(
+      ApplicationState appState) {
+    return newInstance(appState.getSubmitTime(), appState.getStartTime(),
+        appState.getUser(), appState.getApplicationSubmissionContext(),
+        appState.getState(), appState.getDiagnostics(),
+        appState.getFinishTime());
+  }
+
+  public abstract ApplicationStateDataProto getProto();
+
   /**
    * The time at which the application was received by the Resource Manager
    * @return submitTime
    */
   @Public
   @Unstable
-  public long getSubmitTime();
+  public abstract long getSubmitTime();
   
   @Public
   @Unstable
-  public void setSubmitTime(long submitTime);
+  public abstract void setSubmitTime(long submitTime);
 
   /**
    * Get the <em>start time</em> of the application.
@@ -63,11 +90,11 @@ public interface ApplicationStateData {
    */
   @Public
   @Unstable
-  public void setUser(String user);
+  public abstract void setUser(String user);
   
   @Public
   @Unstable
-  public String getUser();
+  public abstract String getUser();
   
   /**
    * The {@link ApplicationSubmissionContext} for the application
@@ -76,34 +103,34 @@ public interface ApplicationStateData {
    */
   @Public
   @Unstable
-  public ApplicationSubmissionContext getApplicationSubmissionContext();
+  public abstract ApplicationSubmissionContext getApplicationSubmissionContext();
   
   @Public
   @Unstable
-  public void setApplicationSubmissionContext(
+  public abstract void setApplicationSubmissionContext(
       ApplicationSubmissionContext context);
 
   /**
    * Get the final state of the application.
    * @return the final state of the application.
    */
-  public RMAppState getState();
+  public abstract RMAppState getState();
 
-  public void setState(RMAppState state);
+  public abstract void setState(RMAppState state);
 
   /**
    * Get the diagnostics information for the application master.
    * @return the diagnostics information for the application master.
    */
-  public String getDiagnostics();
+  public abstract String getDiagnostics();
 
-  public void setDiagnostics(String diagnostics);
+  public abstract void setDiagnostics(String diagnostics);
 
   /**
    * The finish time of the application.
    * @return the finish time of the application.,
    */
-  public long getFinishTime();
+  public abstract long getFinishTime();
 
-  public void setFinishTime(long finishTime);
+  public abstract void setFinishTime(long finishTime);
 }

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/impl/pb/ApplicationAttemptStateDataPBImpl.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/impl/pb/ApplicationAttemptStateDataPBImpl.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/impl/pb/ApplicationAttemptStateDataPBImpl.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/impl/pb/ApplicationAttemptStateDataPBImpl.java Wed Jun 11 12:20:48 2014
@@ -25,10 +25,7 @@ import org.apache.hadoop.yarn.api.record
 import org.apache.hadoop.yarn.api.records.FinalApplicationStatus;
 import org.apache.hadoop.yarn.api.records.impl.pb.ApplicationAttemptIdPBImpl;
 import org.apache.hadoop.yarn.api.records.impl.pb.ContainerPBImpl;
-import org.apache.hadoop.yarn.api.records.impl.pb.ProtoBase;
 import org.apache.hadoop.yarn.api.records.impl.pb.ProtoUtils;
-import org.apache.hadoop.yarn.factories.RecordFactory;
-import org.apache.hadoop.yarn.factory.providers.RecordFactoryProvider;
 import org.apache.hadoop.yarn.proto.YarnProtos.FinalApplicationStatusProto;
 import org.apache.hadoop.yarn.proto.YarnServerResourceManagerServiceProtos.ApplicationAttemptStateDataProto;
 import org.apache.hadoop.yarn.proto.YarnServerResourceManagerServiceProtos.ApplicationAttemptStateDataProtoOrBuilder;
@@ -36,12 +33,10 @@ import org.apache.hadoop.yarn.proto.Yarn
 import org.apache.hadoop.yarn.server.resourcemanager.recovery.records.ApplicationAttemptStateData;
 import org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptState;
 
-public class ApplicationAttemptStateDataPBImpl
-extends ProtoBase<ApplicationAttemptStateDataProto> 
-implements ApplicationAttemptStateData {
-  private static final RecordFactory recordFactory = RecordFactoryProvider
-      .getRecordFactory(null);
+import com.google.protobuf.TextFormat;
 
+public class ApplicationAttemptStateDataPBImpl extends
+    ApplicationAttemptStateData {
   ApplicationAttemptStateDataProto proto = 
       ApplicationAttemptStateDataProto.getDefaultInstance();
   ApplicationAttemptStateDataProto.Builder builder = null;
@@ -60,7 +55,8 @@ implements ApplicationAttemptStateData {
     this.proto = proto;
     viaProto = true;
   }
-  
+
+  @Override
   public ApplicationAttemptStateDataProto getProto() {
     mergeLocalToProto();
     proto = viaProto ? proto : builder.build();
@@ -76,7 +72,8 @@ implements ApplicationAttemptStateData {
       builder.setMasterContainer(((ContainerPBImpl)masterContainer).getProto());
     }
     if(this.appAttemptTokens != null) {
-      builder.setAppAttemptTokens(convertToProtoFormat(this.appAttemptTokens));
+      builder.setAppAttemptTokens(ProtoUtils.convertToProtoFormat(
+          this.appAttemptTokens));
     }
   }
 
@@ -148,7 +145,8 @@ implements ApplicationAttemptStateData {
     if(!p.hasAppAttemptTokens()) {
       return null;
     }
-    this.appAttemptTokens = convertFromProtoFormat(p.getAppAttemptTokens());
+    this.appAttemptTokens = ProtoUtils.convertFromProtoFormat(
+        p.getAppAttemptTokens());
     return appAttemptTokens;
   }
 
@@ -249,24 +247,26 @@ implements ApplicationAttemptStateData {
     builder.setFinalApplicationStatus(convertToProtoFormat(finishState));
   }
 
-  public static ApplicationAttemptStateData newApplicationAttemptStateData(
-      ApplicationAttemptId attemptId, Container container,
-      ByteBuffer attemptTokens, long startTime, RMAppAttemptState finalState,
-      String finalTrackingUrl, String diagnostics,
-      FinalApplicationStatus amUnregisteredFinalStatus) {
-    ApplicationAttemptStateData attemptStateData =
-        recordFactory.newRecordInstance(ApplicationAttemptStateData.class);
-    attemptStateData.setAttemptId(attemptId);
-    attemptStateData.setMasterContainer(container);
-    attemptStateData.setAppAttemptTokens(attemptTokens);
-    attemptStateData.setState(finalState);
-    attemptStateData.setFinalTrackingUrl(finalTrackingUrl);
-    attemptStateData.setDiagnostics(diagnostics);
-    attemptStateData.setStartTime(startTime);
-    attemptStateData.setFinalApplicationStatus(amUnregisteredFinalStatus);
-    return attemptStateData;
+  @Override
+  public int hashCode() {
+    return getProto().hashCode();
   }
 
