Merge trunk to HDFS-4685.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/branches/HDFS-4685@1556665 13f79535-47bb-0310-9956-ffa450edef68

hadoop-common-project/hadoop-common/CHANGES.txt

@@ -412,6 +412,9 @@ Release 2.4.0 - UNRELEASED
     HADOOP-10198. DomainSocket: add support for socketpair.
     (Colin Patrick McCabe via wang)
 
+    HADOOP-10208. Remove duplicate initialization in StringUtils.getStringCollection.
+    (Benoy Antony via jing9)
+
   OPTIMIZATIONS
 
     HADOOP-9748. Reduce blocking on UGI.ensureInitialized (daryn)

hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/ha/ActiveStandbyElector.java

@@ -766,8 +766,9 @@ public class ActiveStandbyElector implements StatCallback, StringCallback {
     zkClient = getNewZooKeeper();
     LOG.debug("Created new connection for " + this);
   }
-  
+
-  void terminateConnection() {
+  @InterfaceAudience.Private
+  public void terminateConnection() {
     if (zkClient == null) {
       return;
     }

hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/util/StringUtils.java

@@ -343,7 +343,6 @@ public class StringUtils {
     if (str == null)
       return values;
     StringTokenizer tokenizer = new StringTokenizer(str, delim);
-    values = new ArrayList<String>();
     while (tokenizer.hasMoreTokens()) {
       values.add(tokenizer.nextToken());
     }

hadoop-common-project/hadoop-nfs/src/main/java/org/apache/hadoop/mount/MountdBase.java

@@ -23,6 +23,7 @@ import org.apache.hadoop.oncrpc.RpcProgram;
 import org.apache.hadoop.oncrpc.SimpleTcpServer;
 import org.apache.hadoop.oncrpc.SimpleUdpServer;
 import org.apache.hadoop.portmap.PortmapMapping;
+import org.apache.hadoop.util.ShutdownHookManager;
 
 /**
  * Main class for starting mountd daemon. This daemon implements the NFS
@@ -71,8 +72,24 @@ abstract public class MountdBase {
     startUDPServer();
     startTCPServer();
     if (register) {
+      ShutdownHookManager.get().addShutdownHook(new Unregister(),
+          SHUTDOWN_HOOK_PRIORITY);
       rpcProgram.register(PortmapMapping.TRANSPORT_UDP, udpBoundPort);
       rpcProgram.register(PortmapMapping.TRANSPORT_TCP, tcpBoundPort);
     }
   }
+  
+  /**
+   * Priority of the mountd shutdown hook.
+   */
+  public static final int SHUTDOWN_HOOK_PRIORITY = 10;
+
+  private class Unregister implements Runnable {
+    @Override
+    public synchronized void run() {
+      rpcProgram.unregister(PortmapMapping.TRANSPORT_UDP, udpBoundPort);
+      rpcProgram.unregister(PortmapMapping.TRANSPORT_TCP, tcpBoundPort);
+    }
+  }
+  
 }

hadoop-common-project/hadoop-nfs/src/main/java/org/apache/hadoop/nfs/nfs3/Nfs3Base.java

@@ -23,6 +23,7 @@ import org.apache.hadoop.conf.Configuration;
 import org.apache.hadoop.oncrpc.RpcProgram;
 import org.apache.hadoop.oncrpc.SimpleTcpServer;
 import org.apache.hadoop.portmap.PortmapMapping;
+import org.apache.hadoop.util.ShutdownHookManager;
 
 /**
  * Nfs server. Supports NFS v3 using {@link RpcProgram}.
@@ -50,6 +51,8 @@ public abstract class Nfs3Base {
     startTCPServer(); // Start TCP server
     
     if (register) {
+      ShutdownHookManager.get().addShutdownHook(new Unregister(),
+          SHUTDOWN_HOOK_PRIORITY);
       rpcProgram.register(PortmapMapping.TRANSPORT_TCP, nfsBoundPort);
     }
   }
@@ -61,4 +64,16 @@ public abstract class Nfs3Base {
     tcpServer.run();
     nfsBoundPort = tcpServer.getBoundPort();
   }
+  
+  /**
+   * Priority of the nfsd shutdown hook.
+   */
+  public static final int SHUTDOWN_HOOK_PRIORITY = 10;
+
+  private class Unregister implements Runnable {
+    @Override
+    public synchronized void run() {
+      rpcProgram.unregister(PortmapMapping.TRANSPORT_TCP, nfsBoundPort);
+    }
+  }
 }

hadoop-common-project/hadoop-nfs/src/main/java/org/apache/hadoop/oncrpc/RpcProgram.java

@@ -78,23 +78,41 @@ public abstract class RpcProgram extends SimpleChannelUpstreamHandler {
     for (int vers = lowProgVersion; vers <= highProgVersion; vers++) {
       PortmapMapping mapEntry = new PortmapMapping(progNumber, vers, transport,
           port);
-      register(mapEntry);
+      register(mapEntry, true);
+    }
+  }
+  
+  /**
+   * Unregister this program with the local portmapper.
+   */
+  public void unregister(int transport, int boundPort) {
+    if (boundPort != port) {
+      LOG.info("The bound port is " + boundPort
+          + ", different with configured port " + port);
+      port = boundPort;
+    }
+    // Unregister all the program versions with portmapper for a given transport
+    for (int vers = lowProgVersion; vers <= highProgVersion; vers++) {
+      PortmapMapping mapEntry = new PortmapMapping(progNumber, vers, transport,
+          port);
+      register(mapEntry, false);
     }
   }
   
   /**
    * Register the program with Portmap or Rpcbind
    */
-  protected void register(PortmapMapping mapEntry) {
+  protected void register(PortmapMapping mapEntry, boolean set) {
-    XDR mappingRequest = PortmapRequest.create(mapEntry);
+    XDR mappingRequest = PortmapRequest.create(mapEntry, set);
     SimpleUdpClient registrationClient = new SimpleUdpClient(host, RPCB_PORT,
         mappingRequest);
     try {
       registrationClient.run();
     } catch (IOException e) {
-      LOG.error("Registration failure with " + host + ":" + port
+      String request = set ? "Registration" : "Unregistration";
+      LOG.error(request + " failure with " + host + ":" + port
           + ", portmap entry: " + mapEntry);
-      throw new RuntimeException("Registration failure");
+      throw new RuntimeException(request + " failure");
     }
   }
 

hadoop-common-project/hadoop-nfs/src/main/java/org/apache/hadoop/portmap/PortmapRequest.java

@@ -31,13 +31,14 @@ public class PortmapRequest {
     return PortmapMapping.deserialize(xdr);
   }
 
-  public static XDR create(PortmapMapping mapping) {
+  public static XDR create(PortmapMapping mapping, boolean set) {
     XDR request = new XDR();
+    int procedure = set ? RpcProgramPortmap.PMAPPROC_SET
+        : RpcProgramPortmap.PMAPPROC_UNSET;
     RpcCall call = RpcCall.getInstance(
         RpcUtil.getNewXid(String.valueOf(RpcProgramPortmap.PROGRAM)),
-        RpcProgramPortmap.PROGRAM, RpcProgramPortmap.VERSION,
+        RpcProgramPortmap.PROGRAM, RpcProgramPortmap.VERSION, procedure,
-        RpcProgramPortmap.PMAPPROC_SET, new CredentialsNone(),
+        new CredentialsNone(), new VerifierNone());
-        new VerifierNone());
     call.write(request);
     return mapping.serialize(request);
   }

hadoop-hdfs-project/hadoop-hdfs-nfs/src/main/java/org/apache/hadoop/hdfs/nfs/nfs3/DFSClientCache.java

@@ -19,6 +19,10 @@ package org.apache.hadoop.hdfs.nfs.nfs3;
 
 import java.io.IOException;
 import java.security.PrivilegedExceptionAction;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.Map.Entry;
+import java.util.concurrent.ConcurrentMap;
 import java.util.concurrent.ExecutionException;
 import java.util.concurrent.TimeUnit;
 
@@ -29,7 +33,9 @@ import org.apache.hadoop.fs.FSDataInputStream;
 import org.apache.hadoop.hdfs.DFSClient;
 import org.apache.hadoop.hdfs.DFSInputStream;
 import org.apache.hadoop.hdfs.server.namenode.NameNode;
+import org.apache.hadoop.io.MultipleIOException;
 import org.apache.hadoop.security.UserGroupInformation;
+import org.apache.hadoop.util.ShutdownHookManager;
 
 import com.google.common.annotations.VisibleForTesting;
 import com.google.common.base.Objects;
@@ -94,7 +100,7 @@ class DFSClientCache {
   DFSClientCache(Configuration config) {
     this(config, DEFAULT_DFS_CLIENT_CACHE_SIZE);
   }
-
+  
   DFSClientCache(Configuration config, int clientCache) {
     this.config = config;
     this.clientCache = CacheBuilder.newBuilder()
@@ -107,8 +113,52 @@ class DFSClientCache {
         .expireAfterAccess(DEFAULT_DFS_INPUTSTREAM_CACHE_TTL, TimeUnit.SECONDS)
         .removalListener(inputStreamRemovalListener())
         .build(inputStreamLoader());
+    
+    ShutdownHookManager.get().addShutdownHook(new CacheFinalizer(),
+        SHUTDOWN_HOOK_PRIORITY);
   }
 
+  /**
+   * Priority of the FileSystem shutdown hook.
+   */
+  public static final int SHUTDOWN_HOOK_PRIORITY = 10;
+  
+  private class CacheFinalizer implements Runnable {
+    @Override
+    public synchronized void run() {
+      try {
+        closeAll(true);
+      } catch (IOException e) {
+        LOG.info("DFSClientCache.closeAll() threw an exception:\n", e);
+      }
+    }
+  }
+  
+  /**
+   * Close all DFSClient instances in the Cache.
+   * @param onlyAutomatic only close those that are marked for automatic closing
+   */
+  synchronized void closeAll(boolean onlyAutomatic) throws IOException {
+    List<IOException> exceptions = new ArrayList<IOException>();
+
+    ConcurrentMap<String, DFSClient> map = clientCache.asMap();
+
+    for (Entry<String, DFSClient> item : map.entrySet()) {
+      final DFSClient client = item.getValue();
+      if (client != null) {
+        try {
+          client.close();
+        } catch (IOException ioe) {
+          exceptions.add(ioe);
+        }
+      }
+    }
+
+    if (!exceptions.isEmpty()) {
+      throw MultipleIOException.createIOException(exceptions);
+    }
+  }
+  
   private CacheLoader<String, DFSClient> clientLoader() {
     return new CacheLoader<String, DFSClient>() {
       @Override

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

@@ -242,6 +242,9 @@ Trunk (Unreleased)
     HDFS-5651. Remove dfs.namenode.caching.enabled and improve CRM locking.
     (cmccabe via wang)
 
+    HDFS-5715. Use Snapshot ID to indicate the corresponding Snapshot for a
+    FileDiff/DirectoryDiff. (jing9)
+
   OPTIMIZATIONS
 
     HDFS-5349. DNA_CACHE and DNA_UNCACHE should be by blockId only. (cmccabe)
@@ -467,6 +470,11 @@ Trunk (Unreleased)
     HDFS-5589. Namenode loops caching and uncaching when data should be
     uncached (awang via cmccabe)
 
+    HDFS-5724. modifyCacheDirective logging audit log command wrongly as
+    addCacheDirective (Uma Maheswara Rao G via Colin Patrick McCabe)
+
+    HDFS-5726. Fix compilation error in AbstractINodeDiff for JDK7. (jing9)
+
 Release 2.4.0 - UNRELEASED
 
   INCOMPATIBLE CHANGES
@@ -1055,6 +1063,9 @@ Release 2.3.0 - UNRELEASED
 
     HDFS-5671. Fix socket leak in DFSInputStream#getBlockReader. (JamesLi via umamahesh) 
 
+    HDFS-5649. Unregister NFS and Mount service when NFS gateway is shutting down.
+    (brandonli)
+
 Release 2.2.0 - 2013-10-13
 
   INCOMPATIBLE CHANGES

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/CacheReplicationMonitor.java

@@ -49,6 +49,7 @@ import org.apache.hadoop.hdfs.server.namenode.FSNamesystem;
 import org.apache.hadoop.hdfs.server.namenode.INode;
 import org.apache.hadoop.hdfs.server.namenode.INodeDirectory;
 import org.apache.hadoop.hdfs.server.namenode.INodeFile;
+import org.apache.hadoop.hdfs.server.namenode.snapshot.Snapshot;
 import org.apache.hadoop.hdfs.util.ReadOnlyList;
 import org.apache.hadoop.util.GSet;
 import org.apache.hadoop.util.Time;
@@ -339,7 +340,8 @@ public class CacheReplicationMonitor extends Thread implements Closeable {
         }
       } else if (node.isDirectory()) {
         INodeDirectory dir = node.asDirectory();
-        ReadOnlyList<INode> children = dir.getChildrenList(null);
+        ReadOnlyList<INode> children = dir
+            .getChildrenList(Snapshot.CURRENT_STATE_ID);
         for (INode child : children) {
           if (child.isFile()) {
             rescanFile(directive, child.asFile());

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

@@ -53,7 +53,6 @@ import org.apache.hadoop.fs.InvalidRequestException;
 import org.apache.hadoop.fs.UnresolvedLinkException;
 import org.apache.hadoop.fs.permission.FsAction;
 import org.apache.hadoop.hdfs.DFSUtil;
-import org.apache.hadoop.hdfs.protocol.Block;
 import org.apache.hadoop.hdfs.protocol.CacheDirective;
 import org.apache.hadoop.hdfs.protocol.CacheDirectiveEntry;
 import org.apache.hadoop.hdfs.protocol.CacheDirectiveInfo;
@@ -69,6 +68,7 @@ import org.apache.hadoop.hdfs.server.blockmanagement.DatanodeDescriptor;
 import org.apache.hadoop.hdfs.server.blockmanagement.DatanodeDescriptor.CachedBlocksList;
 import org.apache.hadoop.hdfs.server.blockmanagement.DatanodeDescriptor.CachedBlocksList.Type;
 import org.apache.hadoop.hdfs.server.namenode.metrics.NameNodeMetrics;
+import org.apache.hadoop.hdfs.server.namenode.snapshot.Snapshot;
 import org.apache.hadoop.hdfs.server.namenode.startupprogress.Phase;
 import org.apache.hadoop.hdfs.server.namenode.startupprogress.StartupProgress;
 import org.apache.hadoop.hdfs.server.namenode.startupprogress.StartupProgress.Counter;
@@ -399,7 +399,8 @@ public final class CacheManager {
       requestedBytes = file.computeFileSize();
     } else if (node.isDirectory()) {
       INodeDirectory dir = node.asDirectory();
-      ReadOnlyList<INode> children = dir.getChildrenList(null);
+      ReadOnlyList<INode> children = dir
+          .getChildrenList(Snapshot.CURRENT_STATE_ID);
       requestedFiles = children.size();
       for (INode child : children) {
         if (child.isFile()) {

