HDFS-4231. BackupNode: Introduce BackupState. Contributed by Konstantin Shvachko.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/branches/branch-2@1416289 13f79535-47bb-0310-9956-ffa450edef68

hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt

@@ -273,6 +273,8 @@ Release 2.0.3-alpha - Unreleased
     of it is undefined after the iteration or modifications of the map.
     (szetszwo)
 
+    HDFS-4231. BackupNode: Introduce BackupState. (shv)
+
 Release 2.0.2-alpha - 2012-09-07 
 
   INCOMPATIBLE CHANGES

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/BackupNode.java

@@ -24,6 +24,7 @@ import java.net.SocketTimeoutException;
 import org.apache.hadoop.classification.InterfaceAudience;
 import org.apache.hadoop.conf.Configuration;
 import org.apache.hadoop.fs.CommonConfigurationKeys;
+import org.apache.hadoop.ha.ServiceFailedException;
 import org.apache.hadoop.hdfs.DFSConfigKeys;
 import org.apache.hadoop.hdfs.DFSUtil;
 import org.apache.hadoop.hdfs.NameNodeProxies;
@@ -35,6 +36,7 @@ import org.apache.hadoop.hdfs.protocolPB.JournalProtocolPB;
 import org.apache.hadoop.hdfs.protocolPB.JournalProtocolServerSideTranslatorPB;
 import org.apache.hadoop.hdfs.server.common.HdfsServerConstants.NamenodeRole;
 import org.apache.hadoop.hdfs.server.common.Storage;
+import org.apache.hadoop.hdfs.server.namenode.ha.HAState;
 import org.apache.hadoop.hdfs.server.protocol.FenceResponse;
 import org.apache.hadoop.hdfs.server.protocol.JournalInfo;
 import org.apache.hadoop.hdfs.server.protocol.JournalProtocol;
@@ -404,14 +406,23 @@ public class BackupNode extends NameNode {
       + HdfsConstants.LAYOUT_VERSION + " actual "+ nsInfo.getLayoutVersion();
     return nsInfo;
   }
-  
+
   @Override
+  protected String getNameServiceId(Configuration conf) {
+    return DFSUtil.getBackupNameServiceId(conf);
+  }
+
+  protected HAState createHAState() {
+    return new BackupState();
+  }
+
+  @Override // NameNode
   protected NameNodeHAContext createHAContext() {
     return new BNHAContext();
   }
-  
+
   private class BNHAContext extends NameNodeHAContext {
-    @Override // NameNode
+    @Override // NameNodeHAContext
     public void checkOperation(OperationCategory op)
         throws StandbyException {
       if (op == OperationCategory.UNCHECKED ||
@@ -425,10 +436,42 @@ public class BackupNode extends NameNode {
         throw new StandbyException(msg);
       }
     }
-  }
-  
-  @Override
-  protected String getNameServiceId(Configuration conf) {
-    return DFSUtil.getBackupNameServiceId(conf);
+
+    @Override // NameNodeHAContext
+    public void prepareToStopStandbyServices() throws ServiceFailedException {
+    }
+
+    /**
+     * Start services for BackupNode.
+     * <p>
+     * The following services should be muted
+     * (not run or not pass any control commands to DataNodes)
+     * on BackupNode:
+     * {@link LeaseManager.Monitor} protected by SafeMode.
+     * {@link BlockManager.ReplicationMonitor} protected by SafeMode.
+     * {@link HeartbeatManager.Monitor} protected by SafeMode.
+     * {@link DecommissionManager.Monitor} need to prohibit refreshNodes().
+     * {@link PendingReplicationBlocks.PendingReplicationMonitor} harmless,
+     * because ReplicationMonitor is muted.
+     */
+    @Override
+    public void startActiveServices() throws IOException {
+      try {
+        namesystem.startActiveServices();
+      } catch (Throwable t) {
+        doImmediateShutdown(t);
+      }
+    }
+
+    @Override
+    public void stopActiveServices() throws IOException {
+      try {
+        if (namesystem != null) {
+          namesystem.stopActiveServices();
+        }
+      } catch (Throwable t) {
+        doImmediateShutdown(t);
+      }
+    }
   }
 }