+  @Override
+  public boolean equals(Object other) {
+    if (other == null)
+      return false;
+    if (other.getClass().isAssignableFrom(this.getClass())) {
+      return this.getProto().equals(this.getClass().cast(other).getProto());
+    }
+    return false;
+  }
+
+  @Override
+  public String toString() {
+    return TextFormat.shortDebugString(getProto());
+  }
+  
   private static String RM_APP_ATTEMPT_PREFIX = "RMATTEMPT_";
   public static RMAppAttemptStateProto convertToProtoFormat(RMAppAttemptState e) {
     return RMAppAttemptStateProto.valueOf(RM_APP_ATTEMPT_PREFIX + e.name());

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/impl/pb/ApplicationStateDataPBImpl.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/impl/pb/ApplicationStateDataPBImpl.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/impl/pb/ApplicationStateDataPBImpl.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/impl/pb/ApplicationStateDataPBImpl.java Wed Jun 11 12:20:48 2014
@@ -20,21 +20,15 @@ package org.apache.hadoop.yarn.server.re
 
 import org.apache.hadoop.yarn.api.records.ApplicationSubmissionContext;
 import org.apache.hadoop.yarn.api.records.impl.pb.ApplicationSubmissionContextPBImpl;
-import org.apache.hadoop.yarn.api.records.impl.pb.ProtoBase;
-import org.apache.hadoop.yarn.factories.RecordFactory;
-import org.apache.hadoop.yarn.factory.providers.RecordFactoryProvider;
 import org.apache.hadoop.yarn.proto.YarnServerResourceManagerServiceProtos.ApplicationStateDataProto;
 import org.apache.hadoop.yarn.proto.YarnServerResourceManagerServiceProtos.ApplicationStateDataProtoOrBuilder;
 import org.apache.hadoop.yarn.proto.YarnServerResourceManagerServiceProtos.RMAppStateProto;
 import org.apache.hadoop.yarn.server.resourcemanager.recovery.records.ApplicationStateData;
 import org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMAppState;
 
-public class ApplicationStateDataPBImpl 
-extends ProtoBase<ApplicationStateDataProto> 
-implements ApplicationStateData {
-  private static final RecordFactory recordFactory = RecordFactoryProvider
-      .getRecordFactory(null);
+import com.google.protobuf.TextFormat;
 
+public class ApplicationStateDataPBImpl extends ApplicationStateData {
   ApplicationStateDataProto proto = 
             ApplicationStateDataProto.getDefaultInstance();
   ApplicationStateDataProto.Builder builder = null;
@@ -51,7 +45,8 @@ implements ApplicationStateData {
     this.proto = proto;
     viaProto = true;
   }
-  
+
+  @Override
   public ApplicationStateDataProto getProto() {
     mergeLocalToProto();
     proto = viaProto ? proto : builder.build();
@@ -136,7 +131,7 @@ implements ApplicationStateData {
     }
     applicationSubmissionContext = 
         new ApplicationSubmissionContextPBImpl(
-                                          p.getApplicationSubmissionContext());
+            p.getApplicationSubmissionContext());
     return applicationSubmissionContext;
   }
 
@@ -200,21 +195,24 @@ implements ApplicationStateData {
     builder.setFinishTime(finishTime);
   }
 
-  public static ApplicationStateData newApplicationStateData(long submitTime,
-      long startTime, String user,
-      ApplicationSubmissionContext submissionContext, RMAppState state,
-      String diagnostics, long finishTime) {
-
-    ApplicationStateData appState =
-        recordFactory.newRecordInstance(ApplicationStateData.class);
-    appState.setSubmitTime(submitTime);
-    appState.setStartTime(startTime);
-    appState.setUser(user);
-    appState.setApplicationSubmissionContext(submissionContext);
-    appState.setState(state);
-    appState.setDiagnostics(diagnostics);
-    appState.setFinishTime(finishTime);
-    return appState;
+  @Override
+  public int hashCode() {
+    return getProto().hashCode();
+  }
+
+  @Override
+  public boolean equals(Object other) {
+    if (other == null)
+      return false;
+    if (other.getClass().isAssignableFrom(this.getClass())) {
+      return this.getProto().equals(this.getClass().cast(other).getProto());
+    }
+    return false;
+  }
+
+  @Override
+  public String toString() {
+    return TextFormat.shortDebugString(getProto());
   }
 
   private static String RM_APP_PREFIX = "RMAPP_";

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmapp/RMAppImpl.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmapp/RMAppImpl.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmapp/RMAppImpl.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmapp/RMAppImpl.java Wed Jun 11 12:20:48 2014
@@ -71,6 +71,7 @@ import org.apache.hadoop.yarn.server.res
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.YarnScheduler;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.event.AppAddedSchedulerEvent;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.event.AppRemovedSchedulerEvent;
+import org.apache.hadoop.yarn.server.resourcemanager.RMServerUtils;
 import org.apache.hadoop.yarn.server.utils.BuilderUtils;
 import org.apache.hadoop.yarn.state.InvalidStateTransitonException;
 import org.apache.hadoop.yarn.state.MultipleArcTransition;
@@ -293,11 +294,6 @@ public class RMAppImpl implements RMApp,
   private final StateMachine<RMAppState, RMAppEventType, RMAppEvent>
                                                                  stateMachine;
 
-  private static final ApplicationResourceUsageReport
-    DUMMY_APPLICATION_RESOURCE_USAGE_REPORT =
-      BuilderUtils.newApplicationResourceUsageReport(-1, -1,
-          Resources.createResource(-1, -1), Resources.createResource(-1, -1),
-          Resources.createResource(-1, -1));
   private static final int DUMMY_APPLICATION_ATTEMPT_NUMBER = -1;
   
   public RMAppImpl(ApplicationId applicationId, RMContext rmContext,
@@ -498,7 +494,7 @@ public class RMAppImpl implements RMApp,
       String origTrackingUrl = UNAVAILABLE;
       int rpcPort = -1;
       ApplicationResourceUsageReport appUsageReport =
-          DUMMY_APPLICATION_RESOURCE_USAGE_REPORT;
+          RMServerUtils.DUMMY_APPLICATION_RESOURCE_USAGE_REPORT;
       FinalApplicationStatus finishState = getFinalApplicationStatus();
       String diags = UNAVAILABLE;
       float progress = 0.0f;
@@ -723,29 +719,36 @@ public class RMAppImpl implements RMApp,
     }
   }
 
+  // synchronously recover attempt to ensure any incoming external events
+  // to be processed after the attempt processes the recover event.
+  private void recoverAppAttempts() {
+    for (RMAppAttempt attempt : getAppAttempts().values()) {
+      attempt.handle(new RMAppAttemptEvent(attempt.getAppAttemptId(),
+        RMAppAttemptEventType.RECOVER));
+    }
+  }
+
   private static final class RMAppRecoveredTransition implements
       MultipleArcTransition<RMAppImpl, RMAppEvent, RMAppState> {
 
     @Override
     public RMAppState transition(RMAppImpl app, RMAppEvent event) {
 
-      for (RMAppAttempt attempt : app.getAppAttempts().values()) {
-        // synchronously recover attempt to ensure any incoming external events
-        // to be processed after the attempt processes the recover event.
-        attempt.handle(
-          new RMAppAttemptEvent(attempt.getAppAttemptId(),
-            RMAppAttemptEventType.RECOVER));
-      }
-
       // The app has completed.
       if (app.recoveredFinalState != null) {
+        app.recoverAppAttempts();
         new FinalTransition(app.recoveredFinalState).transition(app, event);
         return app.recoveredFinalState;
       }
 