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

@@ -616,14 +616,14 @@ public class FSDirectory implements Closeable {
     INode srcChild = srcIIP.getLastINode();
     final byte[] srcChildName = srcChild.getLocalNameBytes();
     final boolean isSrcInSnapshot = srcChild.isInLatestSnapshot(
-        srcIIP.getLatestSnapshot());
+        srcIIP.getLatestSnapshotId());
     final boolean srcChildIsReference = srcChild.isReference();
     
     // Record the snapshot on srcChild. After the rename, before any new 
     // snapshot is taken on the dst tree, changes will be recorded in the latest
     // snapshot of the src tree.
     if (isSrcInSnapshot) {
-      srcChild = srcChild.recordModification(srcIIP.getLatestSnapshot());
+      srcChild = srcChild.recordModification(srcIIP.getLatestSnapshotId());
       srcIIP.setLastINode(srcChild);
     }
     
@@ -631,17 +631,16 @@ public class FSDirectory implements Closeable {
     final INodeReference.WithCount withCount;
     Quota.Counts oldSrcCounts = Quota.Counts.newInstance();
     int srcRefDstSnapshot = srcChildIsReference ? srcChild.asReference()
-        .getDstSnapshotId() : Snapshot.INVALID_ID;
+        .getDstSnapshotId() : Snapshot.CURRENT_STATE_ID;
     if (isSrcInSnapshot) {
       final INodeReference.WithName withName = 
           srcIIP.getINode(-2).asDirectory().replaceChild4ReferenceWithName(
-              srcChild, srcIIP.getLatestSnapshot()); 
+              srcChild, srcIIP.getLatestSnapshotId()); 
       withCount = (INodeReference.WithCount) withName.getReferredINode();
       srcChild = withName;
       srcIIP.setLastINode(srcChild);
       // get the counts before rename
-      withCount.getReferredINode().computeQuotaUsage(oldSrcCounts, true,
+      withCount.getReferredINode().computeQuotaUsage(oldSrcCounts, true);
-          Snapshot.INVALID_ID);
     } else if (srcChildIsReference) {
       // srcChild is reference but srcChild is not in latest snapshot
       withCount = (WithCount) srcChild.asReference().getReferredINode();
@@ -677,10 +676,9 @@ public class FSDirectory implements Closeable {
         toDst = srcChild;
       } else {
         withCount.getReferredINode().setLocalName(dstChildName);
-        Snapshot dstSnapshot = dstIIP.getLatestSnapshot();
+        int dstSnapshotId = dstIIP.getLatestSnapshotId();
         final INodeReference.DstReference ref = new INodeReference.DstReference(
-            dstParent.asDirectory(), withCount,
+            dstParent.asDirectory(), withCount, dstSnapshotId);
-            dstSnapshot == null ? Snapshot.INVALID_ID : dstSnapshot.getId());
         toDst = ref;
       }
       
@@ -692,9 +690,9 @@ public class FSDirectory implements Closeable {
         }
         // update modification time of dst and the parent of src
         final INode srcParent = srcIIP.getINode(-2);
-        srcParent.updateModificationTime(timestamp, srcIIP.getLatestSnapshot());
+        srcParent.updateModificationTime(timestamp, srcIIP.getLatestSnapshotId());
         dstParent = dstIIP.getINode(-2); // refresh dstParent
-        dstParent.updateModificationTime(timestamp, dstIIP.getLatestSnapshot());
+        dstParent.updateModificationTime(timestamp, dstIIP.getLatestSnapshotId());
         // update moved leases with new filename
         getFSNamesystem().unprotectedChangeLease(src, dst);     
 
@@ -702,7 +700,7 @@ public class FSDirectory implements Closeable {
         if (isSrcInSnapshot) {
           // get the counts after rename
           Quota.Counts newSrcCounts = srcChild.computeQuotaUsage(
-              Quota.Counts.newInstance(), false, Snapshot.INVALID_ID);
+              Quota.Counts.newInstance(), false);
           newSrcCounts.subtract(oldSrcCounts);
           srcParent.addSpaceConsumed(newSrcCounts.get(Quota.NAMESPACE),
               newSrcCounts.get(Quota.DISKSPACE), false);
@@ -734,8 +732,7 @@ public class FSDirectory implements Closeable {
         if (isSrcInSnapshot) {
           // srcParent must have snapshot feature since isSrcInSnapshot is true
           // and src node has been removed from srcParent 
-          srcParent.undoRename4ScrParent(oldSrcChild.asReference(), srcChild,
+          srcParent.undoRename4ScrParent(oldSrcChild.asReference(), srcChild);
-              srcIIP.getLatestSnapshot());
         } else {
           // original srcChild is not in latest snapshot, we only need to add
           // the srcChild back
@@ -838,7 +835,7 @@ public class FSDirectory implements Closeable {
       }
       if (dstInode.isDirectory()) {
         final ReadOnlyList<INode> children = dstInode.asDirectory()
-            .getChildrenList(null);
+            .getChildrenList(Snapshot.CURRENT_STATE_ID);
         if (!children.isEmpty()) {
           error = "rename destination directory is not empty: " + dst;
           NameNode.stateChangeLog.warn(
@@ -869,31 +866,30 @@ public class FSDirectory implements Closeable {
     INode srcChild = srcIIP.getLastINode();
     final byte[] srcChildName = srcChild.getLocalNameBytes();
     final boolean isSrcInSnapshot = srcChild.isInLatestSnapshot(
-        srcIIP.getLatestSnapshot());
+        srcIIP.getLatestSnapshotId());
     final boolean srcChildIsReference = srcChild.isReference();
     
     // Record the snapshot on srcChild. After the rename, before any new 
     // snapshot is taken on the dst tree, changes will be recorded in the latest
     // snapshot of the src tree.
     if (isSrcInSnapshot) {
-      srcChild = srcChild.recordModification(srcIIP.getLatestSnapshot());
+      srcChild = srcChild.recordModification(srcIIP.getLatestSnapshotId());
       srcIIP.setLastINode(srcChild);
     }
     
     // check srcChild for reference
     final INodeReference.WithCount withCount;
     int srcRefDstSnapshot = srcChildIsReference ? srcChild.asReference()
-        .getDstSnapshotId() : Snapshot.INVALID_ID;
+        .getDstSnapshotId() : Snapshot.CURRENT_STATE_ID;
     Quota.Counts oldSrcCounts = Quota.Counts.newInstance();    
     if (isSrcInSnapshot) {
       final INodeReference.WithName withName = srcIIP.getINode(-2).asDirectory()
-          .replaceChild4ReferenceWithName(srcChild, srcIIP.getLatestSnapshot()); 
+          .replaceChild4ReferenceWithName(srcChild, srcIIP.getLatestSnapshotId()); 
       withCount = (INodeReference.WithCount) withName.getReferredINode();
       srcChild = withName;
       srcIIP.setLastINode(srcChild);
       // get the counts before rename
-      withCount.getReferredINode().computeQuotaUsage(oldSrcCounts, true,
+      withCount.getReferredINode().computeQuotaUsage(oldSrcCounts, true);
-          Snapshot.INVALID_ID);
     } else if (srcChildIsReference) {
       // srcChild is reference but srcChild is not in latest snapshot
       withCount = (WithCount) srcChild.asReference().getReferredINode();
@@ -937,10 +933,9 @@ public class FSDirectory implements Closeable {
         toDst = srcChild;
       } else {
         withCount.getReferredINode().setLocalName(dstChildName);
-        Snapshot dstSnapshot = dstIIP.getLatestSnapshot();
+        int dstSnapshotId = dstIIP.getLatestSnapshotId();
         final INodeReference.DstReference ref = new INodeReference.DstReference(
-            dstIIP.getINode(-2).asDirectory(), withCount,
+            dstIIP.getINode(-2).asDirectory(), withCount, dstSnapshotId);
-            dstSnapshot == null ? Snapshot.INVALID_ID : dstSnapshot.getId());
         toDst = ref;
       }
 
@@ -954,9 +949,9 @@ public class FSDirectory implements Closeable {
         }
 
         final INode srcParent = srcIIP.getINode(-2);
-        srcParent.updateModificationTime(timestamp, srcIIP.getLatestSnapshot());
+        srcParent.updateModificationTime(timestamp, srcIIP.getLatestSnapshotId());
         dstParent = dstIIP.getINode(-2);
-        dstParent.updateModificationTime(timestamp, dstIIP.getLatestSnapshot());
+        dstParent.updateModificationTime(timestamp, dstIIP.getLatestSnapshotId());
         // update moved lease with new filename
         getFSNamesystem().unprotectedChangeLease(src, dst);
 
@@ -966,8 +961,8 @@ public class FSDirectory implements Closeable {
           undoRemoveDst = false;
           BlocksMapUpdateInfo collectedBlocks = new BlocksMapUpdateInfo();
           List<INode> removedINodes = new ChunkedArrayList<INode>();
-          filesDeleted = removedDst.cleanSubtree(null,
+          filesDeleted = removedDst.cleanSubtree(Snapshot.CURRENT_STATE_ID,
-              dstIIP.getLatestSnapshot(), collectedBlocks, removedINodes, true)
+              dstIIP.getLatestSnapshotId(), collectedBlocks, removedINodes, true)
               .get(Quota.NAMESPACE);
           getFSNamesystem().removePathAndBlocks(src, collectedBlocks,
               removedINodes);
@@ -983,7 +978,7 @@ public class FSDirectory implements Closeable {
         if (isSrcInSnapshot) {
           // get the counts after rename
           Quota.Counts newSrcCounts = srcChild.computeQuotaUsage(
-              Quota.Counts.newInstance(), false, Snapshot.INVALID_ID);
+              Quota.Counts.newInstance(), false);
           newSrcCounts.subtract(oldSrcCounts);
           srcParent.addSpaceConsumed(newSrcCounts.get(Quota.NAMESPACE),
               newSrcCounts.get(Quota.DISKSPACE), false);
@@ -1014,8 +1009,7 @@ public class FSDirectory implements Closeable {
         }
         
         if (srcParent.isWithSnapshot()) {
-          srcParent.undoRename4ScrParent(oldSrcChild.asReference(), srcChild,
+          srcParent.undoRename4ScrParent(oldSrcChild.asReference(), srcChild);
-              srcIIP.getLatestSnapshot());
         } else {
           // srcParent is not an INodeDirectoryWithSnapshot, we only need to add
           // the srcChild back
@@ -1026,7 +1020,7 @@ public class FSDirectory implements Closeable {
         // Rename failed - restore dst
         if (dstParent.isDirectory() && dstParent.asDirectory().isWithSnapshot()) {
           dstParent.asDirectory().undoRename4DstParent(removedDst,
-              dstIIP.getLatestSnapshot());
+              dstIIP.getLatestSnapshotId());
         } else {
           addLastINodeNoQuotaCheck(dstIIP, removedDst);
         }
@@ -1090,7 +1084,7 @@ public class FSDirectory implements Closeable {
       updateCount(iip, 0, dsDelta, true);
     }
 
-    file = file.setFileReplication(replication, iip.getLatestSnapshot(),
+    file = file.setFileReplication(replication, iip.getLatestSnapshotId(),
         inodeMap);
     
     final short newBR = file.getBlockReplication(); 
@@ -1157,7 +1151,7 @@ public class FSDirectory implements Closeable {
     if (inode == null) {
       throw new FileNotFoundException("File does not exist: " + src);
     }
-    inode.setPermission(permissions, inodesInPath.getLatestSnapshot());
+    inode.setPermission(permissions, inodesInPath.getLatestSnapshotId());
   }
 
   void setOwner(String src, String username, String groupname)
@@ -1182,10 +1176,10 @@ public class FSDirectory implements Closeable {
       throw new FileNotFoundException("File does not exist: " + src);
     }
     if (username != null) {
-      inode = inode.setUser(username, inodesInPath.getLatestSnapshot());
+      inode = inode.setUser(username, inodesInPath.getLatestSnapshotId());
     }
     if (groupname != null) {
-      inode.setGroup(groupname, inodesInPath.getLatestSnapshot());
+      inode.setGroup(groupname, inodesInPath.getLatestSnapshotId());
     }
   }
 
@@ -1227,12 +1221,12 @@ public class FSDirectory implements Closeable {
     final INode[] trgINodes = trgIIP.getINodes();
     final INodeFile trgInode = trgIIP.getLastINode().asFile();
     INodeDirectory trgParent = trgINodes[trgINodes.length-2].asDirectory();
-    final Snapshot trgLatestSnapshot = trgIIP.getLatestSnapshot();
+    final int trgLatestSnapshot = trgIIP.getLatestSnapshotId();
     
     final INodeFile [] allSrcInodes = new INodeFile[srcs.length];
     for(int i = 0; i < srcs.length; i++) {
       final INodesInPath iip = getINodesInPath4Write(srcs[i]);
-      final Snapshot latest = iip.getLatestSnapshot();
+      final int latest = iip.getLatestSnapshotId();
       final INode inode = iip.getLastINode();
 
       // check if the file in the latest snapshot
@@ -1356,7 +1350,7 @@ public class FSDirectory implements Closeable {
         //not found or not a directory
         return false;
       }
-      final Snapshot s = inodesInPath.getPathSnapshot();
+      final int s = inodesInPath.getPathSnapshotId();
       return !inode.asDirectory().getChildrenList(s).isEmpty();
     } finally {
       readUnlock();
@@ -1410,7 +1404,7 @@ public class FSDirectory implements Closeable {
     }
 
     // record modification
-    final Snapshot latestSnapshot = iip.getLatestSnapshot();
+    final int latestSnapshot = iip.getLatestSnapshotId();
     targetNode = targetNode.recordModification(latestSnapshot);
     iip.setLastINode(targetNode);
 