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/BackupState.java

@@ -0,0 +1,51 @@
+package org.apache.hadoop.hdfs.server.namenode;
+
+import java.io.IOException;
+
+import org.apache.hadoop.ha.HAServiceProtocol.HAServiceState;
+import org.apache.hadoop.ha.ServiceFailedException;
+import org.apache.hadoop.hdfs.server.namenode.NameNode.OperationCategory;
+import org.apache.hadoop.hdfs.server.namenode.ha.HAContext;
+import org.apache.hadoop.hdfs.server.namenode.ha.HAState;
+import org.apache.hadoop.ipc.StandbyException;
+
+public class BackupState extends HAState {
+
+  public BackupState() {
+    super(HAServiceState.STANDBY);
+  }
+
+  @Override // HAState
+  public void checkOperation(HAContext context, OperationCategory op)
+      throws StandbyException {
+    context.checkOperation(op);
+  }
+
+  @Override // HAState
+  public boolean shouldPopulateReplQueues() {
+    return false;
+  }
+
+  @Override
+  public void enterState(HAContext context) throws ServiceFailedException {
+    try {
+      context.startActiveServices();
+    } catch (IOException e) {
+      throw new ServiceFailedException("Failed to start backup services", e);
+    }
+  }
+
+  @Override
+  public void exitState(HAContext context) throws ServiceFailedException {
+    try {
+      context.stopActiveServices();
+    } catch (IOException e) {
+      throw new ServiceFailedException("Failed to stop backup services", e);
+    }
+  }
+
+  @Override
+  public void prepareToExitState(HAContext context) throws ServiceFailedException {
+    context.prepareToStopStandbyServices();
+  }
+}

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java

@@ -121,6 +121,7 @@ import org.apache.hadoop.fs.UnresolvedLinkException;
 import org.apache.hadoop.fs.permission.FsAction;
 import org.apache.hadoop.fs.permission.FsPermission;
 import org.apache.hadoop.fs.permission.PermissionStatus;
+import org.apache.hadoop.ha.HAServiceProtocol.HAServiceState;
 import org.apache.hadoop.ha.ServiceFailedException;
 import org.apache.hadoop.hdfs.DFSUtil;
 import org.apache.hadoop.hdfs.HAUtil;
@@ -161,7 +162,6 @@ import org.apache.hadoop.hdfs.server.common.Storage.StorageDirectory;
 import org.apache.hadoop.hdfs.server.common.Util;
 import org.apache.hadoop.hdfs.server.namenode.LeaseManager.Lease;
 import org.apache.hadoop.hdfs.server.namenode.NameNode.OperationCategory;
-import org.apache.hadoop.hdfs.server.namenode.ha.ActiveState;
 import org.apache.hadoop.hdfs.server.namenode.ha.EditLogTailer;
 import org.apache.hadoop.hdfs.server.namenode.ha.HAContext;
 import org.apache.hadoop.hdfs.server.namenode.ha.HAState;
@@ -3413,9 +3413,9 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
   private NNHAStatusHeartbeat createHaStatusHeartbeat() {
     HAState state = haContext.getState();
     NNHAStatusHeartbeat.State hbState;
-    if (state instanceof ActiveState) {
+    if (state.getServiceState() == HAServiceState.ACTIVE) {
       hbState = NNHAStatusHeartbeat.State.ACTIVE;
-    } else if (state instanceof StandbyState) {
+    } else if (state.getServiceState() == HAServiceState.STANDBY) {
       hbState = NNHAStatusHeartbeat.State.STANDBY;      
     } else {
       throw new AssertionError("Invalid state: " + state.getClass());

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/NameNode.java

@@ -600,11 +600,7 @@ public class NameNode {
     String nsId = getNameServiceId(conf);
     String namenodeId = HAUtil.getNameNodeId(conf, nsId);
     this.haEnabled = HAUtil.isHAEnabled(conf, nsId);
-    if (!haEnabled) {
-      state = ACTIVE_STATE;
-    } else {
-      state = STANDBY_STATE;
-    }
+    state = createHAState();
     this.allowStaleStandbyReads = HAUtil.shouldAllowStandbyReads(conf);
     this.haContext = createHAContext();
     try {
@@ -621,6 +617,10 @@ public class NameNode {
     }
   }
 
+  protected HAState createHAState() {
+    return !haEnabled ? ACTIVE_STATE : STANDBY_STATE;
+  }
+
   protected HAContext createHAContext() {
     return new NameNodeHAContext();
   }
@@ -1284,7 +1284,7 @@ public class NameNode {
    *          before exit.
    * @throws ExitException thrown only for testing.
    */
-  private synchronized void doImmediateShutdown(Throwable t)
+  protected synchronized void doImmediateShutdown(Throwable t)
       throws ExitException {
     String message = "Error encountered requiring NN shutdown. " +
         "Shutting down immediately.";

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestBackupNode.java

@@ -35,6 +35,7 @@ import org.apache.hadoop.conf.Configuration;
 import org.apache.hadoop.fs.FileSystem;
 import org.apache.hadoop.fs.FileUtil;
 import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.ha.HAServiceProtocol.HAServiceState;
 import org.apache.hadoop.hdfs.DFSConfigKeys;
 import org.apache.hadoop.hdfs.HAUtil;
 import org.apache.hadoop.hdfs.HdfsConfiguration;
@@ -98,6 +99,9 @@ public class TestBackupNode {
     BackupNode bn = (BackupNode)NameNode.createNameNode(
         new String[]{startupOpt.getName()}, c);
     assertTrue(bn.getRole() + " must be in SafeMode.", bn.isInSafeMode());
+    assertTrue(bn.getRole() + " must be in StandbyState",
+               bn.getNamesystem().getHAState()
+                 .equalsIgnoreCase(HAServiceState.STANDBY.name()));
     return bn;
   }