-      // Last attempt is in final state, do not add to scheduler and just return
-      // ACCEPTED waiting for last RMAppAttempt to send finished or failed event
-      // back.
+      // Notify scheduler about the app on recovery
+      new AddApplicationToSchedulerTransition().transition(app, event);
+
+      // recover attempts
+      app.recoverAppAttempts();
+
+      // Last attempt is in final state, return ACCEPTED waiting for last
+      // RMAppAttempt to send finished or failed event back.
       if (app.currentAttempt != null
           && (app.currentAttempt.getState() == RMAppAttemptState.KILLED
               || app.currentAttempt.getState() == RMAppAttemptState.FINISHED
@@ -754,9 +757,6 @@ public class RMAppImpl implements RMApp,
         return RMAppState.ACCEPTED;
       }
 
-      // Notify scheduler about the app on recovery
-      new AddApplicationToSchedulerTransition().transition(app, event);
-
       // No existent attempts means the attempt associated with this app was not
       // started or started but not yet saved.
       if (app.attempts.isEmpty()) {
@@ -1055,8 +1055,12 @@ public class RMAppImpl implements RMApp,
       if (app.finishTime == 0 ) {
         app.finishTime = System.currentTimeMillis();
       }
-      app.handler.handle(new AppRemovedSchedulerEvent(app.applicationId,
-        finalState));
+      // Recovered apps that are completed were not added to scheduler, so no
+      // need to remove them from scheduler.
+      if (app.recoveredFinalState == null) {
+        app.handler.handle(new AppRemovedSchedulerEvent(app.applicationId,
+          finalState));
+      }
       app.handler.handle(
           new RMAppManagerEvent(app.applicationId,
           RMAppManagerEventType.APP_COMPLETED));

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmapp/attempt/RMAppAttemptImpl.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmapp/attempt/RMAppAttemptImpl.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmapp/attempt/RMAppAttemptImpl.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmapp/attempt/RMAppAttemptImpl.java Wed Jun 11 12:20:48 2014
@@ -267,15 +267,17 @@ public class RMAppAttemptImpl implements
       .addTransition(RMAppAttemptState.ALLOCATED, RMAppAttemptState.FINAL_SAVING,
           RMAppAttemptEventType.CONTAINER_FINISHED,
           new FinalSavingTransition(
-            new AMContainerCrashedTransition(), RMAppAttemptState.FAILED))
+            new AMContainerCrashedBeforeRunningTransition(), RMAppAttemptState.FAILED))
 
        // Transitions from LAUNCHED State
       .addTransition(RMAppAttemptState.LAUNCHED, RMAppAttemptState.RUNNING,
           RMAppAttemptEventType.REGISTERED, new AMRegisteredTransition())
-      .addTransition(RMAppAttemptState.LAUNCHED, RMAppAttemptState.FINAL_SAVING,
+      .addTransition(RMAppAttemptState.LAUNCHED,
+          EnumSet.of(RMAppAttemptState.LAUNCHED, RMAppAttemptState.FINAL_SAVING),
           RMAppAttemptEventType.CONTAINER_FINISHED,
-          new FinalSavingTransition(
-            new AMContainerCrashedTransition(), RMAppAttemptState.FAILED))
+          new ContainerFinishedTransition(
+            new AMContainerCrashedBeforeRunningTransition(),
+            RMAppAttemptState.LAUNCHED))
       .addTransition(
           RMAppAttemptState.LAUNCHED, RMAppAttemptState.FINAL_SAVING,
           RMAppAttemptEventType.EXPIRE,
@@ -302,7 +304,9 @@ public class RMAppAttemptImpl implements
           RMAppAttemptState.RUNNING,
           EnumSet.of(RMAppAttemptState.RUNNING, RMAppAttemptState.FINAL_SAVING),
           RMAppAttemptEventType.CONTAINER_FINISHED,
-          new ContainerFinishedTransition())
+          new ContainerFinishedTransition(
+            new AMContainerCrashedAtRunningTransition(),
+            RMAppAttemptState.RUNNING))
       .addTransition(
           RMAppAttemptState.RUNNING, RMAppAttemptState.FINAL_SAVING,
           RMAppAttemptEventType.EXPIRE,
@@ -671,9 +675,7 @@ public class RMAppAttemptImpl implements
       ApplicationResourceUsageReport report =
           scheduler.getAppResourceUsageReport(this.getAppAttemptId());
       if (report == null) {
-        Resource none = Resource.newInstance(0, 0);
-        report = ApplicationResourceUsageReport.newInstance(0, 0, none, none,
-            none);
+        report = RMServerUtils.DUMMY_APPLICATION_RESOURCE_USAGE_REPORT;
       }
       return report;
     } finally {
@@ -904,6 +906,12 @@ public class RMAppAttemptImpl implements
         }
         return appAttempt.recoveredFinalState;
       } else {
+        // Add the current attempt to the scheduler.
+        if (appAttempt.rmContext.isWorkPreservingRecoveryEnabled()) {
+          appAttempt.eventHandler.handle(new AppAttemptAddedSchedulerEvent(
+            appAttempt.getAppAttemptId(), false));
+        }
+
         /*
          * Since the application attempt's final state is not saved that means
          * for AM container (previous attempt) state must be one of these.
@@ -1207,17 +1215,16 @@ public class RMAppAttemptImpl implements
     }
   }
 
-  private static final class AMContainerCrashedTransition extends
+  private static final class AMContainerCrashedBeforeRunningTransition extends
       BaseFinalTransition {
 
-    public AMContainerCrashedTransition() {
+    public AMContainerCrashedBeforeRunningTransition() {
       super(RMAppAttemptState.FAILED);
     }
 
     @Override
     public void transition(RMAppAttemptImpl appAttempt,
         RMAppAttemptEvent event) {
-
       RMAppAttemptContainerFinishedEvent finishEvent =
           ((RMAppAttemptContainerFinishedEvent)event);
 
@@ -1410,6 +1417,16 @@ public class RMAppAttemptImpl implements
       implements
       MultipleArcTransition<RMAppAttemptImpl, RMAppAttemptEvent, RMAppAttemptState> {
 
+    // The transition To Do after attempt final state is saved.
+    private BaseTransition transitionToDo;
+    private RMAppAttemptState currentState;
+
+    public ContainerFinishedTransition(BaseTransition transitionToDo,
+        RMAppAttemptState currentState) {
+      this.transitionToDo = transitionToDo;
+      this.currentState = currentState;
+    }
+
     @Override
     public RMAppAttemptState transition(RMAppAttemptImpl appAttempt,
         RMAppAttemptEvent event) {
@@ -1426,14 +1443,13 @@ public class RMAppAttemptImpl implements
             containerStatus.getContainerId())) {
         // Remember the follow up transition and save the final attempt state.
         appAttempt.rememberTargetTransitionsAndStoreState(event,
-          new ContainerFinishedFinalStateSavedTransition(),
-          RMAppAttemptState.FAILED, RMAppAttemptState.FAILED);
+          transitionToDo, RMAppAttemptState.FAILED, RMAppAttemptState.FAILED);
         return RMAppAttemptState.FINAL_SAVING;
       }
 