@@ -1431,8 +1425,8 @@ public class FSDirectory implements Closeable {
     if (!targetNode.isInLatestSnapshot(latestSnapshot)) {
       targetNode.destroyAndCollectBlocks(collectedBlocks, removedINodes);
     } else {
-      Quota.Counts counts = targetNode.cleanSubtree(null, latestSnapshot,
+      Quota.Counts counts = targetNode.cleanSubtree(Snapshot.CURRENT_STATE_ID,
-          collectedBlocks, removedINodes, true);
+          latestSnapshot, collectedBlocks, removedINodes, true);
       parent.addSpaceConsumed(-counts.get(Quota.NAMESPACE),
           -counts.get(Quota.DISKSPACE), true);
       removed = counts.get(Quota.NAMESPACE);
@@ -1469,7 +1463,7 @@ public class FSDirectory implements Closeable {
           }
         }
       } 
-      for (INode child : targetDir.getChildrenList(null)) {
+      for (INode child : targetDir.getChildrenList(Snapshot.CURRENT_STATE_ID)) {
         checkSnapshot(child, snapshottableDirs);
       }
     }
@@ -1493,7 +1487,7 @@ public class FSDirectory implements Closeable {
         return getSnapshotsListing(srcs, startAfter);
       }
       final INodesInPath inodesInPath = rootDir.getLastINodeInPath(srcs, true);
-      final Snapshot snapshot = inodesInPath.getPathSnapshot();
+      final int snapshot = inodesInPath.getPathSnapshotId();
       final INode targetNode = inodesInPath.getINode(0);
       if (targetNode == null)
         return null;
@@ -1545,7 +1539,8 @@ public class FSDirectory implements Closeable {
     final HdfsFileStatus listing[] = new HdfsFileStatus[numOfListing];
     for (int i = 0; i < numOfListing; i++) {
       Root sRoot = snapshots.get(i + skipSize).getRoot();
-      listing[i] = createFileStatus(sRoot.getLocalNameBytes(), sRoot, null);
+      listing[i] = createFileStatus(sRoot.getLocalNameBytes(), sRoot,
+          Snapshot.CURRENT_STATE_ID);
     }
     return new DirectoryListing(
         listing, snapshots.size() - skipSize - numOfListing);
@@ -1568,7 +1563,7 @@ public class FSDirectory implements Closeable {
       final INodesInPath inodesInPath = rootDir.getLastINodeInPath(srcs, resolveLink);
       final INode i = inodesInPath.getINode(0);
       return i == null? null: createFileStatus(HdfsFileStatus.EMPTY_NAME, i,
-          inodesInPath.getPathSnapshot());
+          inodesInPath.getPathSnapshotId());
     } finally {
       readUnlock();
     }
@@ -2131,7 +2126,7 @@ public class FSDirectory implements Closeable {
     }
 
     final INodeDirectory parent = pathComponents[pos-1].asDirectory();
-    final int count = parent.getChildrenList(null).size();
+    final int count = parent.getChildrenList(Snapshot.CURRENT_STATE_ID).size();
     if (count >= maxDirItems) {
       final MaxDirectoryItemsExceededException e
           = new MaxDirectoryItemsExceededException(maxDirItems, count);
@@ -2195,7 +2190,7 @@ public class FSDirectory implements Closeable {
     final INodeDirectory parent = inodes[pos-1].asDirectory();
     boolean added = false;
     try {
-      added = parent.addChild(child, true, iip.getLatestSnapshot());
+      added = parent.addChild(child, true, iip.getLatestSnapshotId());
     } catch (QuotaExceededException e) {
       updateCountNoQuotaCheck(iip, pos,
           -counts.get(Quota.NAMESPACE), -counts.get(Quota.DISKSPACE));
@@ -2230,7 +2225,7 @@ public class FSDirectory implements Closeable {
    */
   private long removeLastINode(final INodesInPath iip)
       throws QuotaExceededException {
-    final Snapshot latestSnapshot = iip.getLatestSnapshot();
+    final int latestSnapshot = iip.getLatestSnapshotId();
     final INode last = iip.getLastINode();
     final INodeDirectory parent = iip.getINode(-2).asDirectory();
     if (!parent.removeChild(last, latestSnapshot)) {
@@ -2384,7 +2379,7 @@ public class FSDirectory implements Closeable {
         return null;
       }
 
-      final Snapshot latest = iip.getLatestSnapshot();
+      final int latest = iip.getLatestSnapshotId();
       dirNode = dirNode.recordModification(latest);
       dirNode.setQuota(nsQuota, dsQuota);
       return dirNode;
@@ -2427,11 +2422,11 @@ public class FSDirectory implements Closeable {
    * Sets the access time on the file/directory. Logs it in the transaction log.
    */
   void setTimes(String src, INode inode, long mtime, long atime, boolean force,
-      Snapshot latest) throws QuotaExceededException {
+      int latestSnapshotId) throws QuotaExceededException {
     boolean status = false;
     writeLock();
     try {
-      status = unprotectedSetTimes(inode, mtime, atime, force, latest);
+      status = unprotectedSetTimes(inode, mtime, atime, force, latestSnapshotId);
     } finally {
       writeUnlock();
     }
@@ -2445,11 +2440,11 @@ public class FSDirectory implements Closeable {
     assert hasWriteLock();
     final INodesInPath i = getLastINodeInPath(src); 
     return unprotectedSetTimes(i.getLastINode(), mtime, atime, force,
-        i.getLatestSnapshot());
+        i.getLatestSnapshotId());
   }
 
   private boolean unprotectedSetTimes(INode inode, long mtime,
-      long atime, boolean force, Snapshot latest) throws QuotaExceededException {
+      long atime, boolean force, int latest) throws QuotaExceededException {
     assert hasWriteLock();
     boolean status = false;
     if (mtime != -1) {
@@ -2457,7 +2452,7 @@ public class FSDirectory implements Closeable {
       status = true;
     }
     if (atime != -1) {
-      long inodeTime = inode.getAccessTime(null);
+      long inodeTime = inode.getAccessTime();
 
       // if the last access time update was within the last precision interval, then
       // no need to store access time
@@ -2497,7 +2492,7 @@ public class FSDirectory implements Closeable {
    * @throws IOException if any error occurs
    */
   private HdfsFileStatus createFileStatus(byte[] path, INode node,
-      boolean needLocation, Snapshot snapshot) throws IOException {
+      boolean needLocation, int snapshot) throws IOException {
     if (needLocation) {
       return createLocatedFileStatus(path, node, snapshot);
     } else {
@@ -2508,7 +2503,7 @@ public class FSDirectory implements Closeable {
    * Create FileStatus by file INode 
    */
    HdfsFileStatus createFileStatus(byte[] path, INode node,
-       Snapshot snapshot) {
+       int snapshot) {
      long size = 0;     // length is zero for directories
      short replication = 0;
      long blocksize = 0;
@@ -2541,7 +2536,7 @@ public class FSDirectory implements Closeable {
    * Create FileStatus with location info by file INode
    */
   private HdfsLocatedFileStatus createLocatedFileStatus(byte[] path,
-      INode node, Snapshot snapshot) throws IOException {
+      INode node, int snapshot) throws IOException {
     assert hasReadLock();
     long size = 0; // length is zero for directories
     short replication = 0;
@@ -2553,7 +2548,7 @@ public class FSDirectory implements Closeable {
       replication = fileNode.getFileReplication(snapshot);
       blocksize = fileNode.getPreferredBlockSize();
 
-      final boolean inSnapshot = snapshot != null; 
+      final boolean inSnapshot = snapshot != Snapshot.CURRENT_STATE_ID; 
       final boolean isUc = inSnapshot ? false : fileNode.isUnderConstruction();
       final long fileSize = !inSnapshot && isUc ? 
           fileNode.computeFileSizeNotIncludingLastUcBlock() : size;

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

@@ -24,14 +24,12 @@ import java.io.IOException;
 import java.io.InputStream;
 import java.util.Arrays;
 import java.util.EnumMap;
-import java.util.EnumSet;
 import java.util.List;
 
 import org.apache.commons.logging.Log;
 import org.apache.commons.logging.LogFactory;
 import org.apache.hadoop.classification.InterfaceAudience;
 import org.apache.hadoop.classification.InterfaceStability;
-import org.apache.hadoop.fs.CacheFlag;
 import org.apache.hadoop.hdfs.protocol.Block;
 import org.apache.hadoop.hdfs.protocol.CacheDirectiveInfo;
 import org.apache.hadoop.hdfs.protocol.HdfsConstants;
@@ -80,6 +78,7 @@ import org.apache.hadoop.hdfs.server.namenode.FSEditLogOp.UpdateBlocksOp;
 import org.apache.hadoop.hdfs.server.namenode.FSEditLogOp.UpdateMasterKeyOp;
 import org.apache.hadoop.hdfs.server.namenode.INode.BlocksMapUpdateInfo;
 import org.apache.hadoop.hdfs.server.namenode.LeaseManager.Lease;
+import org.apache.hadoop.hdfs.server.namenode.snapshot.Snapshot;
 import org.apache.hadoop.hdfs.server.namenode.startupprogress.Phase;
 import org.apache.hadoop.hdfs.server.namenode.startupprogress.StartupProgress;
 import org.apache.hadoop.hdfs.server.namenode.startupprogress.StartupProgress.Counter;
@@ -328,7 +327,7 @@ public class FSEditLogLoader {
         // add the op into retry cache if necessary
         if (toAddRetryCache) {
           HdfsFileStatus stat = fsNamesys.dir.createFileStatus(
-              HdfsFileStatus.EMPTY_NAME, newFile, null);
+              HdfsFileStatus.EMPTY_NAME, newFile, Snapshot.CURRENT_STATE_ID);
           fsNamesys.addCacheEntryWithPayload(addCloseOp.rpcClientId,
               addCloseOp.rpcCallId, stat);
         }
@@ -341,7 +340,7 @@ public class FSEditLogLoader {
           }
           LocatedBlock lb = fsNamesys.prepareFileForWrite(addCloseOp.path,
               oldFile, addCloseOp.clientName, addCloseOp.clientMachine, null,
-              false, iip.getLatestSnapshot(), false);
+              false, iip.getLatestSnapshotId(), false);
           newFile = INodeFile.valueOf(fsDir.getINode(addCloseOp.path),
               addCloseOp.path, true);
           
@@ -357,8 +356,8 @@ public class FSEditLogLoader {
       // update the block list.
       
       // Update the salient file attributes.
-      newFile.setAccessTime(addCloseOp.atime, null);
+      newFile.setAccessTime(addCloseOp.atime, Snapshot.CURRENT_STATE_ID);
-      newFile.setModificationTime(addCloseOp.mtime, null);
+      newFile.setModificationTime(addCloseOp.mtime, Snapshot.CURRENT_STATE_ID);
       updateBlocks(fsDir, addCloseOp, newFile);
       break;
     }
@@ -376,8 +375,8 @@ public class FSEditLogLoader {
       final INodeFile file = INodeFile.valueOf(iip.getINode(0), addCloseOp.path);
 
       // Update the salient file attributes.
-      file.setAccessTime(addCloseOp.atime, null);
+      file.setAccessTime(addCloseOp.atime, Snapshot.CURRENT_STATE_ID);
-      file.setModificationTime(addCloseOp.mtime, null);
+      file.setModificationTime(addCloseOp.mtime, Snapshot.CURRENT_STATE_ID);
       updateBlocks(fsDir, addCloseOp, file);
 
       // Now close the file

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

@@ -53,6 +53,7 @@ import org.apache.hadoop.hdfs.server.common.Util;
 import org.apache.hadoop.hdfs.server.namenode.FSImageStorageInspector.FSImageFile;
 import org.apache.hadoop.hdfs.server.namenode.NNStorage.NameNodeDirType;
 import org.apache.hadoop.hdfs.server.namenode.NNStorage.NameNodeFile;
+import org.apache.hadoop.hdfs.server.namenode.snapshot.Snapshot;
 import org.apache.hadoop.hdfs.server.namenode.startupprogress.Phase;
 import org.apache.hadoop.hdfs.server.namenode.startupprogress.StartupProgress;
 import org.apache.hadoop.hdfs.server.protocol.CheckpointCommand;
@@ -770,7 +771,7 @@ public class FSImage implements Closeable {
 
     dir.computeQuotaUsage4CurrentDirectory(counts);
     
-    for (INode child : dir.getChildrenList(null)) {
+    for (INode child : dir.getChildrenList(Snapshot.CURRENT_STATE_ID)) {
       if (child.isDirectory()) {
         updateCountForQuotaRecursively(child.asDirectory(), counts);
       } else {

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

@@ -1144,7 +1144,8 @@ public class FSImageFormat {
         return;
       }
       
-      final ReadOnlyList<INode> children = current.getChildrenList(null);
+      final ReadOnlyList<INode> children = current
+          .getChildrenList(Snapshot.CURRENT_STATE_ID);
       int dirNum = 0;
       List<INodeDirectory> snapshotDirs = null;
       DirectoryWithSnapshotFeature sf = current.getDirectoryWithSnapshotFeature();

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

@@ -178,7 +178,15 @@ import org.apache.hadoop.hdfs.security.token.block.BlockTokenSecretManager;
 import org.apache.hadoop.hdfs.security.token.block.BlockTokenSecretManager.AccessMode;
 import org.apache.hadoop.hdfs.security.token.delegation.DelegationTokenIdentifier;
 import org.apache.hadoop.hdfs.security.token.delegation.DelegationTokenSecretManager;
-import org.apache.hadoop.hdfs.server.blockmanagement.*;
+import org.apache.hadoop.hdfs.server.blockmanagement.BlockCollection;
+import org.apache.hadoop.hdfs.server.blockmanagement.BlockInfo;
+import org.apache.hadoop.hdfs.server.blockmanagement.BlockInfoUnderConstruction;
+import org.apache.hadoop.hdfs.server.blockmanagement.BlockManager;
+import org.apache.hadoop.hdfs.server.blockmanagement.DatanodeDescriptor;
+import org.apache.hadoop.hdfs.server.blockmanagement.DatanodeManager;
+import org.apache.hadoop.hdfs.server.blockmanagement.DatanodeStatistics;
+import org.apache.hadoop.hdfs.server.blockmanagement.DatanodeStorageInfo;
+import org.apache.hadoop.hdfs.server.blockmanagement.OutOfV1GenerationStampsException;
 import org.apache.hadoop.hdfs.server.common.GenerationStamp;
 import org.apache.hadoop.hdfs.server.common.HdfsServerConstants.BlockUCState;
 import org.apache.hadoop.hdfs.server.common.HdfsServerConstants.NamenodeRole;
@@ -1619,11 +1627,11 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
             if (isReadOp) {
               continue;
             }
-            dir.setTimes(src, inode, -1, now, false, iip.getLatestSnapshot());
+            dir.setTimes(src, inode, -1, now, false, iip.getLatestSnapshotId());
           }
         }
         final long fileSize = iip.isSnapshot() ?
-            inode.computeFileSize(iip.getPathSnapshot())
+            inode.computeFileSize(iip.getPathSnapshotId())
             : inode.computeFileSizeNotIncludingLastUcBlock();
         boolean isUc = inode.isUnderConstruction();
         if (iip.isSnapshot()) {
@@ -1871,7 +1879,7 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
       final INodesInPath iip = dir.getINodesInPath4Write(src);
       final INode inode = iip.getLastINode();
       if (inode != null) {
-        dir.setTimes(src, inode, mtime, atime, true, iip.getLatestSnapshot());
+        dir.setTimes(src, inode, mtime, atime, true, iip.getLatestSnapshotId());
         resultingStat = getAuditFileInfo(src, false);
       } else {
         throw new FileNotFoundException("File/Directory " + src + " does not exist.");
@@ -2267,7 +2275,7 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
       final DatanodeDescriptor clientNode = 
           blockManager.getDatanodeManager().getDatanodeByHost(clientMachine);
       return prepareFileForWrite(src, myFile, holder, clientMachine, clientNode,
-          true, iip.getLatestSnapshot(), logRetryCache);
+          true, iip.getLatestSnapshotId(), logRetryCache);
     } catch (IOException ie) {
       NameNode.stateChangeLog.warn("DIR* NameSystem.append: " +ie.getMessage());
       throw ie;
@@ -2292,7 +2300,7 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
    */
   LocatedBlock prepareFileForWrite(String src, INodeFile file,
       String leaseHolder, String clientMachine, DatanodeDescriptor clientNode,
-      boolean writeToEditLog, Snapshot latestSnapshot, boolean logRetryCache)
+      boolean writeToEditLog, int latestSnapshot, boolean logRetryCache)
       throws IOException {
     file = file.recordModification(latestSnapshot);
     final INodeFile cons = file.toUnderConstruction(leaseHolder, clientMachine,
@@ -2927,7 +2935,7 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
     }
 
     finalizeINodeFileUnderConstruction(src, pendingFile,
-        iip.getLatestSnapshot());
+        iip.getLatestSnapshotId());
     return true;
   }
 