       // Normal container.Put it in completedcontainers list
       appAttempt.justFinishedContainers.add(containerStatus);
-      return RMAppAttemptState.RUNNING;
+      return this.currentState;
     }
   }
 
@@ -1451,7 +1467,7 @@ public class RMAppAttemptImpl implements
     }
   }
 
-  private static class ContainerFinishedFinalStateSavedTransition extends
+  private static class AMContainerCrashedAtRunningTransition extends
       BaseTransition {
     @Override
     public void

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmcontainer/RMContainerEventType.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmcontainer/RMContainerEventType.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmcontainer/RMContainerEventType.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmcontainer/RMContainerEventType.java Wed Jun 11 12:20:48 2014
@@ -33,5 +33,7 @@ public enum RMContainerEventType {
   RELEASED,
 
   // Source: ContainerAllocationExpirer  
-  EXPIRE
+  EXPIRE,
+
+  RECOVER
 }

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmcontainer/RMContainerImpl.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmcontainer/RMContainerImpl.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmcontainer/RMContainerImpl.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmcontainer/RMContainerImpl.java Wed Jun 11 12:20:48 2014
@@ -35,12 +35,14 @@ import org.apache.hadoop.yarn.api.record
 import org.apache.hadoop.yarn.api.records.Priority;
 import org.apache.hadoop.yarn.api.records.Resource;
 import org.apache.hadoop.yarn.event.EventHandler;
+import org.apache.hadoop.yarn.server.api.protocolrecords.NMContainerStatus;
 import org.apache.hadoop.yarn.server.resourcemanager.RMContext;
 import org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.event.RMAppAttemptContainerAcquiredEvent;
 import org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.event.RMAppAttemptContainerAllocatedEvent;
 import org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.event.RMAppAttemptContainerFinishedEvent;
 import org.apache.hadoop.yarn.server.resourcemanager.rmnode.RMNodeCleanContainerEvent;
 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;
@@ -65,6 +67,9 @@ public class RMContainerImpl implements 
         RMContainerEventType.KILL)
     .addTransition(RMContainerState.NEW, RMContainerState.RESERVED,
         RMContainerEventType.RESERVED, new ContainerReservedTransition())
+    .addTransition(RMContainerState.NEW,
+        EnumSet.of(RMContainerState.RUNNING, RMContainerState.COMPLETED),
+        RMContainerEventType.RECOVER, new ContainerRecoveredTransition())
 
     // Transitions from RESERVED state
     .addTransition(RMContainerState.RESERVED, RMContainerState.RESERVED, 
@@ -341,6 +346,38 @@ public class RMContainerImpl implements 
     }
   }
 
+  private static final class ContainerRecoveredTransition
+      implements
+      MultipleArcTransition<RMContainerImpl, RMContainerEvent, RMContainerState> {
+    @Override
+    public RMContainerState transition(RMContainerImpl container,
+        RMContainerEvent event) {
+      NMContainerStatus report =
+          ((RMContainerRecoverEvent) event).getContainerReport();
+      if (report.getContainerState().equals(ContainerState.COMPLETE)) {
+        ContainerStatus status =
+            ContainerStatus.newInstance(report.getContainerId(),
+              report.getContainerState(), report.getDiagnostics(),
+              report.getContainerExitStatus());
+
+        new FinishedTransition().transition(container,
+          new RMContainerFinishedEvent(container.containerId, status,
+            RMContainerEventType.FINISHED));
+        return RMContainerState.COMPLETED;
+      } else if (report.getContainerState().equals(ContainerState.RUNNING)) {
+        // Tell the appAttempt
+        container.eventHandler.handle(new RMAppAttemptContainerAcquiredEvent(
+            container.getApplicationAttemptId(), container.getContainer()));
+        return RMContainerState.RUNNING;
+      } else {
+        // This can never happen.
+        LOG.warn("RMContainer received unexpected recover event with container"
+            + " state " + report.getContainerState() + " while recovering.");
+        return RMContainerState.RUNNING;
+      }
+    }
+  }
+
   private static final class ContainerReservedTransition extends
   BaseTransition {
 
@@ -398,7 +435,6 @@ public class RMContainerImpl implements 
       // Inform AppAttempt
       container.eventHandler.handle(new RMAppAttemptContainerFinishedEvent(
           container.appAttemptId, finishedEvent.getRemoteContainerStatus()));
-
       container.rmContext.getRMApplicationHistoryWriter()
           .containerFinished(container);
     }

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmnode/RMNodeImpl.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmnode/RMNodeImpl.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmnode/RMNodeImpl.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmnode/RMNodeImpl.java Wed Jun 11 12:20:48 2014
@@ -48,6 +48,7 @@ import org.apache.hadoop.yarn.api.record
 import org.apache.hadoop.yarn.event.EventHandler;
 import org.apache.hadoop.yarn.factories.RecordFactory;
 import org.apache.hadoop.yarn.factory.providers.RecordFactoryProvider;
+import org.apache.hadoop.yarn.server.api.protocolrecords.NMContainerStatus;
 import org.apache.hadoop.yarn.server.api.protocolrecords.NodeHeartbeatResponse;
 import org.apache.hadoop.yarn.server.api.records.NodeHealthStatus;
 import org.apache.hadoop.yarn.server.resourcemanager.ClusterMetrics;
@@ -460,13 +461,9 @@ public class RMNodeImpl implements RMNod
     @Override
     public void transition(RMNodeImpl rmNode, RMNodeEvent event) {
       // Inform the scheduler
+      RMNodeStartedEvent startEvent = (RMNodeStartedEvent) event;
+      List<NMContainerStatus> containers = null;
 
-      rmNode.context.getDispatcher().getEventHandler().handle(
-          new NodeAddedSchedulerEvent(rmNode));
-      rmNode.context.getDispatcher().getEventHandler().handle(
-          new NodesListManagerEvent(
-              NodesListManagerEventType.NODE_USABLE, rmNode));
- 
       String host = rmNode.nodeId.getHost();
       if (rmNode.context.getInactiveRMNodes().containsKey(host)) {
         // Old node rejoining
@@ -476,10 +473,17 @@ public class RMNodeImpl implements RMNod
       } else {
         // Increment activeNodes explicitly because this is a new node.
         ClusterMetrics.getMetrics().incrNumActiveNodes();
+        containers = startEvent.getContainerRecoveryReports();
       }
+
+      rmNode.context.getDispatcher().getEventHandler()
+        .handle(new NodeAddedSchedulerEvent(rmNode, containers));
+      rmNode.context.getDispatcher().getEventHandler().handle(
+        new NodesListManagerEvent(
+            NodesListManagerEventType.NODE_USABLE, rmNode));
     }
   }
-  
+
   public static class ReconnectNodeTransition implements
       SingleArcTransition<RMNodeImpl, RMNodeEvent> {
 
@@ -513,7 +517,7 @@ public class RMNodeImpl implements RMNod
         }
         rmNode.context.getRMNodes().put(newNode.getNodeID(), newNode);
         rmNode.context.getDispatcher().getEventHandler().handle(
-            new RMNodeEvent(newNode.getNodeID(), RMNodeEventType.STARTED));
+            new RMNodeStartedEvent(newNode.getNodeID(), null));
       }
       rmNode.context.getDispatcher().getEventHandler().handle(
           new NodesListManagerEvent(

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/AbstractYarnScheduler.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/AbstractYarnScheduler.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/AbstractYarnScheduler.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/AbstractYarnScheduler.java Wed Jun 11 12:20:48 2014
@@ -32,14 +32,21 @@ import org.apache.hadoop.yarn.api.record
 import org.apache.hadoop.yarn.api.records.ApplicationResourceUsageReport;
 import org.apache.hadoop.yarn.api.records.Container;
 import org.apache.hadoop.yarn.api.records.ContainerId;
+import org.apache.hadoop.yarn.api.records.ContainerState;
 import org.apache.hadoop.yarn.api.records.NodeId;
 import org.apache.hadoop.yarn.api.records.Resource;
 import org.apache.hadoop.yarn.exceptions.YarnException;
+import org.apache.hadoop.yarn.server.api.protocolrecords.NMContainerStatus;
 import org.apache.hadoop.yarn.server.resourcemanager.RMContext;
 import org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMApp;
 import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer;
+import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainerImpl;
+import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainerRecoverEvent;
+import org.apache.hadoop.yarn.server.resourcemanager.rmnode.RMNode;
+import org.apache.hadoop.yarn.server.resourcemanager.rmnode.RMNodeCleanContainerEvent;
 import org.apache.hadoop.yarn.util.resource.Resources;
 
+@SuppressWarnings("unchecked")
 public abstract class AbstractYarnScheduler
     <T extends SchedulerApplicationAttempt, N extends SchedulerNode>
     extends AbstractService implements ResourceScheduler {
@@ -47,8 +54,7 @@ public abstract class AbstractYarnSchedu
   private static final Log LOG = LogFactory.getLog(AbstractYarnScheduler.class);
 
   // Nodes in the cluster, indexed by NodeId
-  protected Map<NodeId, N> nodes =
-      new ConcurrentHashMap<NodeId, N>();
+  protected Map<NodeId, N> nodes = new ConcurrentHashMap<NodeId, N>();
 
   // Whole capacity of the cluster
   protected Resource clusterResource = Resource.newInstance(0, 0);
@@ -58,6 +64,7 @@ public abstract class AbstractYarnSchedu
 
   protected RMContext rmContext;
   protected Map<ApplicationId, SchedulerApplication<T>> applications;
+
   protected final static List<Container> EMPTY_CONTAINER_LIST =
       new ArrayList<Container>();
   protected static final Allocation EMPTY_ALLOCATION = new Allocation(
@@ -169,4 +176,90 @@ public abstract class AbstractYarnSchedu
     throw new YarnException(getClass().getSimpleName()
         + " does not support moving apps between queues");
   }
+
+  private void killOrphanContainerOnNode(RMNode node,
+      NMContainerStatus container) {
+    if (!container.getContainerState().equals(ContainerState.COMPLETE)) {
+      this.rmContext.getDispatcher().getEventHandler().handle(
+        new RMNodeCleanContainerEvent(node.getNodeID(),
+          container.getContainerId()));
+    }
+  }
+
+  public synchronized void recoverContainersOnNode(
+      List<NMContainerStatus> containerReports, RMNode nm) {
+    if (!rmContext.isWorkPreservingRecoveryEnabled()
+        || containerReports == null
+        || (containerReports != null && containerReports.isEmpty())) {
+      return;
+    }
+
+    for (NMContainerStatus container : containerReports) {
+      ApplicationId appId =
+          container.getContainerId().getApplicationAttemptId().getApplicationId();
+      RMApp rmApp = rmContext.getRMApps().get(appId);
+      if (rmApp == null) {
+        LOG.error("Skip recovering container " + container
+            + " for unknown application.");
+        killOrphanContainerOnNode(nm, container);
+        continue;
+      }
+
+      // Unmanaged AM recovery is addressed in YARN-1815
+      if (rmApp.getApplicationSubmissionContext().getUnmanagedAM()) {
+        LOG.info("Skip recovering container " + container + " for unmanaged AM."
+            + rmApp.getApplicationId());
+        killOrphanContainerOnNode(nm, container);
+        continue;
+      }
+
+      SchedulerApplication<T> schedulerApp = applications.get(appId);
+      if (schedulerApp == null) {
+        LOG.info("Skip recovering container  " + container
+            + " for unknown SchedulerApplication. Application current state is "
+            + rmApp.getState());
+        killOrphanContainerOnNode(nm, container);
+        continue;
+      }
+
+      LOG.info("Recovering container " + container);
+      SchedulerApplicationAttempt schedulerAttempt =
+          schedulerApp.getCurrentAppAttempt();
+
+      // create container
+      RMContainer rmContainer = recoverAndCreateContainer(container, nm);
+
+      // recover RMContainer
+      rmContainer.handle(new RMContainerRecoverEvent(container.getContainerId(),
+        container));
+
+      // recover scheduler node
+      nodes.get(nm.getNodeID()).recoverContainer(rmContainer);
+
+      // recover queue: update headroom etc.
+      Queue queue = schedulerAttempt.getQueue();
+      queue.recoverContainer(clusterResource, schedulerAttempt, rmContainer);
+
+      // recover scheduler attempt
+      schedulerAttempt.recoverContainer(rmContainer);
+    }
+  }
+
+  private RMContainer recoverAndCreateContainer(NMContainerStatus report,
+      RMNode node) {
+    Container container =
+        Container.newInstance(report.getContainerId(), node.getNodeID(),
+          node.getHttpAddress(), report.getAllocatedResource(),
+          report.getPriority(), null);
+    ApplicationAttemptId attemptId =
+        container.getId().getApplicationAttemptId();
+    RMContainer rmContainer =
+        new RMContainerImpl(container, attemptId, node.getNodeID(),
+          applications.get(attemptId.getApplicationId()).getUser(), rmContext);
+    return rmContainer;
+  }
+
+  public SchedulerNode getSchedulerNode(NodeId nodeId) {
+    return nodes.get(nodeId);
+  }
 }

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/AppSchedulingInfo.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/AppSchedulingInfo.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/AppSchedulingInfo.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/AppSchedulingInfo.java Wed Jun 11 12:20:48 2014
@@ -39,6 +39,8 @@ import org.apache.hadoop.yarn.api.record
 import org.apache.hadoop.yarn.api.records.Resource;
 import org.apache.hadoop.yarn.api.records.ResourceRequest;
 import org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptState;
+import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer;
+import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainerState;
 import org.apache.hadoop.yarn.util.resource.Resources;
 