@@ -3636,7 +3644,7 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
     // then reap lease immediately and close the file.
     if(nrCompleteBlocks == nrBlocks) {
       finalizeINodeFileUnderConstruction(src, pendingFile,
-          iip.getLatestSnapshot());
+          iip.getLatestSnapshotId());
       NameNode.stateChangeLog.warn("BLOCK*"
         + " internalReleaseLease: All existing blocks are COMPLETE,"
         + " lease removed, file closed.");
@@ -3685,7 +3693,7 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
       if(penultimateBlockMinReplication &&
           blockManager.checkMinReplication(lastBlock)) {
         finalizeINodeFileUnderConstruction(src, pendingFile,
-            iip.getLatestSnapshot());
+            iip.getLatestSnapshotId());
         NameNode.stateChangeLog.warn("BLOCK*"
           + " internalReleaseLease: Committed blocks are minimally replicated,"
           + " lease removed, file closed.");
@@ -3716,7 +3724,7 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
         // We can remove this block and close the file.
         pendingFile.removeLastBlock(lastBlock);
         finalizeINodeFileUnderConstruction(src, pendingFile,
-            iip.getLatestSnapshot());
+            iip.getLatestSnapshotId());
         NameNode.stateChangeLog.warn("BLOCK* internalReleaseLease: "
             + "Removed empty last block and closed file.");
         return true;
@@ -3777,7 +3785,7 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
   }
 
   private void finalizeINodeFileUnderConstruction(String src,
-      INodeFile pendingFile, Snapshot latestSnapshot) throws IOException,
+      INodeFile pendingFile, int latestSnapshot) throws IOException,
       UnresolvedLinkException {
     assert hasWriteLock();
     FileUnderConstructionFeature uc = pendingFile.getFileUnderConstructionFeature();
@@ -3977,7 +3985,7 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
 
     //remove lease, close file
     finalizeINodeFileUnderConstruction(src, pendingFile,
-        Snapshot.findLatestSnapshot(pendingFile, null));
+        Snapshot.findLatestSnapshot(pendingFile, Snapshot.CURRENT_STATE_ID));
 
     return src;
   }
@@ -7131,7 +7139,7 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
         getEditLog().logSync();
       }
       if (isAuditEnabled() && isExternalInvocation()) {
-        logAuditEvent(success, "addCacheDirective", null, null, null);
+        logAuditEvent(success, "modifyCacheDirective", null, null, null);
       }
       RetryCache.setState(cacheEntry, success);
     }

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

@@ -17,7 +17,6 @@
  */
 package org.apache.hadoop.hdfs.server.namenode;
 
-import java.io.IOException;
 import java.util.Arrays;
 import java.util.Collections;
 import java.util.HashSet;
@@ -29,7 +28,6 @@ import org.apache.commons.logging.LogFactory;
 import org.apache.hadoop.fs.UnresolvedLinkException;
 import org.apache.hadoop.fs.permission.FsAction;
 import org.apache.hadoop.fs.permission.FsPermission;
-import org.apache.hadoop.hdfs.server.namenode.snapshot.Snapshot;
 import org.apache.hadoop.security.AccessControlException;
 import org.apache.hadoop.security.UserGroupInformation;
 
@@ -142,54 +140,54 @@ class FSPermissionChecker {
     // check if (parentAccess != null) && file exists, then check sb
     // If resolveLink, the check is performed on the link target.
     final INodesInPath inodesInPath = root.getINodesInPath(path, resolveLink);
-    final Snapshot snapshot = inodesInPath.getPathSnapshot();
+    final int snapshotId = inodesInPath.getPathSnapshotId();
     final INode[] inodes = inodesInPath.getINodes();
     int ancestorIndex = inodes.length - 2;
     for(; ancestorIndex >= 0 && inodes[ancestorIndex] == null;
         ancestorIndex--);
-    checkTraverse(inodes, ancestorIndex, snapshot);
+    checkTraverse(inodes, ancestorIndex, snapshotId);
 
     final INode last = inodes[inodes.length - 1];
     if (parentAccess != null && parentAccess.implies(FsAction.WRITE)
         && inodes.length > 1 && last != null) {
-      checkStickyBit(inodes[inodes.length - 2], last, snapshot);
+      checkStickyBit(inodes[inodes.length - 2], last, snapshotId);
     }
     if (ancestorAccess != null && inodes.length > 1) {
-      check(inodes, ancestorIndex, snapshot, ancestorAccess);
+      check(inodes, ancestorIndex, snapshotId, ancestorAccess);
     }
     if (parentAccess != null && inodes.length > 1) {
-      check(inodes, inodes.length - 2, snapshot, parentAccess);
+      check(inodes, inodes.length - 2, snapshotId, parentAccess);
     }
     if (access != null) {
-      check(last, snapshot, access);
+      check(last, snapshotId, access);
     }
     if (subAccess != null) {
-      checkSubAccess(last, snapshot, subAccess);
+      checkSubAccess(last, snapshotId, subAccess);
     }
     if (doCheckOwner) {
-      checkOwner(last, snapshot);
+      checkOwner(last, snapshotId);
     }
   }
 
   /** Guarded by {@link FSNamesystem#readLock()} */
-  private void checkOwner(INode inode, Snapshot snapshot
+  private void checkOwner(INode inode, int snapshotId
       ) throws AccessControlException {
-    if (inode != null && user.equals(inode.getUserName(snapshot))) {
+    if (inode != null && user.equals(inode.getUserName(snapshotId))) {
       return;
     }
     throw new AccessControlException("Permission denied");
   }
 
   /** Guarded by {@link FSNamesystem#readLock()} */
-  private void checkTraverse(INode[] inodes, int last, Snapshot snapshot
+  private void checkTraverse(INode[] inodes, int last, int snapshotId
       ) throws AccessControlException {
     for(int j = 0; j <= last; j++) {
-      check(inodes[j], snapshot, FsAction.EXECUTE);
+      check(inodes[j], snapshotId, FsAction.EXECUTE);
     }
   }
 
   /** Guarded by {@link FSNamesystem#readLock()} */
-  private void checkSubAccess(INode inode, Snapshot snapshot, FsAction access
+  private void checkSubAccess(INode inode, int snapshotId, FsAction access
       ) throws AccessControlException {
     if (inode == null || !inode.isDirectory()) {
       return;
@@ -198,9 +196,9 @@ class FSPermissionChecker {
     Stack<INodeDirectory> directories = new Stack<INodeDirectory>();
     for(directories.push(inode.asDirectory()); !directories.isEmpty(); ) {
       INodeDirectory d = directories.pop();
-      check(d, snapshot, access);
+      check(d, snapshotId, access);
 
-      for(INode child : d.getChildrenList(snapshot)) {
+      for(INode child : d.getChildrenList(snapshotId)) {
         if (child.isDirectory()) {
           directories.push(child.asDirectory());
         }
@@ -209,23 +207,23 @@ class FSPermissionChecker {
   }
 
   /** Guarded by {@link FSNamesystem#readLock()} */
-  private void check(INode[] inodes, int i, Snapshot snapshot, FsAction access
+  private void check(INode[] inodes, int i, int snapshotId, FsAction access
       ) throws AccessControlException {
-    check(i >= 0? inodes[i]: null, snapshot, access);
+    check(i >= 0? inodes[i]: null, snapshotId, access);
   }
 
   /** Guarded by {@link FSNamesystem#readLock()} */
-  private void check(INode inode, Snapshot snapshot, FsAction access
+  private void check(INode inode, int snapshotId, FsAction access
       ) throws AccessControlException {
     if (inode == null) {
       return;
     }
-    FsPermission mode = inode.getFsPermission(snapshot);
+    FsPermission mode = inode.getFsPermission(snapshotId);
 
-    if (user.equals(inode.getUserName(snapshot))) { //user class
+    if (user.equals(inode.getUserName(snapshotId))) { //user class
       if (mode.getUserAction().implies(access)) { return; }
     }
-    else if (groups.contains(inode.getGroupName(snapshot))) { //group class
+    else if (groups.contains(inode.getGroupName(snapshotId))) { //group class
       if (mode.getGroupAction().implies(access)) { return; }
     }
     else { //other class
@@ -236,19 +234,19 @@ class FSPermissionChecker {
   }
 
   /** Guarded by {@link FSNamesystem#readLock()} */
-  private void checkStickyBit(INode parent, INode inode, Snapshot snapshot
+  private void checkStickyBit(INode parent, INode inode, int snapshotId
       ) throws AccessControlException {
-    if(!parent.getFsPermission(snapshot).getStickyBit()) {
+    if(!parent.getFsPermission(snapshotId).getStickyBit()) {
       return;
     }
 
     // If this user is the directory owner, return
-    if(parent.getUserName(snapshot).equals(user)) {
+    if(parent.getUserName(snapshotId).equals(user)) {
       return;
     }
 
     // if this user is the file owner, return
-    if(inode.getUserName(snapshot).equals(user)) {
+    if(inode.getUserName(snapshotId).equals(user)) {
       return;
     }
 

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

@@ -70,98 +70,101 @@ public abstract class INode implements INodeAttributes, Diff.Element<byte[]> {
   }
 
   /** Get the {@link PermissionStatus} */
-  abstract PermissionStatus getPermissionStatus(Snapshot snapshot);
+  abstract PermissionStatus getPermissionStatus(int snapshotId);
 
   /** The same as getPermissionStatus(null). */
   final PermissionStatus getPermissionStatus() {
-    return getPermissionStatus(null);
+    return getPermissionStatus(Snapshot.CURRENT_STATE_ID);
   }
 
   /**
-   * @param snapshot
+   * @param snapshotId
-   *          if it is not null, get the result from the given snapshot;
+   *          if it is not {@link Snapshot#CURRENT_STATE_ID}, get the result
-   *          otherwise, get the result from the current inode.
+   *          from the given snapshot; otherwise, get the result from the
+   *          current inode.
    * @return user name
    */
-  abstract String getUserName(Snapshot snapshot);
+  abstract String getUserName(int snapshotId);
 
-  /** The same as getUserName(null). */
+  /** The same as getUserName(Snapshot.CURRENT_STATE_ID). */
   @Override
   public final String getUserName() {
-    return getUserName(null);
+    return getUserName(Snapshot.CURRENT_STATE_ID);
   }
 
   /** Set user */
   abstract void setUser(String user);
 
   /** Set user */
-  final INode setUser(String user, Snapshot latest)
+  final INode setUser(String user, int latestSnapshotId)
       throws QuotaExceededException {
-    final INode nodeToUpdate = recordModification(latest);
+    final INode nodeToUpdate = recordModification(latestSnapshotId);
     nodeToUpdate.setUser(user);
     return nodeToUpdate;
   }
   /**
-   * @param snapshot
+   * @param snapshotId
-   *          if it is not null, get the result from the given snapshot;
+   *          if it is not {@link Snapshot#CURRENT_STATE_ID}, get the result
-   *          otherwise, get the result from the current inode.
+   *          from the given snapshot; otherwise, get the result from the
+   *          current inode.
    * @return group name
    */
-  abstract String getGroupName(Snapshot snapshot);
+  abstract String getGroupName(int snapshotId);
 
-  /** The same as getGroupName(null). */
+  /** The same as getGroupName(Snapshot.CURRENT_STATE_ID). */
   @Override
   public final String getGroupName() {
-    return getGroupName(null);
+    return getGroupName(Snapshot.CURRENT_STATE_ID);
   }
 
   /** Set group */
   abstract void setGroup(String group);
 
   /** Set group */
-  final INode setGroup(String group, Snapshot latest)
+  final INode setGroup(String group, int latestSnapshotId)
       throws QuotaExceededException {
-    final INode nodeToUpdate = recordModification(latest);
+    final INode nodeToUpdate = recordModification(latestSnapshotId);
     nodeToUpdate.setGroup(group);
     return nodeToUpdate;
   }
 
   /**
-   * @param snapshot
+   * @param snapshotId
-   *          if it is not null, get the result from the given snapshot;
+   *          if it is not {@link Snapshot#CURRENT_STATE_ID}, get the result
-   *          otherwise, get the result from the current inode.
+   *          from the given snapshot; otherwise, get the result from the
+   *          current inode.
    * @return permission.
    */
-  abstract FsPermission getFsPermission(Snapshot snapshot);
+  abstract FsPermission getFsPermission(int snapshotId);
   
-  /** The same as getFsPermission(null). */
+  /** The same as getFsPermission(Snapshot.CURRENT_STATE_ID). */
   @Override
   public final FsPermission getFsPermission() {
-    return getFsPermission(null);
+    return getFsPermission(Snapshot.CURRENT_STATE_ID);
   }
 
   /** Set the {@link FsPermission} of this {@link INode} */
   abstract void setPermission(FsPermission permission);
 
   /** Set the {@link FsPermission} of this {@link INode} */
-  INode setPermission(FsPermission permission, Snapshot latest) 
+  INode setPermission(FsPermission permission, int latestSnapshotId) 
       throws QuotaExceededException {
-    final INode nodeToUpdate = recordModification(latest);
+    final INode nodeToUpdate = recordModification(latestSnapshotId);
     nodeToUpdate.setPermission(permission);
     return nodeToUpdate;
   }
 
   /**
-   * @return if the given snapshot is null, return this;
+   * @return if the given snapshot id is {@link Snapshot#CURRENT_STATE_ID},
-   *     otherwise return the corresponding snapshot inode.
+   *         return this; otherwise return the corresponding snapshot inode.
    */
-  public INodeAttributes getSnapshotINode(final Snapshot snapshot) {
+  public INodeAttributes getSnapshotINode(final int snapshotId) {
     return this;
   }
 
   /** Is this inode in the latest snapshot? */
-  public final boolean isInLatestSnapshot(final Snapshot latest) {
+  public final boolean isInLatestSnapshot(final int latestSnapshotId) {
-    if (latest == null) {
+    if (latestSnapshotId == Snapshot.CURRENT_STATE_ID) {
       return false;
     }
     // if parent is a reference node, parent must be a renamed node. We can 
@@ -173,10 +176,11 @@ public abstract class INode implements INodeAttributes, Diff.Element<byte[]> {
     if (parentDir == null) { // root
       return true;
     }
-    if (!parentDir.isInLatestSnapshot(latest)) {
+    if (!parentDir.isInLatestSnapshot(latestSnapshotId)) {
       return false;
     }
-    final INode child = parentDir.getChild(getLocalNameBytes(), latest);
+    final INode child = parentDir.getChild(getLocalNameBytes(),
+        latestSnapshotId);
     if (this == child) {
       return true;
     }
@@ -203,21 +207,22 @@ public abstract class INode implements INodeAttributes, Diff.Element<byte[]> {
    * operation, or the snapshot belonging to the DST tree.
    * 
    * @param latestInDst
-   *          the latest snapshot in the DST tree above the reference node
+   *          id of the latest snapshot in the DST tree above the reference node
    * @return True: the modification should be recorded in the snapshot that
    *         belongs to the SRC tree. False: the modification should be
    *         recorded in the snapshot that belongs to the DST tree.
    */
-  public final boolean shouldRecordInSrcSnapshot(final Snapshot latestInDst) {
+  public final boolean shouldRecordInSrcSnapshot(final int latestInDst) {
     Preconditions.checkState(!isReference());
 
-    if (latestInDst == null) {
+    if (latestInDst == Snapshot.CURRENT_STATE_ID) {
       return true;
     }
     INodeReference withCount = getParentReference();
     if (withCount != null) {
       int dstSnapshotId = withCount.getParentReference().getDstSnapshotId();
-      if (dstSnapshotId >= latestInDst.getId()) {
+      if (dstSnapshotId != Snapshot.CURRENT_STATE_ID
+          && dstSnapshotId >= latestInDst) {
         return true;
       }
     }
@@ -228,13 +233,14 @@ public abstract class INode implements INodeAttributes, Diff.Element<byte[]> {
    * This inode is being modified.  The previous version of the inode needs to
    * be recorded in the latest snapshot.
    *
-   * @param latest the latest snapshot that has been taken.
+   * @param latestSnapshotId The id of the latest snapshot that has been taken.
-   *        Note that it is null if no snapshots have been taken.
+   *                         Note that it is {@link Snapshot#CURRENT_STATE_ID} 
+   *                         if no snapshots have been taken.
    * @return The current inode, which usually is the same object of this inode.
    *         However, in some cases, this inode may be replaced with a new inode
    *         for maintaining snapshots. The current inode is then the new inode.
    */
-  abstract INode recordModification(final Snapshot latest)
+  abstract INode recordModification(final int latestSnapshotId)
       throws QuotaExceededException;
 
   /** Check whether it's a reference. */
@@ -330,12 +336,13 @@ public abstract class INode implements INodeAttributes, Diff.Element<byte[]> {
    * snapshot in its diff list. Recursively clean its children.
    * </pre>
    * 
-   * @param snapshot
+   * @param snapshotId
-   *          The snapshot to delete. Null means to delete the current
+   *          The id of the snapshot to delete. 
+   *          {@link Snapshot#CURRENT_STATE_ID} means to delete the current
    *          file/directory.
-   * @param prior
+   * @param priorSnapshotId
-   *          The latest snapshot before the to-be-deleted snapshot. When
+   *          The id of the latest snapshot before the to-be-deleted snapshot.
-   *          deleting a current inode, this parameter captures the latest
+   *          When deleting a current inode, this parameter captures the latest
    *          snapshot.
    * @param collectedBlocks
    *          blocks collected from the descents for further block
@@ -345,8 +352,8 @@ public abstract class INode implements INodeAttributes, Diff.Element<byte[]> {
    *          inodeMap
    * @return quota usage delta when deleting a snapshot
    */
-  public abstract Quota.Counts cleanSubtree(final Snapshot snapshot,
+  public abstract Quota.Counts cleanSubtree(final int snapshotId,
-      Snapshot prior, BlocksMapUpdateInfo collectedBlocks,
+      int priorSnapshotId, BlocksMapUpdateInfo collectedBlocks,
       List<INode> removedINodes, boolean countDiffChange)
       throws QuotaExceededException;
   
@@ -460,9 +467,10 @@ public abstract class INode implements INodeAttributes, Diff.Element<byte[]> {
    * @param counts The subtree counts for returning.
    * @param useCache Whether to use cached quota usage. Note that 
    *                 {@link WithName} node never uses cache for its subtree.
-   * @param lastSnapshotId {@link Snapshot#INVALID_ID} indicates the computation
+   * @param lastSnapshotId {@link Snapshot#CURRENT_STATE_ID} indicates the 
-   *                       is in the current tree. Otherwise the id indicates
+   *                       computation is in the current tree. Otherwise the id
-   *                       the computation range for a {@link WithName} node.
+   *                       indicates the computation range for a 
+   *                       {@link WithName} node.
    * @return The same objects as the counts parameter.
    */
   public abstract Quota.Counts computeQuotaUsage(Quota.Counts counts,
@@ -470,7 +478,7 @@ public abstract class INode implements INodeAttributes, Diff.Element<byte[]> {
 
   public final Quota.Counts computeQuotaUsage(Quota.Counts counts,
       boolean useCache) {
-    return computeQuotaUsage(counts, useCache, Snapshot.INVALID_ID);
+    return computeQuotaUsage(counts, useCache, Snapshot.CURRENT_STATE_ID);
   }
   
   /**
@@ -558,21 +566,22 @@ public abstract class INode implements INodeAttributes, Diff.Element<byte[]> {
   }
 
   /**
-   * @param snapshot
+   * @param snapshotId
-   *          if it is not null, get the result from the given snapshot;
+   *          if it is not {@link Snapshot#CURRENT_STATE_ID}, get the result
-   *          otherwise, get the result from the current inode.
+   *          from the given snapshot; otherwise, get the result from the
+   *          current inode.
    * @return modification time.
    */
-  abstract long getModificationTime(Snapshot snapshot);
+  abstract long getModificationTime(int snapshotId);
 
-  /** The same as getModificationTime(null). */
+  /** The same as getModificationTime(Snapshot.CURRENT_STATE_ID). */
   @Override
   public final long getModificationTime() {
-    return getModificationTime(null);
+    return getModificationTime(Snapshot.CURRENT_STATE_ID);
   }
 
   /** Update modification time if it is larger than the current value. */
-  public abstract INode updateModificationTime(long mtime, Snapshot latest) 
+  public abstract INode updateModificationTime(long mtime, int latestSnapshotId) 
       throws QuotaExceededException;
 
   /** Set the last modification time of inode. */
@@ -580,24 +589,25 @@ public abstract class INode implements INodeAttributes, Diff.Element<byte[]> {
 
   /** Set the last modification time of inode. */
   public final INode setModificationTime(long modificationTime,
-      Snapshot latest) throws QuotaExceededException {
+      int latestSnapshotId) throws QuotaExceededException {
-    final INode nodeToUpdate = recordModification(latest);
+    final INode nodeToUpdate = recordModification(latestSnapshotId);
     nodeToUpdate.setModificationTime(modificationTime);
     return nodeToUpdate;
   }
 
   /**
-   * @param snapshot
+   * @param snapshotId
-   *          if it is not null, get the result from the given snapshot;
+   *          if it is not {@link Snapshot#CURRENT_STATE_ID}, get the result
-   *          otherwise, get the result from the current inode.
+   *          from the given snapshot; otherwise, get the result from the
+   *          current inode.
    * @return access time
    */
-  abstract long getAccessTime(Snapshot snapshot);
+  abstract long getAccessTime(int snapshotId);
 
-  /** The same as getAccessTime(null). */
+  /** The same as getAccessTime(Snapshot.CURRENT_STATE_ID). */
   @Override
   public final long getAccessTime() {
-    return getAccessTime(null);
+    return getAccessTime(Snapshot.CURRENT_STATE_ID);
   }
 
   /**
@@ -608,9 +618,9 @@ public abstract class INode implements INodeAttributes, Diff.Element<byte[]> {
   /**
    * Set last access time of inode.
    */
-  public final INode setAccessTime(long accessTime, Snapshot latest)
+  public final INode setAccessTime(long accessTime, int latestSnapshotId)
       throws QuotaExceededException {
-    final INode nodeToUpdate = recordModification(latest);
+    final INode nodeToUpdate = recordModification(latestSnapshotId);
     nodeToUpdate.setAccessTime(accessTime);
     return nodeToUpdate;
   }
@@ -679,13 +689,15 @@ public abstract class INode implements INodeAttributes, Diff.Element<byte[]> {
   @VisibleForTesting
   public final StringBuffer dumpTreeRecursively() {
     final StringWriter out = new StringWriter(); 
-    dumpTreeRecursively(new PrintWriter(out, true), new StringBuilder(), null);
+    dumpTreeRecursively(new PrintWriter(out, true), new StringBuilder(),
+        Snapshot.CURRENT_STATE_ID);
     return out.getBuffer();
   }
 
   @VisibleForTesting
   public final void dumpTreeRecursively(PrintStream out) {
-    dumpTreeRecursively(new PrintWriter(out, true), new StringBuilder(), null);
+    dumpTreeRecursively(new PrintWriter(out, true), new StringBuilder(),
+        Snapshot.CURRENT_STATE_ID);
   }
 
   /**
@@ -694,7 +706,7 @@ public abstract class INode implements INodeAttributes, Diff.Element<byte[]> {
    */
   @VisibleForTesting
   public void dumpTreeRecursively(PrintWriter out, StringBuilder prefix,
-      Snapshot snapshot) {
+      int snapshotId) {
     out.print(prefix);
     out.print(" ");
     final String name = getLocalName();
@@ -703,7 +715,7 @@ public abstract class INode implements INodeAttributes, Diff.Element<byte[]> {
     out.print(getObjectString());
     out.print("), ");
     out.print(getParentString());
-    out.print(", " + getPermissionStatus(snapshot));
+    out.print(", " + getPermissionStatus(snapshotId));
   }
   
   /**

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

@@ -204,9 +204,9 @@ public class INodeDirectory extends INodeWithAdditionalFields
   }
   
   @Override
-  public INodeDirectoryAttributes getSnapshotINode(Snapshot snapshot) {
+  public INodeDirectoryAttributes getSnapshotINode(int snapshotId) {
     DirectoryWithSnapshotFeature sf = getDirectoryWithSnapshotFeature();
-    return sf == null ? this : sf.getDiffs().getSnapshotINode(snapshot, this);
+    return sf == null ? this : sf.getDiffs().getSnapshotINode(snapshotId, this);
   }
   
   @Override
@@ -217,12 +217,13 @@ public class INodeDirectory extends INodeWithAdditionalFields
 
   /** Replace itself with an {@link INodeDirectorySnapshottable}. */
   public INodeDirectorySnapshottable replaceSelf4INodeDirectorySnapshottable(
-      Snapshot latest, final INodeMap inodeMap) throws QuotaExceededException {
+      int latestSnapshotId, final INodeMap inodeMap)
+      throws QuotaExceededException {
     Preconditions.checkState(!(this instanceof INodeDirectorySnapshottable),
         "this is already an INodeDirectorySnapshottable, this=%s", this);
     final INodeDirectorySnapshottable s = new INodeDirectorySnapshottable(this);
     replaceSelf(s, inodeMap).getDirectoryWithSnapshotFeature().getDiffs()
-        .saveSelf2Snapshot(latest, s, this);
+        .saveSelf2Snapshot(latestSnapshotId, s, this);
     return s;
   }
 
@@ -289,8 +290,8 @@ public class INodeDirectory extends INodeWithAdditionalFields
   }
 
   INodeReference.WithName replaceChild4ReferenceWithName(INode oldChild,
-      Snapshot latest) {
+      int latestSnapshotId) {
-    Preconditions.checkArgument(latest != null);
+    Preconditions.checkArgument(latestSnapshotId != Snapshot.CURRENT_STATE_ID);
     if (oldChild instanceof INodeReference.WithName) {
       return (INodeReference.WithName)oldChild;
     }
@@ -304,22 +305,23 @@ public class INodeDirectory extends INodeWithAdditionalFields
       withCount = new INodeReference.WithCount(null, oldChild);
     }
     final INodeReference.WithName ref = new INodeReference.WithName(this,
-        withCount, oldChild.getLocalNameBytes(), latest.getId());
+        withCount, oldChild.getLocalNameBytes(), latestSnapshotId);
     replaceChild(oldChild, ref, null);
     return ref;
   }
 
   @Override
-  public INodeDirectory recordModification(Snapshot latest) 
+  public INodeDirectory recordModification(int latestSnapshotId) 
       throws QuotaExceededException {
-    if (isInLatestSnapshot(latest) && !shouldRecordInSrcSnapshot(latest)) {
+    if (isInLatestSnapshot(latestSnapshotId)
+        && !shouldRecordInSrcSnapshot(latestSnapshotId)) {
       // add snapshot feature if necessary
       DirectoryWithSnapshotFeature sf = getDirectoryWithSnapshotFeature();
       if (sf == null) {
         sf = addSnapshotFeature(null);
       }
       // record self in the diff list if necessary
-      sf.getDiffs().saveSelf2Snapshot(latest, this, null);
+      sf.getDiffs().saveSelf2Snapshot(latestSnapshotId, this, null);
     }
     return this;
   }
@@ -329,9 +331,9 @@ public class INodeDirectory extends INodeWithAdditionalFields
    * 
    * @return the child inode, which may be replaced.
    */
-  public INode saveChild2Snapshot(final INode child, final Snapshot latest,
+  public INode saveChild2Snapshot(final INode child, final int latestSnapshotId,
       final INode snapshotCopy) throws QuotaExceededException {
-    if (latest == null) {
+    if (latestSnapshotId == Snapshot.CURRENT_STATE_ID) {
       return child;
     }
     