 /**
@@ -409,4 +411,25 @@ public class AppSchedulingInfo {
     //    this.requests = appInfo.getRequests();
     this.blacklist = appInfo.getBlackList();
   }
+
+  public synchronized void recoverContainer(RMContainer rmContainer) {
+    // ContainerIdCounter on recovery will be addressed in YARN-2052
+    this.containerIdCounter.incrementAndGet();
+
+    QueueMetrics metrics = queue.getMetrics();
+    if (pending) {
+      // If there was any container to recover, the application was
+      // running from scheduler's POV.
+      pending = false;
+      metrics.runAppAttempt(applicationId, user);
+    }
+
+    // Container is completed. Skip recovering resources.
+    if (rmContainer.getState().equals(RMContainerState.COMPLETED)) {
+      return;
+    }
+
+    metrics.allocateResources(user, 1, rmContainer.getAllocatedResource(),
+      false);
+  }
 }

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/Queue.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/Queue.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/Queue.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/Queue.java Wed Jun 11 12:20:48 2014
@@ -26,6 +26,8 @@ import org.apache.hadoop.security.UserGr
 import org.apache.hadoop.yarn.api.records.QueueACL;
 import org.apache.hadoop.yarn.api.records.QueueInfo;
 import org.apache.hadoop.yarn.api.records.QueueUserACLInfo;
+import org.apache.hadoop.yarn.api.records.Resource;
+import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer;
 
 @Evolving
 @LimitedPrivate("yarn")
@@ -60,4 +62,13 @@ public interface Queue {
   boolean hasAccess(QueueACL acl, UserGroupInformation user);
   
   public ActiveUsersManager getActiveUsersManager();
+
+  /**
+   * Recover the state of the queue for a given container.
+   * @param clusterResource the resource of the cluster
+   * @param schedulerAttempt the application for which the container was allocated
+   * @param rmContainer the container that was recovered.
+   */
+  public void recoverContainer(Resource clusterResource,
+      SchedulerApplicationAttempt schedulerAttempt, RMContainer rmContainer);
 }

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/SchedulerApplicationAttempt.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/SchedulerApplicationAttempt.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/SchedulerApplicationAttempt.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/SchedulerApplicationAttempt.java Wed Jun 11 12:20:48 2014
@@ -47,6 +47,7 @@ import org.apache.hadoop.yarn.server.res
 import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainerEventType;
 import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainerImpl;
 import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainerReservedEvent;
+import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainerState;
 import org.apache.hadoop.yarn.server.resourcemanager.rmnode.RMNodeCleanContainerEvent;
 import org.apache.hadoop.yarn.util.resource.Resources;
 
@@ -534,5 +535,24 @@ public class SchedulerApplicationAttempt
 
     appSchedulingInfo.move(newQueue);
     this.queue = newQueue;
-  }  
+  }
+
+  public synchronized void recoverContainer(RMContainer rmContainer) {
+    // recover app scheduling info
+    appSchedulingInfo.recoverContainer(rmContainer);
+
+    if (rmContainer.getState().equals(RMContainerState.COMPLETED)) {
+      return;
+    }
+    LOG.info("SchedulerAttempt " + getApplicationAttemptId()
+      + " is recovering container " + rmContainer.getContainerId());
+    liveContainers.put(rmContainer.getContainerId(), rmContainer);
+    Resources.addTo(currentConsumption, rmContainer.getContainer()
+      .getResource());
+    // resourceLimit: updated when LeafQueue#recoverContainer#allocateResource
+    // is called.
+    // newlyAllocatedContainers.add(rmContainer);
+    // schedulingOpportunities
+    // lastScheduledContainer
+  }
 }

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/SchedulerNode.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/SchedulerNode.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/SchedulerNode.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/SchedulerNode.java Wed Jun 11 12:20:48 2014
@@ -27,7 +27,6 @@ import org.apache.commons.logging.Log;
 import org.apache.commons.logging.LogFactory;
 import org.apache.hadoop.classification.InterfaceAudience.Private;
 import org.apache.hadoop.classification.InterfaceStability.Unstable;
-import org.apache.hadoop.yarn.api.records.ApplicationId;
 import org.apache.hadoop.yarn.api.records.Container;
 import org.apache.hadoop.yarn.api.records.ContainerId;
 import org.apache.hadoop.yarn.api.records.NodeId;
@@ -35,10 +34,10 @@ import org.apache.hadoop.yarn.api.record
 import org.apache.hadoop.yarn.api.records.Resource;
 import org.apache.hadoop.yarn.conf.YarnConfiguration;
 import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer;
+import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainerState;
 import org.apache.hadoop.yarn.server.resourcemanager.rmnode.RMNode;
 import org.apache.hadoop.yarn.util.resource.Resources;
 
-import com.google.common.base.Preconditions;
 
 /**
  * Represents a YARN Cluster Node from the viewpoint of the scheduler.
@@ -119,13 +118,10 @@ public abstract class SchedulerNode {
    * The Scheduler has allocated containers on this node to the given
    * application.
    * 
-   * @param applicationId
-   *          application
    * @param rmContainer
    *          allocated container
    */
-  public synchronized void allocateContainer(ApplicationId applicationId,
-      RMContainer rmContainer) {
+  public synchronized void allocateContainer(RMContainer rmContainer) {
     Container container = rmContainer.getContainer();
     deductAvailableResource(container.getResource());
     ++numContainers;
@@ -166,8 +162,8 @@ public abstract class SchedulerNode {
     return this.totalResourceCapability;
   }
 
-  private synchronized boolean isValidContainer(Container c) {
-    if (launchedContainers.containsKey(c.getId())) {
+  public synchronized boolean isValidContainer(ContainerId containerId) {
+    if (launchedContainers.containsKey(containerId)) {
       return true;
     }
     return false;
@@ -185,7 +181,7 @@ public abstract class SchedulerNode {
    *          container to be released
    */
   public synchronized void releaseContainer(Container container) {
-    if (!isValidContainer(container)) {
+    if (!isValidContainer(container.getId())) {
       LOG.error("Invalid container released " + container);
       return;
     }
@@ -274,4 +270,12 @@ public abstract class SchedulerNode {
     // we can only adjust available resource if total resource is changed.
     Resources.addTo(this.availableResource, deltaResource);
   }
+
+
+  public synchronized void recoverContainer(RMContainer rmContainer) {
+    if (rmContainer.getState().equals(RMContainerState.COMPLETED)) {
+      return;
+    }
+    allocateContainer(rmContainer);
+  }
 }

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/CSQueue.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/CSQueue.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/CSQueue.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/CSQueue.java Wed Jun 11 12:20:48 2014
@@ -28,7 +28,6 @@ import org.apache.hadoop.security.Access
 import org.apache.hadoop.security.UserGroupInformation;
 import org.apache.hadoop.yarn.api.records.ApplicationAttemptId;
 import org.apache.hadoop.yarn.api.records.ApplicationId;
-import org.apache.hadoop.yarn.api.records.Container;
 import org.apache.hadoop.yarn.api.records.ContainerStatus;
 import org.apache.hadoop.yarn.api.records.QueueACL;
 import org.apache.hadoop.yarn.api.records.QueueState;
@@ -235,15 +234,6 @@ extends org.apache.hadoop.yarn.server.re
   public ActiveUsersManager getActiveUsersManager();
   