@@ -340,42 +342,45 @@ public class INodeDirectory extends INodeWithAdditionalFields
     if (sf == null) {
       sf = this.addSnapshotFeature(null);
     }
-    return sf.saveChild2Snapshot(this, child, latest, snapshotCopy);
+    return sf.saveChild2Snapshot(this, child, latestSnapshotId, snapshotCopy);
   }
 
   /**
    * @param name the name of the child
-   * @param snapshot
+   * @param snapshotId
-   *          if it is not null, get the result from the given snapshot;
+   *          if it is not {@link Snapshot#CURRENT_STATE_ID}, get the result
-   *          otherwise, get the result from the current directory.
+   *          from the corresponding snapshot; otherwise, get the result from
+   *          the current directory.
    * @return the child inode.
    */
-  public INode getChild(byte[] name, Snapshot snapshot) {
+  public INode getChild(byte[] name, int snapshotId) {
     DirectoryWithSnapshotFeature sf;
-    if (snapshot == null || (sf = getDirectoryWithSnapshotFeature()) == null) {
+    if (snapshotId == Snapshot.CURRENT_STATE_ID || 
+        (sf = getDirectoryWithSnapshotFeature()) == null) {
       ReadOnlyList<INode> c = getCurrentChildrenList();
       final int i = ReadOnlyList.Util.binarySearch(c, name);
       return i < 0 ? null : c.get(i);
     }
     
-    return sf.getChild(this, name, snapshot);
+    return sf.getChild(this, name, snapshotId);
   }
   
   /**
-   * @param snapshot
+   * @param snapshotId
-   *          if it is not null, get the result from the given snapshot;
+   *          if it is not {@link Snapshot#CURRENT_STATE_ID}, get the result
-   *          otherwise, get the result from the current directory.
+   *          from the corresponding snapshot; otherwise, get the result from
+   *          the current directory.
    * @return the current children list if the specified snapshot is null;
    *         otherwise, return the children list corresponding to the snapshot.
    *         Note that the returned list is never null.
    */
-  public ReadOnlyList<INode> getChildrenList(final Snapshot snapshot) {
+  public ReadOnlyList<INode> getChildrenList(final int snapshotId) {
     DirectoryWithSnapshotFeature sf;
-    if (snapshot == null
+    if (snapshotId == Snapshot.CURRENT_STATE_ID
         || (sf = this.getDirectoryWithSnapshotFeature()) == null) {
       return getCurrentChildrenList();
     }
-    return sf.getChildrenList(this, snapshot);
+    return sf.getChildrenList(this, snapshotId);
   }
   
   private ReadOnlyList<INode> getCurrentChildrenList() {
@@ -450,15 +455,15 @@ public class INodeDirectory extends INodeWithAdditionalFields
   /**
    * Remove the specified child from this directory.
    */
-  public boolean removeChild(INode child, Snapshot latest)
+  public boolean removeChild(INode child, int latestSnapshotId)
       throws QuotaExceededException {
-    if (isInLatestSnapshot(latest)) {
+    if (isInLatestSnapshot(latestSnapshotId)) {
       // create snapshot feature if necessary
       DirectoryWithSnapshotFeature sf = this.getDirectoryWithSnapshotFeature();
       if (sf == null) {
         sf = this.addSnapshotFeature(null);
       }
-      return sf.removeChild(this, child, latest);
+      return sf.removeChild(this, child, latestSnapshotId);
     }
     return removeChild(child);
   }
@@ -493,24 +498,24 @@ public class INodeDirectory extends INodeWithAdditionalFields
    *         otherwise, return true;
    */
   public boolean addChild(INode node, final boolean setModTime,
-      final Snapshot latest) throws QuotaExceededException {
+      final int latestSnapshotId) throws QuotaExceededException {
     final int low = searchChildren(node.getLocalNameBytes());
     if (low >= 0) {
       return false;
     }
 
-    if (isInLatestSnapshot(latest)) {
+    if (isInLatestSnapshot(latestSnapshotId)) {
       // create snapshot feature if necessary
       DirectoryWithSnapshotFeature sf = this.getDirectoryWithSnapshotFeature();
       if (sf == null) {
         sf = this.addSnapshotFeature(null);
       }
-      return sf.addChild(this, node, setModTime, latest);
+      return sf.addChild(this, node, setModTime, latestSnapshotId);
     }
     addChild(node, low);
     if (setModTime) {
       // update modification time of the parent directory
-      updateModificationTime(node.getModificationTime(), latest);
+      updateModificationTime(node.getModificationTime(), latestSnapshotId);
     }
     return true;
   }
@@ -548,10 +553,9 @@ public class INodeDirectory extends INodeWithAdditionalFields
     // we are computing the quota usage for a specific snapshot here, i.e., the
     // computation only includes files/directories that exist at the time of the
     // given snapshot
-    if (sf != null && lastSnapshotId != Snapshot.INVALID_ID
+    if (sf != null && lastSnapshotId != Snapshot.CURRENT_STATE_ID
         && !(useCache && isQuotaSet())) {
-      Snapshot lastSnapshot = sf.getDiffs().getSnapshotById(lastSnapshotId);
+      ReadOnlyList<INode> childrenList = getChildrenList(lastSnapshotId);
-      ReadOnlyList<INode> childrenList = getChildrenList(lastSnapshot);
       for (INode child : childrenList) {
         child.computeQuotaUsage(counts, useCache, lastSnapshotId);
       }
@@ -607,7 +611,7 @@ public class INodeDirectory extends INodeWithAdditionalFields
 
   ContentSummaryComputationContext computeDirectoryContentSummary(
       ContentSummaryComputationContext summary) {
-    ReadOnlyList<INode> childrenList = getChildrenList(null);
+    ReadOnlyList<INode> childrenList = getChildrenList(Snapshot.CURRENT_STATE_ID);
     // Explicit traversing is done to enable repositioning after relinquishing
     // and reacquiring locks.
     for (int i = 0;  i < childrenList.size(); i++) {
@@ -629,7 +633,7 @@ public class INodeDirectory extends INodeWithAdditionalFields
         break;
       }
       // Obtain the children list again since it may have been modified.
-      childrenList = getChildrenList(null);
+      childrenList = getChildrenList(Snapshot.CURRENT_STATE_ID);
       // Reposition in case the children list is changed. Decrement by 1
       // since it will be incremented when loops.
       i = nextChild(childrenList, childName) - 1;
@@ -668,21 +672,16 @@ public class INodeDirectory extends INodeWithAdditionalFields
    *          The reference node to be removed/replaced
    * @param newChild
    *          The node to be added back
-   * @param latestSnapshot
-   *          The latest snapshot. Note this may not be the last snapshot in the
-   *          diff list, since the src tree of the current rename operation
-   *          may be the dst tree of a previous rename.
    * @throws QuotaExceededException should not throw this exception
    */
   public void undoRename4ScrParent(final INodeReference oldChild,
-      final INode newChild, Snapshot latestSnapshot)
+      final INode newChild) throws QuotaExceededException {
-      throws QuotaExceededException {
     DirectoryWithSnapshotFeature sf = getDirectoryWithSnapshotFeature();
     Preconditions.checkState(sf != null,
         "Directory does not have snapshot feature");
     sf.getDiffs().removeChild(ListType.DELETED, oldChild);
     sf.getDiffs().replaceChild(ListType.CREATED, oldChild, newChild);
-    addChild(newChild, true, null);
+    addChild(newChild, true, Snapshot.CURRENT_STATE_ID);
   }
   
   /**
@@ -691,16 +690,14 @@ public class INodeDirectory extends INodeWithAdditionalFields
    * and delete possible record in the deleted list.  
    */
   public void undoRename4DstParent(final INode deletedChild,
-      Snapshot latestSnapshot) throws QuotaExceededException {
+      int latestSnapshotId) throws QuotaExceededException {
     DirectoryWithSnapshotFeature sf = getDirectoryWithSnapshotFeature();
     Preconditions.checkState(sf != null,
         "Directory does not have snapshot feature");
     boolean removeDeletedChild = sf.getDiffs().removeChild(ListType.DELETED,
         deletedChild);
-    // pass null for inodeMap since the parent node will not get replaced when
+    int sid = removeDeletedChild ? Snapshot.CURRENT_STATE_ID : latestSnapshotId;
-    // undoing rename
+    final boolean added = addChild(deletedChild, true, sid);
-    final boolean added = addChild(deletedChild, true, removeDeletedChild ? null
-        : latestSnapshot);
     // update quota usage if adding is successfully and the old child has not
     // been stored in deleted list before
     if (added && !removeDeletedChild) {
@@ -722,8 +719,8 @@ public class INodeDirectory extends INodeWithAdditionalFields
   }
 
   /** Call cleanSubtree(..) recursively down the subtree. */
-  public Quota.Counts cleanSubtreeRecursively(final Snapshot snapshot,
+  public Quota.Counts cleanSubtreeRecursively(final int snapshot,
-      Snapshot prior, final BlocksMapUpdateInfo collectedBlocks,
+      int prior, final BlocksMapUpdateInfo collectedBlocks,
       final List<INode> removedINodes, final Map<INode, INode> excludedNodes, 
       final boolean countDiffChange) throws QuotaExceededException {
     Quota.Counts counts = Quota.Counts.newInstance();
@@ -732,9 +729,10 @@ public class INodeDirectory extends INodeWithAdditionalFields
     // to its latest previous snapshot. (besides, we also need to consider nodes
     // created after prior but before snapshot. this will be done in 
     // DirectoryWithSnapshotFeature)
-    Snapshot s = snapshot != null && prior != null ? prior : snapshot;
+    int s = snapshot != Snapshot.CURRENT_STATE_ID
+        && prior != Snapshot.NO_SNAPSHOT_ID ? prior : snapshot;
     for (INode child : getChildrenList(s)) {
-      if (snapshot != null && excludedNodes != null
+      if (snapshot != Snapshot.CURRENT_STATE_ID && excludedNodes != null
           && excludedNodes.containsKey(child)) {
         continue;
       } else {
@@ -753,7 +751,7 @@ public class INodeDirectory extends INodeWithAdditionalFields
     if (sf != null) {
       sf.clear(this, collectedBlocks, removedINodes);
     }
-    for (INode child : getChildrenList(null)) {
+    for (INode child : getChildrenList(Snapshot.CURRENT_STATE_ID)) {
       child.destroyAndCollectBlocks(collectedBlocks, removedINodes);
     }
     clear();
@@ -761,18 +759,19 @@ public class INodeDirectory extends INodeWithAdditionalFields
   }
   
   @Override
-  public Quota.Counts cleanSubtree(final Snapshot snapshot, Snapshot prior,
+  public Quota.Counts cleanSubtree(final int snapshotId, int priorSnapshotId,
       final BlocksMapUpdateInfo collectedBlocks,
       final List<INode> removedINodes, final boolean countDiffChange)
       throws QuotaExceededException {
     DirectoryWithSnapshotFeature sf = getDirectoryWithSnapshotFeature();
     // there is snapshot data
     if (sf != null) {
-      return sf.cleanDirectory(this, snapshot, prior, collectedBlocks,
+      return sf.cleanDirectory(this, snapshotId, priorSnapshotId,
-          removedINodes, countDiffChange);
+          collectedBlocks, removedINodes, countDiffChange);
     }
     // there is no snapshot data
-    if (prior == null && snapshot == null) {
+    if (priorSnapshotId == Snapshot.NO_SNAPSHOT_ID
+        && snapshotId == Snapshot.CURRENT_STATE_ID) {
       // destroy the whole subtree and collect blocks that should be deleted
       Quota.Counts counts = Quota.Counts.newInstance();
       this.computeQuotaUsage(counts, true);
@@ -780,7 +779,7 @@ public class INodeDirectory extends INodeWithAdditionalFields
       return counts; 
     } else {
       // process recursively down the subtree
-      Quota.Counts counts = cleanSubtreeRecursively(snapshot, prior,
+      Quota.Counts counts = cleanSubtreeRecursively(snapshotId, priorSnapshotId,
           collectedBlocks, removedINodes, null, countDiffChange);
       if (isQuotaSet()) {
         getDirectoryWithQuotaFeature().addSpaceConsumed2Cache(
@@ -816,7 +815,7 @@ public class INodeDirectory extends INodeWithAdditionalFields
   @VisibleForTesting
   @Override
   public void dumpTreeRecursively(PrintWriter out, StringBuilder prefix,
-      final Snapshot snapshot) {
+      final int snapshot) {
     super.dumpTreeRecursively(out, prefix, snapshot);
     out.print(", childrenSize=" + getChildrenList(snapshot).size());
     final DirectoryWithQuotaFeature q = getDirectoryWithQuotaFeature();
@@ -824,7 +823,7 @@ public class INodeDirectory extends INodeWithAdditionalFields
       out.print(", " + q);
     }
     if (this instanceof Snapshot.Root) {
-      out.print(", snapshotId=" + snapshot.getId());
+      out.print(", snapshotId=" + snapshot);
     }
     out.println();
 
@@ -869,7 +868,7 @@ public class INodeDirectory extends INodeWithAdditionalFields
       for(final Iterator<SnapshotAndINode> i = subs.iterator(); i.hasNext();) {
         final SnapshotAndINode pair = i.next();
         prefix.append(i.hasNext()? DUMPTREE_EXCEPT_LAST_ITEM: DUMPTREE_LAST_ITEM);
-        pair.inode.dumpTreeRecursively(out, prefix, pair.snapshot);
+        pair.inode.dumpTreeRecursively(out, prefix, pair.snapshotId);
         prefix.setLength(prefix.length() - 2);
       }
     }
@@ -877,20 +876,16 @@ public class INodeDirectory extends INodeWithAdditionalFields
 
   /** A pair of Snapshot and INode objects. */
   protected static class SnapshotAndINode {
-    public final Snapshot snapshot;
+    public final int snapshotId;
     public final INode inode;
 
-    public SnapshotAndINode(Snapshot snapshot, INode inode) {
+    public SnapshotAndINode(int snapshot, INode inode) {
-      this.snapshot = snapshot;
+      this.snapshotId = snapshot;
       this.inode = inode;
     }
-
-    public SnapshotAndINode(Snapshot snapshot) {
-      this(snapshot, snapshot.getRoot());
-    }
   }
 
-  public final int getChildrenNum(final Snapshot snapshot) {
+  public final int getChildrenNum(final int snapshotId) {
-    return getChildrenList(snapshot).size();
+    return getChildrenList(snapshotId).size();
   }
 }

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

@@ -17,6 +17,9 @@
  */
 package org.apache.hadoop.hdfs.server.namenode;
 