   /**
-   * Recover the state of the queue
-   * @param clusterResource the resource of the cluster
-   * @param application the application for which the container was allocated
-   * @param container the container that was recovered.
-   */
-  public void recoverContainer(Resource clusterResource, FiCaSchedulerApp application, 
-      Container container);
-  
-  /**
    * Adds all applications in the queue and its subqueues to the given collection.
    * @param apps the collection to add the applications to
    */

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/CapacityScheduler.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/CapacityScheduler.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/CapacityScheduler.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/CapacityScheduler.java Wed Jun 11 12:20:48 2014
@@ -872,6 +872,8 @@ public class CapacityScheduler extends
     {
       NodeAddedSchedulerEvent nodeAddedEvent = (NodeAddedSchedulerEvent)event;
       addNode(nodeAddedEvent.getAddedRMNode());
+      recoverContainersOnNode(nodeAddedEvent.getContainerReports(),
+        nodeAddedEvent.getAddedRMNode());
     }
     break;
     case NODE_REMOVED:

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/LeafQueue.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/LeafQueue.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/LeafQueue.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/LeafQueue.java Wed Jun 11 12:20:48 2014
@@ -59,15 +59,17 @@ import org.apache.hadoop.yarn.server.res
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.NodeType;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.QueueMetrics;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.SchedulerAppUtils;
+import org.apache.hadoop.yarn.server.resourcemanager.scheduler.SchedulerApplicationAttempt;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.common.fica.FiCaSchedulerApp;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.common.fica.FiCaSchedulerNode;
-import org.apache.hadoop.yarn.server.resourcemanager.security.RMContainerTokenSecretManager;
 import org.apache.hadoop.yarn.server.utils.BuilderUtils;
 import org.apache.hadoop.yarn.server.utils.Lock;
 import org.apache.hadoop.yarn.server.utils.Lock.NoLock;
 import org.apache.hadoop.yarn.util.resource.ResourceCalculator;
 import org.apache.hadoop.yarn.util.resource.Resources;
 
+import com.google.common.annotations.VisibleForTesting;
+
 @Private
 @Unstable
 public class LeafQueue implements CSQueue {
@@ -564,7 +566,8 @@ public class LeafQueue implements CSQueu
         "numContainers=" + getNumContainers();  
   }
 
-  private synchronized User getUser(String userName) {
+  @VisibleForTesting
+  public synchronized User getUser(String userName) {
     User user = users.get(userName);
     if (user == null) {
       user = new User();
@@ -1346,8 +1349,7 @@ public class LeafQueue implements CSQueu
       }
 
       // Inform the node
-      node.allocateContainer(application.getApplicationId(), 
-          allocatedContainer);
+      node.allocateContainer(allocatedContainer);
 
       LOG.info("assignedContainer" +
           " application attempt=" + application.getApplicationAttemptId() +
@@ -1446,7 +1448,7 @@ public class LeafQueue implements CSQueu
   }
 
   synchronized void allocateResource(Resource clusterResource, 
-      FiCaSchedulerApp application, Resource resource) {
+      SchedulerApplicationAttempt application, Resource resource) {
     // Update queue metrics
     Resources.addTo(usedResources, resource);
     CSQueueUtils.updateQueueStatistics(
@@ -1530,7 +1532,8 @@ public class LeafQueue implements CSQueu
     return metrics;
   }
 
-  static class User {
+  @VisibleForTesting
+  public static class User {
     Resource consumed = Resources.createResource(0, 0);
     int pendingApplications = 0;
     int activeApplications = 0;
@@ -1580,13 +1583,16 @@ public class LeafQueue implements CSQueu
 
   @Override
   public void recoverContainer(Resource clusterResource,
-      FiCaSchedulerApp application, Container container) {
+      SchedulerApplicationAttempt attempt, RMContainer rmContainer) {
+    if (rmContainer.getState().equals(RMContainerState.COMPLETED)) {
+      return;
+    }
     // Careful! Locking order is important! 
     synchronized (this) {
-      allocateResource(clusterResource, application, container.getResource());
+      allocateResource(clusterResource, attempt, rmContainer.getContainer()
+        .getResource());
     }
-    getParent().recoverContainer(clusterResource, application, container);
-
+    getParent().recoverContainer(clusterResource, attempt, rmContainer);
   }
 
   /**
@@ -1613,5 +1619,4 @@ public class LeafQueue implements CSQueu
       apps.add(app.getApplicationAttemptId());
     }
   }
-
 }

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/ParentQueue.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/ParentQueue.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/ParentQueue.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/ParentQueue.java Wed Jun 11 12:20:48 2014
@@ -38,7 +38,6 @@ import org.apache.hadoop.security.UserGr
 import org.apache.hadoop.security.authorize.AccessControlList;
 import org.apache.hadoop.yarn.api.records.ApplicationAttemptId;
 import org.apache.hadoop.yarn.api.records.ApplicationId;
-import org.apache.hadoop.yarn.api.records.Container;
 import org.apache.hadoop.yarn.api.records.ContainerStatus;
 import org.apache.hadoop.yarn.api.records.QueueACL;
 import org.apache.hadoop.yarn.api.records.QueueInfo;
@@ -49,9 +48,11 @@ import org.apache.hadoop.yarn.factories.
 import org.apache.hadoop.yarn.factory.providers.RecordFactoryProvider;
 import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer;
 import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainerEventType;
+import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainerState;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.ActiveUsersManager;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.NodeType;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.QueueMetrics;
+import org.apache.hadoop.yarn.server.resourcemanager.scheduler.SchedulerApplicationAttempt;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.common.fica.FiCaSchedulerApp;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.common.fica.FiCaSchedulerNode;
 import org.apache.hadoop.yarn.util.resource.ResourceCalculator;
@@ -770,13 +771,16 @@ public class ParentQueue implements CSQu
   
   @Override
   public void recoverContainer(Resource clusterResource,
-      FiCaSchedulerApp application, Container container) {
+      SchedulerApplicationAttempt attempt, RMContainer rmContainer) {
+    if (rmContainer.getState().equals(RMContainerState.COMPLETED)) {
+      return;
+    }
     // Careful! Locking order is important! 
     synchronized (this) {
-      allocateResource(clusterResource, container.getResource());
+      allocateResource(clusterResource,rmContainer.getContainer().getResource());
     }
     if (parent != null) {
-      parent.recoverContainer(clusterResource, application, container);
+      parent.recoverContainer(clusterResource, attempt, rmContainer);
     }
   }
 

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/common/fica/FiCaSchedulerNode.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/common/fica/FiCaSchedulerNode.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/common/fica/FiCaSchedulerNode.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/common/fica/FiCaSchedulerNode.java Wed Jun 11 12:20:48 2014
@@ -24,6 +24,7 @@ import org.apache.commons.logging.LogFac
 import org.apache.hadoop.yarn.api.records.ApplicationAttemptId;
 import org.apache.hadoop.yarn.api.records.Priority;
 import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer;
+import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainerState;
 import org.apache.hadoop.yarn.server.resourcemanager.rmnode.RMNode;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.SchedulerApplicationAttempt;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.SchedulerNode;

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/event/NodeAddedSchedulerEvent.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/event/NodeAddedSchedulerEvent.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/event/NodeAddedSchedulerEvent.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/event/NodeAddedSchedulerEvent.java Wed Jun 11 12:20:48 2014
@@ -18,19 +18,34 @@
 
 package org.apache.hadoop.yarn.server.resourcemanager.scheduler.event;
 