+import static org.apache.hadoop.hdfs.server.namenode.snapshot.Snapshot.CURRENT_STATE_ID;
+import static org.apache.hadoop.hdfs.server.namenode.snapshot.Snapshot.NO_SNAPSHOT_ID;
+
 import java.io.FileNotFoundException;
 import java.io.IOException;
 import java.io.PrintWriter;
@@ -272,26 +275,27 @@ public class INodeFile extends INodeWithAdditionalFields
   }
 
   @Override
-  public INodeFileAttributes getSnapshotINode(final Snapshot snapshot) {
+  public INodeFileAttributes getSnapshotINode(final int snapshotId) {
     FileWithSnapshotFeature sf = this.getFileWithSnapshotFeature();
     if (sf != null) {
-      return sf.getDiffs().getSnapshotINode(snapshot, this);
+      return sf.getDiffs().getSnapshotINode(snapshotId, this);
     } else {
       return this;
     }
   }
 
   @Override
-  public INodeFile recordModification(final Snapshot latest) 
+  public INodeFile recordModification(final int latestSnapshotId) 
       throws QuotaExceededException {
-    if (isInLatestSnapshot(latest) && !shouldRecordInSrcSnapshot(latest)) {
+    if (isInLatestSnapshot(latestSnapshotId)
+        && !shouldRecordInSrcSnapshot(latestSnapshotId)) {
       // the file is in snapshot, create a snapshot feature if it does not have
       FileWithSnapshotFeature sf = this.getFileWithSnapshotFeature();
       if (sf == null) {
         sf = addSnapshotFeature(null);
       }
       // record self in the diff list if necessary
-      sf.getDiffs().saveSelf2Snapshot(latest, this, null);
+      sf.getDiffs().saveSelf2Snapshot(latestSnapshotId, this, null);
     }
     return this;
   }
@@ -307,23 +311,22 @@ public class INodeFile extends INodeWithAdditionalFields
   /* End of Snapshot Feature */
 
   /** @return the replication factor of the file. */
-  public final short getFileReplication(Snapshot snapshot) {
+  public final short getFileReplication(int snapshot) {
-    if (snapshot != null) {
+    if (snapshot != CURRENT_STATE_ID) {
       return getSnapshotINode(snapshot).getFileReplication();
     }
-
     return HeaderFormat.getReplication(header);
   }
 
   /** The same as getFileReplication(null). */
   @Override // INodeFileAttributes
   public final short getFileReplication() {
-    return getFileReplication(null);
+    return getFileReplication(CURRENT_STATE_ID);
   }
 
   @Override // BlockCollection
   public short getBlockReplication() {
-    short max = getFileReplication(null);
+    short max = getFileReplication(CURRENT_STATE_ID);
     FileWithSnapshotFeature sf = this.getFileWithSnapshotFeature();
     if (sf != null) {
       short maxInSnapshot = sf.getMaxBlockRepInDiffs();
@@ -341,9 +344,10 @@ public class INodeFile extends INodeWithAdditionalFields
   }
 
   /** Set the replication factor of this file. */
-  public final INodeFile setFileReplication(short replication, Snapshot latest,
+  public final INodeFile setFileReplication(short replication,
-      final INodeMap inodeMap) throws QuotaExceededException {
+      int latestSnapshotId, final INodeMap inodeMap)
-    final INodeFile nodeToUpdate = recordModification(latest);
+      throws QuotaExceededException {
+    final INodeFile nodeToUpdate = recordModification(latestSnapshotId);
     nodeToUpdate.setFileReplication(replication);
     return nodeToUpdate;
   }
@@ -421,22 +425,22 @@ public class INodeFile extends INodeWithAdditionalFields
   }
 
   @Override
-  public Quota.Counts cleanSubtree(final Snapshot snapshot, Snapshot prior,
+  public Quota.Counts cleanSubtree(final int snapshot, int priorSnapshotId,
       final BlocksMapUpdateInfo collectedBlocks,
       final List<INode> removedINodes, final boolean countDiffChange)
       throws QuotaExceededException {
     FileWithSnapshotFeature sf = getFileWithSnapshotFeature();
     if (sf != null) {
-      return sf.cleanFile(this, snapshot, prior, collectedBlocks,
+      return sf.cleanFile(this, snapshot, priorSnapshotId, collectedBlocks,
           removedINodes, countDiffChange);
     }
     Quota.Counts counts = Quota.Counts.newInstance();
-    if (snapshot == null && prior == null) {
+    if (snapshot == CURRENT_STATE_ID && priorSnapshotId == NO_SNAPSHOT_ID) {
       // this only happens when deleting the current file and the file is not
       // in any snapshot
       computeQuotaUsage(counts, false);
       destroyAndCollectBlocks(collectedBlocks, removedINodes);
-    } else if (snapshot == null && prior != null) {
+    } else if (snapshot == CURRENT_STATE_ID && priorSnapshotId != NO_SNAPSHOT_ID) {
       // when deleting the current file and the file is in snapshot, we should
       // clean the 0-sized block if the file is UC
       FileUnderConstructionFeature uc = getFileUnderConstructionFeature();
@@ -480,17 +484,18 @@ public class INodeFile extends INodeWithAdditionalFields
     FileWithSnapshotFeature sf = getFileWithSnapshotFeature();
     if (sf != null) {
       FileDiffList fileDiffList = sf.getDiffs();
-      Snapshot last = fileDiffList.getLastSnapshot();
+      int last = fileDiffList.getLastSnapshotId();
       List<FileDiff> diffs = fileDiffList.asList();
 
-      if (lastSnapshotId == Snapshot.INVALID_ID || last == null) {
+      if (lastSnapshotId == Snapshot.CURRENT_STATE_ID
+          || last == Snapshot.CURRENT_STATE_ID) {
         nsDelta += diffs.size();
         dsDelta = diskspaceConsumed();
-      } else if (last.getId() < lastSnapshotId) {
+      } else if (last < lastSnapshotId) {
         dsDelta = computeFileSize(true, false) * getFileReplication();
       } else {      
-        Snapshot s = fileDiffList.getSnapshotById(lastSnapshotId);
+        int sid = fileDiffList.getSnapshotById(lastSnapshotId);
-        dsDelta = diskspaceConsumed(s);
+        dsDelta = diskspaceConsumed(sid);
       }
     } else {
       dsDelta = diskspaceConsumed();
@@ -501,7 +506,7 @@ public class INodeFile extends INodeWithAdditionalFields
   }
 
   @Override
-  public final ContentSummaryComputationContext  computeContentSummary(
+  public final ContentSummaryComputationContext computeContentSummary(
       final ContentSummaryComputationContext summary) {
     computeContentSummary4Snapshot(summary.getCounts());
     computeContentSummary4Current(summary.getCounts());
@@ -540,23 +545,21 @@ public class INodeFile extends INodeWithAdditionalFields
 
   /** The same as computeFileSize(null). */
   public final long computeFileSize() {
-    return computeFileSize(null);
+    return computeFileSize(CURRENT_STATE_ID);
   }
 
   /**
    * Compute file size of the current file if the given snapshot is null;
    * otherwise, get the file size from the given snapshot.
    */
-  public final long computeFileSize(Snapshot snapshot) {
+  public final long computeFileSize(int snapshotId) {
     FileWithSnapshotFeature sf = this.getFileWithSnapshotFeature();
-    if (snapshot != null && sf != null) {
+    if (snapshotId != CURRENT_STATE_ID && sf != null) {
-      final FileDiff d = sf.getDiffs().getDiff(
+      final FileDiff d = sf.getDiffs().getDiffById(snapshotId);
-          snapshot);
       if (d != null) {
         return d.getFileSize();
       }
     }
-
     return computeFileSize(true, false);
   }
 
@@ -607,9 +610,10 @@ public class INodeFile extends INodeWithAdditionalFields
     return computeFileSize(true, true) * getBlockReplication();
   }
 
-  public final long diskspaceConsumed(Snapshot lastSnapshot) {
+  public final long diskspaceConsumed(int lastSnapshotId) {
-    if (lastSnapshot != null) {
+    if (lastSnapshotId != CURRENT_STATE_ID) {
-      return computeFileSize(lastSnapshot) * getFileReplication(lastSnapshot);
+      return computeFileSize(lastSnapshotId)
+          * getFileReplication(lastSnapshotId);
     } else {
       return diskspaceConsumed();
     }
@@ -638,9 +642,9 @@ public class INodeFile extends INodeWithAdditionalFields
   @VisibleForTesting
   @Override
   public void dumpTreeRecursively(PrintWriter out, StringBuilder prefix,
-      final Snapshot snapshot) {
+      final int snapshotId) {
-    super.dumpTreeRecursively(out, prefix, snapshot);
+    super.dumpTreeRecursively(out, prefix, snapshotId);
-    out.print(", fileSize=" + computeFileSize(snapshot));
+    out.print(", fileSize=" + computeFileSize(snapshotId));
     // only compare the first block
     out.print(", blocks=");
     out.print(blocks == null || blocks.length == 0? null: blocks[0]);

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

@@ -23,7 +23,6 @@ import org.apache.hadoop.fs.permission.FsPermission;
 import org.apache.hadoop.fs.permission.PermissionStatus;
 import org.apache.hadoop.hdfs.protocol.QuotaExceededException;
 import org.apache.hadoop.hdfs.server.namenode.Quota.Counts;
-import org.apache.hadoop.hdfs.server.namenode.snapshot.Snapshot;
 import org.apache.hadoop.util.GSet;
 import org.apache.hadoop.util.LightWeightGSet;
 
@@ -89,7 +88,8 @@ public class INodeMap {
         "", "", new FsPermission((short) 0)), 0, 0) {
       
       @Override
-      INode recordModification(Snapshot latest) throws QuotaExceededException {
+      INode recordModification(int latestSnapshotId)
+          throws QuotaExceededException {
         return null;
       }
       
@@ -112,7 +112,7 @@ public class INodeMap {
       }
       
       @Override
-      public Counts cleanSubtree(Snapshot snapshot, Snapshot prior,
+      public Counts cleanSubtree(int snapshotId, int priorSnapshotId,
           BlocksMapUpdateInfo collectedBlocks, List<INode> removedINodes,
           boolean countDiffChange) throws QuotaExceededException {
         return null;

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

@@ -91,7 +91,7 @@ public abstract class INodeReference extends INode {
    * method to identify the snapshot which is the latest snapshot before the
    * reference node's creation. 
    */
-  static Snapshot getPriorSnapshot(INodeReference ref) {
+  static int getPriorSnapshot(INodeReference ref) {
     WithCount wc = (WithCount) ref.getReferredINode();
     WithName wn = null;
     if (ref instanceof DstReference) {
@@ -111,7 +111,7 @@ public abstract class INodeReference extends INode {
         }
       }
     }
-    return null;
+    return Snapshot.NO_SNAPSHOT_ID;
   }
   
   private INode referred;
@@ -185,13 +185,13 @@ public abstract class INodeReference extends INode {
   }
   
   @Override
-  public final PermissionStatus getPermissionStatus(Snapshot snapshot) {
+  public final PermissionStatus getPermissionStatus(int snapshotId) {
-    return referred.getPermissionStatus(snapshot);
+    return referred.getPermissionStatus(snapshotId);
   }
   
   @Override
-  public final String getUserName(Snapshot snapshot) {
+  public final String getUserName(int snapshotId) {
-    return referred.getUserName(snapshot);
+    return referred.getUserName(snapshotId);
   }
   
   @Override
@@ -200,8 +200,8 @@ public abstract class INodeReference extends INode {
   }
   
   @Override
-  public final String getGroupName(Snapshot snapshot) {
+  public final String getGroupName(int snapshotId) {
-    return referred.getGroupName(snapshot);
+    return referred.getGroupName(snapshotId);
   }
   
   @Override
@@ -210,8 +210,8 @@ public abstract class INodeReference extends INode {
   }
   
   @Override
-  public final FsPermission getFsPermission(Snapshot snapshot) {
+  public final FsPermission getFsPermission(int snapshotId) {
-    return referred.getFsPermission(snapshot);
+    return referred.getFsPermission(snapshotId);
   }
   @Override
   public final short getFsPermissionShort() {
@@ -229,14 +229,14 @@ public abstract class INodeReference extends INode {
   }
 
   @Override
-  public final long getModificationTime(Snapshot snapshot) {
+  public final long getModificationTime(int snapshotId) {
-    return referred.getModificationTime(snapshot);
+    return referred.getModificationTime(snapshotId);
   }
   
   @Override
-  public final INode updateModificationTime(long mtime, Snapshot latest) 
+  public final INode updateModificationTime(long mtime, int latestSnapshotId) 
       throws QuotaExceededException {
-    return referred.updateModificationTime(mtime, latest);
+    return referred.updateModificationTime(mtime, latestSnapshotId);
   }
   
   @Override
@@ -245,8 +245,8 @@ public abstract class INodeReference extends INode {
   }
   
   @Override
-  public final long getAccessTime(Snapshot snapshot) {
+  public final long getAccessTime(int snapshotId) {
-    return referred.getAccessTime(snapshot);
+    return referred.getAccessTime(snapshotId);
   }
   
   @Override
@@ -255,15 +255,15 @@ public abstract class INodeReference extends INode {
   }
 
   @Override
-  final INode recordModification(Snapshot latest)
+  final INode recordModification(int latestSnapshotId)
       throws QuotaExceededException {
-    referred.recordModification(latest);
+    referred.recordModification(latestSnapshotId);
     // reference is never replaced 
     return this;
   }
 
   @Override // used by WithCount
-  public Quota.Counts cleanSubtree(Snapshot snapshot, Snapshot prior,
+  public Quota.Counts cleanSubtree(int snapshot, int prior,
       BlocksMapUpdateInfo collectedBlocks, final List<INode> removedINodes,
       final boolean countDiffChange) throws QuotaExceededException {
     return referred.cleanSubtree(snapshot, prior, collectedBlocks,
@@ -291,8 +291,8 @@ public abstract class INodeReference extends INode {
   }
   
   @Override
-  public final INodeAttributes getSnapshotINode(Snapshot snapshot) {
+  public final INodeAttributes getSnapshotINode(int snapshotId) {
-    return referred.getSnapshotINode(snapshot);
+    return referred.getSnapshotINode(snapshotId);
   }
 
   @Override
@@ -308,7 +308,7 @@ public abstract class INodeReference extends INode {
 
   @Override
   public void dumpTreeRecursively(PrintWriter out, StringBuilder prefix,
-      final Snapshot snapshot) {
+      final int snapshot) {
     super.dumpTreeRecursively(out, prefix, snapshot);
     if (this instanceof DstReference) {
       out.print(", dstSnapshotId=" + ((DstReference) this).dstSnapshotId);
@@ -327,7 +327,7 @@ public abstract class INodeReference extends INode {
   }
   
   public int getDstSnapshotId() {
-    return Snapshot.INVALID_ID;
+    return Snapshot.CURRENT_STATE_ID;
   }
   