+import java.util.List;
+
+import org.apache.hadoop.yarn.server.api.protocolrecords.NMContainerStatus;
 import org.apache.hadoop.yarn.server.resourcemanager.rmnode.RMNode;
 
 public class NodeAddedSchedulerEvent extends SchedulerEvent {
 
   private final RMNode rmNode;
+  private final List<NMContainerStatus> containerReports;
 
   public NodeAddedSchedulerEvent(RMNode rmNode) {
     super(SchedulerEventType.NODE_ADDED);
     this.rmNode = rmNode;
+    this.containerReports = null;
+  }
+
+  public NodeAddedSchedulerEvent(RMNode rmNode,
+      List<NMContainerStatus> containerReports) {
+    super(SchedulerEventType.NODE_ADDED);
+    this.rmNode = rmNode;
+    this.containerReports = containerReports;
   }
 
   public RMNode getAddedRMNode() {
     return rmNode;
   }
 
+  public List<NMContainerStatus> getContainerReports() {
+    return containerReports;
+  }
 }

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/AppSchedulable.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/AppSchedulable.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/AppSchedulable.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/AppSchedulable.java Wed Jun 11 12:20:48 2014
@@ -264,8 +264,7 @@ public class AppSchedulable extends Sche
       }
 
       // Inform the node
-      node.allocateContainer(app.getApplicationId(),
-          allocatedContainer);
+      node.allocateContainer(allocatedContainer);
 
       // If this container is used to run AM, update the leaf queue's AM usage
       if (app.getLiveContainers().size() == 1 &&

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSLeafQueue.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSLeafQueue.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSLeafQueue.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSLeafQueue.java Wed Jun 11 12:20:48 2014
@@ -36,6 +36,7 @@ import org.apache.hadoop.yarn.api.record
 import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.ActiveUsersManager;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.SchedulerAppUtils;
+import org.apache.hadoop.yarn.server.resourcemanager.scheduler.SchedulerApplicationAttempt;
 import org.apache.hadoop.yarn.util.resource.Resources;
 
 @Private
@@ -318,4 +319,10 @@ public class FSLeafQueue extends FSQueue
       Resources.addTo(amResourceUsage, amResource);
     }
   }
+
+  @Override
+  public void recoverContainer(Resource clusterResource,
+      SchedulerApplicationAttempt schedulerAttempt, RMContainer rmContainer) {
+    // TODO Auto-generated method stub
+  }
 }

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSParentQueue.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSParentQueue.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSParentQueue.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSParentQueue.java Wed Jun 11 12:20:48 2014
@@ -35,7 +35,9 @@ import org.apache.hadoop.yarn.api.record
 import org.apache.hadoop.yarn.api.records.Resource;
 import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer;
 import org.apache.hadoop.yarn.util.resource.Resources;
+import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.ActiveUsersManager;
+import org.apache.hadoop.yarn.server.resourcemanager.scheduler.SchedulerApplicationAttempt;
 
 @Private
 @Unstable
@@ -228,4 +230,11 @@ public class FSParentQueue extends FSQue
     // Should never be called since all applications are submitted to LeafQueues
     return null;
   }
+
+  @Override
+  public void recoverContainer(Resource clusterResource,
+      SchedulerApplicationAttempt schedulerAttempt, RMContainer rmContainer) {
+    // TODO Auto-generated method stub
+    
+  }
 }

Modified: hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fifo/FifoScheduler.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fifo/FifoScheduler.java?rev=1601869&r1=1601868&r2=1601869&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fifo/FifoScheduler.java (original)
+++ hadoop/common/branches/HDFS-5442/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fifo/FifoScheduler.java Wed Jun 11 12:20:48 2014
@@ -76,6 +76,7 @@ import org.apache.hadoop.yarn.server.res
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.QueueMetrics;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.SchedulerAppUtils;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.SchedulerApplication;
+import org.apache.hadoop.yarn.server.resourcemanager.scheduler.SchedulerApplicationAttempt;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.SchedulerApplicationAttempt.ContainersAndNMTokensAllocation;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.SchedulerUtils;
 import org.apache.hadoop.yarn.server.resourcemanager.scheduler.common.fica.FiCaSchedulerApp;
@@ -178,6 +179,17 @@ public class FifoScheduler extends
     public ActiveUsersManager getActiveUsersManager() {
       return activeUsersManager;
     }
+
+    @Override
+    public void recoverContainer(Resource clusterResource,
+        SchedulerApplicationAttempt schedulerAttempt, RMContainer rmContainer) {
+      if (rmContainer.getState().equals(RMContainerState.COMPLETED)) {
+        return;
+      }
+      increaseUsedResources(rmContainer);
+      updateAppHeadRoom(schedulerAttempt);
+      updateAvailableResourcesMetrics();
+    }
   };
 
   public FifoScheduler() {
@@ -488,7 +500,7 @@ public class FifoScheduler extends
       if (attempt == null) {
         continue;
       }
-      attempt.setHeadroom(Resources.subtract(clusterResource, usedResource));
+      updateAppHeadRoom(attempt);
     }
   }
 
@@ -659,11 +671,10 @@ public class FifoScheduler extends
             application.allocate(type, node, priority, request, container);
         
         // Inform the node
-        node.allocateContainer(application.getApplicationId(), 
-            rmContainer);
+        node.allocateContainer(rmContainer);
 
         // Update usage for this container
-        Resources.addTo(usedResource, capability);
+        increaseUsedResources(rmContainer);
       }
 
     }
@@ -707,9 +718,22 @@ public class FifoScheduler extends
       LOG.debug("Node after allocation " + rmNode.getNodeID() + " resource = "
           + node.getAvailableResource());
     }
-    
-    metrics.setAvailableResourcesToQueue(
-        Resources.subtract(clusterResource, usedResource));
+
+    updateAvailableResourcesMetrics();
+  }
+
+  private void increaseUsedResources(RMContainer rmContainer) {
+    Resources.addTo(usedResource, rmContainer.getAllocatedResource());
+  }
+
+  private void updateAppHeadRoom(SchedulerApplicationAttempt schedulerAttempt) {
+    schedulerAttempt.setHeadroom(Resources.subtract(clusterResource,
+      usedResource));
+  }
+
+  private void updateAvailableResourcesMetrics() {
+    metrics.setAvailableResourcesToQueue(Resources.subtract(clusterResource,
+      usedResource));
   }
 
   @Override
@@ -719,6 +743,9 @@ public class FifoScheduler extends
     {
       NodeAddedSchedulerEvent nodeAddedEvent = (NodeAddedSchedulerEvent)event;
       addNode(nodeAddedEvent.getAddedRMNode());
+      recoverContainersOnNode(nodeAddedEvent.getContainerReports(),
+        nodeAddedEvent.getAddedRMNode());
+
     }
     break;
     case NODE_REMOVED:
@@ -923,4 +950,8 @@ public class FifoScheduler extends
       return null;
     }
   }
+
+  public Resource getUsedResource() {
+    return usedResource;
+  }
 }



Mime
View raw message