   /** An anonymous reference with reference count. */
@@ -457,34 +457,35 @@ public abstract class INodeReference extends INode {
       // node happened before the rename of its ancestor. This should be 
       // impossible since for WithName node we only count its children at the 
       // time of the rename. 
-      Preconditions.checkState(this.lastSnapshotId >= lastSnapshotId);
+      Preconditions.checkState(lastSnapshotId == Snapshot.CURRENT_STATE_ID
+          || this.lastSnapshotId >= lastSnapshotId);
       final INode referred = this.getReferredINode().asReference()
           .getReferredINode();
       // We will continue the quota usage computation using the same snapshot id
       // as time line (if the given snapshot id is valid). Also, we cannot use 
       // cache for the referred node since its cached quota may have already 
       // been updated by changes in the current tree.
-      int id = lastSnapshotId > Snapshot.INVALID_ID ? 
+      int id = lastSnapshotId != Snapshot.CURRENT_STATE_ID ? 
           lastSnapshotId : this.lastSnapshotId;
       return referred.computeQuotaUsage(counts, false, id);
     }
     
     @Override
-    public Quota.Counts cleanSubtree(final Snapshot snapshot, Snapshot prior,
+    public Quota.Counts cleanSubtree(final int snapshot, int prior,
         final BlocksMapUpdateInfo collectedBlocks,
         final List<INode> removedINodes, final boolean countDiffChange)
         throws QuotaExceededException {
       // since WithName node resides in deleted list acting as a snapshot copy,
       // the parameter snapshot must be non-null
-      Preconditions.checkArgument(snapshot != null);
+      Preconditions.checkArgument(snapshot != Snapshot.CURRENT_STATE_ID);
-      // if prior is null, we need to check snapshot belonging to the previous
+      // if prior is NO_SNAPSHOT_ID, we need to check snapshot belonging to the
-      // WithName instance
+      // previous WithName instance
-      if (prior == null) {
+      if (prior == Snapshot.NO_SNAPSHOT_ID) {
         prior = getPriorSnapshot(this);
       }
       
-      if (prior != null
+      if (prior != Snapshot.NO_SNAPSHOT_ID
-          && Snapshot.ID_COMPARATOR.compare(snapshot, prior) <= 0) {
+          && Snapshot.ID_INTEGER_COMPARATOR.compare(snapshot, prior) <= 0) {
         return Quota.Counts.newInstance();
       }
 
@@ -496,7 +497,7 @@ public abstract class INodeReference extends INode {
             -counts.get(Quota.DISKSPACE), true);
       }
       
-      if (snapshot.getId() < lastSnapshotId) {
+      if (snapshot < lastSnapshotId) {
         // for a WithName node, when we compute its quota usage, we only count
         // in all the nodes existing at the time of the corresponding rename op.
         // Thus if we are deleting a snapshot before/at the snapshot associated 
@@ -509,16 +510,16 @@ public abstract class INodeReference extends INode {
     @Override
     public void destroyAndCollectBlocks(BlocksMapUpdateInfo collectedBlocks,
         final List<INode> removedINodes) {
-      Snapshot snapshot = getSelfSnapshot();
+      int snapshot = getSelfSnapshot();
       if (removeReference(this) <= 0) {
         getReferredINode().destroyAndCollectBlocks(collectedBlocks,
             removedINodes);
       } else {
-        Snapshot prior = getPriorSnapshot(this);
+        int prior = getPriorSnapshot(this);
         INode referred = getReferredINode().asReference().getReferredINode();
         
-        if (snapshot != null) {
+        if (snapshot != Snapshot.NO_SNAPSHOT_ID) {
-          if (prior != null && snapshot.getId() <= prior.getId()) {
+          if (prior != Snapshot.NO_SNAPSHOT_ID && snapshot <= prior) {
             // the snapshot to be deleted has been deleted while traversing 
             // the src tree of the previous rename operation. This usually 
             // happens when rename's src and dst are under the same 
@@ -545,9 +546,9 @@ public abstract class INodeReference extends INode {
       }
     }
     
-    private Snapshot getSelfSnapshot() {
+    private int getSelfSnapshot() {
       INode referred = getReferredINode().asReference().getReferredINode();
-      Snapshot snapshot = null;
+      int snapshot = Snapshot.NO_SNAPSHOT_ID;
       if (referred.isFile() && referred.asFile().isWithSnapshot()) {
         snapshot = referred.asFile().getDiffs().getPrior(lastSnapshotId);
       } else if (referred.isDirectory()) {
@@ -569,7 +570,7 @@ public abstract class INodeReference extends INode {
      * latest snapshot. Otherwise changes will be recorded to the snapshot
      * belonging to the src of the rename.
      * 
-     * {@link Snapshot#INVALID_ID} means no dstSnapshot (e.g., src of the
+     * {@link Snapshot#NO_SNAPSHOT_ID} means no dstSnapshot (e.g., src of the
      * first-time rename).
      */
     private final int dstSnapshotId;
@@ -587,25 +588,27 @@ public abstract class INodeReference extends INode {
     }
     
     @Override
-    public Quota.Counts cleanSubtree(Snapshot snapshot, Snapshot prior,
+    public Quota.Counts cleanSubtree(int snapshot, int prior,
         BlocksMapUpdateInfo collectedBlocks, List<INode> removedINodes,
         final boolean countDiffChange) throws QuotaExceededException {
-      if (snapshot == null && prior == null) {
+      if (snapshot == Snapshot.CURRENT_STATE_ID
+          && prior == Snapshot.NO_SNAPSHOT_ID) {
         Quota.Counts counts = Quota.Counts.newInstance();
         this.computeQuotaUsage(counts, true);
         destroyAndCollectBlocks(collectedBlocks, removedINodes);
         return counts;
       } else {
-        // if prior is null, we need to check snapshot belonging to the previous
+        // if prior is NO_SNAPSHOT_ID, we need to check snapshot belonging to 
-        // WithName instance
+        // the previous WithName instance
-        if (prior == null) {
+        if (prior == Snapshot.NO_SNAPSHOT_ID) {
           prior = getPriorSnapshot(this);
         }
-        // if prior is not null, and prior is not before the to-be-deleted 
+        // if prior is not NO_SNAPSHOT_ID, and prior is not before the
-        // snapshot, we can quit here and leave the snapshot deletion work to 
+        // to-be-deleted snapshot, we can quit here and leave the snapshot
-        // the src tree of rename
+        // deletion work to the src tree of rename
-        if (snapshot != null && prior != null
+        if (snapshot != Snapshot.CURRENT_STATE_ID
-            && Snapshot.ID_COMPARATOR.compare(snapshot, prior) <= 0) {
+            && prior != Snapshot.NO_SNAPSHOT_ID
+            && Snapshot.ID_INTEGER_COMPARATOR.compare(snapshot, prior) <= 0) {
           return Quota.Counts.newInstance();
         }
         return getReferredINode().cleanSubtree(snapshot, prior,
@@ -632,12 +635,12 @@ public abstract class INodeReference extends INode {
       } else {
         // we will clean everything, including files, directories, and 
         // snapshots, that were created after this prior snapshot
-        Snapshot prior = getPriorSnapshot(this);
+        int prior = getPriorSnapshot(this);
         // prior must be non-null, otherwise we do not have any previous 
         // WithName nodes, and the reference number will be 0.
-        Preconditions.checkState(prior != null);
+        Preconditions.checkState(prior != Snapshot.NO_SNAPSHOT_ID);
         // identify the snapshot created after prior
-        Snapshot snapshot = getSelfSnapshot(prior);
+        int snapshot = getSelfSnapshot(prior);
         
         INode referred = getReferredINode().asReference().getReferredINode();
         if (referred.isFile()) {
@@ -671,23 +674,23 @@ public abstract class INodeReference extends INode {
       }
     }
     
-    private Snapshot getSelfSnapshot(final Snapshot prior) {
+    private int getSelfSnapshot(final int prior) {
       WithCount wc = (WithCount) getReferredINode().asReference();
       INode referred = wc.getReferredINode();
-      Snapshot lastSnapshot = null;
+      int lastSnapshot = Snapshot.CURRENT_STATE_ID;
       if (referred.isFile() && referred.asFile().isWithSnapshot()) {
-        lastSnapshot = referred.asFile().getDiffs().getLastSnapshot();
+        lastSnapshot = referred.asFile().getDiffs().getLastSnapshotId();
       } else if (referred.isDirectory()) {
         DirectoryWithSnapshotFeature sf = referred.asDirectory()
             .getDirectoryWithSnapshotFeature();
         if (sf != null) {
-          lastSnapshot = sf.getLastSnapshot();
+          lastSnapshot = sf.getLastSnapshotId();
         }
       }
-      if (lastSnapshot != null && !lastSnapshot.equals(prior)) {
+      if (lastSnapshot != Snapshot.CURRENT_STATE_ID && lastSnapshot != prior) {
         return lastSnapshot;
       } else {
-        return null;
+        return Snapshot.CURRENT_STATE_ID;
       }
     }
   }

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

@@ -45,10 +45,10 @@ public class INodeSymlink extends INodeWithAdditionalFields {
   }
 
   @Override
-  INode recordModification(Snapshot latest) throws QuotaExceededException {
+  INode recordModification(int latestSnapshotId) throws QuotaExceededException {
-    if (isInLatestSnapshot(latest)) {
+    if (isInLatestSnapshot(latestSnapshotId)) {
       INodeDirectory parent = getParent();
-      parent.saveChild2Snapshot(this, latest, new INodeSymlink(this));
+      parent.saveChild2Snapshot(this, latestSnapshotId, new INodeSymlink(this));
     }
     return this;
   }
@@ -74,10 +74,11 @@ public class INodeSymlink extends INodeWithAdditionalFields {
   }
   
   @Override
-  public Quota.Counts cleanSubtree(final Snapshot snapshot, Snapshot prior,
+  public Quota.Counts cleanSubtree(final int snapshotId, int priorSnapshotId,
       final BlocksMapUpdateInfo collectedBlocks,
       final List<INode> removedINodes, final boolean countDiffChange) {
-    if (snapshot == null && prior == null) {
+    if (snapshotId == Snapshot.CURRENT_STATE_ID
+        && priorSnapshotId == Snapshot.NO_SNAPSHOT_ID) {
       destroyAndCollectBlocks(collectedBlocks, removedINodes);
     }
     return Quota.Counts.newInstance(1, 0);
@@ -105,7 +106,7 @@ public class INodeSymlink extends INodeWithAdditionalFields {
 
   @Override
   public void dumpTreeRecursively(PrintWriter out, StringBuilder prefix,
-      final Snapshot snapshot) {
+      final int snapshot) {
     super.dumpTreeRecursively(out, prefix, snapshot);
     out.println();
   }

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

@@ -155,9 +155,9 @@ public abstract class INodeWithAdditionalFields extends INode
   }
 
   @Override
-  final PermissionStatus getPermissionStatus(Snapshot snapshot) {
+  final PermissionStatus getPermissionStatus(int snapshotId) {
-    return new PermissionStatus(getUserName(snapshot), getGroupName(snapshot),
+    return new PermissionStatus(getUserName(snapshotId), getGroupName(snapshotId),
-        getFsPermission(snapshot));
+        getFsPermission(snapshotId));
   }
 
   private final void updatePermissionStatus(PermissionStatusFormat f, long n) {
@@ -165,9 +165,9 @@ public abstract class INodeWithAdditionalFields extends INode
   }
 
   @Override
-  final String getUserName(Snapshot snapshot) {
+  final String getUserName(int snapshotId) {
-    if (snapshot != null) {
+    if (snapshotId != Snapshot.CURRENT_STATE_ID) {
-      return getSnapshotINode(snapshot).getUserName();
+      return getSnapshotINode(snapshotId).getUserName();
     }
 
     int n = (int)PermissionStatusFormat.USER.retrieve(permission);
@@ -181,9 +181,9 @@ public abstract class INodeWithAdditionalFields extends INode
   }
 
   @Override
-  final String getGroupName(Snapshot snapshot) {
+  final String getGroupName(int snapshotId) {
-    if (snapshot != null) {
+    if (snapshotId != Snapshot.CURRENT_STATE_ID) {
-      return getSnapshotINode(snapshot).getGroupName();
+      return getSnapshotINode(snapshotId).getGroupName();
     }
 
     int n = (int)PermissionStatusFormat.GROUP.retrieve(permission);
@@ -197,9 +197,9 @@ public abstract class INodeWithAdditionalFields extends INode
   }
 
   @Override
-  final FsPermission getFsPermission(Snapshot snapshot) {
+  final FsPermission getFsPermission(int snapshotId) {
-    if (snapshot != null) {
+    if (snapshotId != Snapshot.CURRENT_STATE_ID) {
-      return getSnapshotINode(snapshot).getFsPermission();
+      return getSnapshotINode(snapshotId).getFsPermission();
     }
 
     return new FsPermission(getFsPermissionShort());
@@ -221,9 +221,9 @@ public abstract class INodeWithAdditionalFields extends INode
   }
 
   @Override
-  final long getModificationTime(Snapshot snapshot) {
+  final long getModificationTime(int snapshotId) {
-    if (snapshot != null) {
+    if (snapshotId != Snapshot.CURRENT_STATE_ID) {
-      return getSnapshotINode(snapshot).getModificationTime();
+      return getSnapshotINode(snapshotId).getModificationTime();
     }
 
     return this.modificationTime;
@@ -232,13 +232,13 @@ public abstract class INodeWithAdditionalFields extends INode
 
   /** Update modification time if it is larger than the current value. */
   @Override
-  public final INode updateModificationTime(long mtime, Snapshot latest) 
+  public final INode updateModificationTime(long mtime, int latestSnapshotId) 
       throws QuotaExceededException {
     Preconditions.checkState(isDirectory());
     if (mtime <= modificationTime) {
       return this;
     }
-    return setModificationTime(mtime, latest);
+    return setModificationTime(mtime, latestSnapshotId);
   }
 
   final void cloneModificationTime(INodeWithAdditionalFields that) {
@@ -251,11 +251,10 @@ public abstract class INodeWithAdditionalFields extends INode
   }
 
   @Override
-  final long getAccessTime(Snapshot snapshot) {
+  final long getAccessTime(int snapshotId) {
-    if (snapshot != null) {
+    if (snapshotId != Snapshot.CURRENT_STATE_ID) {
-      return getSnapshotINode(snapshot).getAccessTime();
+      return getSnapshotINode(snapshotId).getAccessTime();
     }
-
     return accessTime;
   }
 

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

@@ -135,8 +135,8 @@ public class INodesInPath {
       if (!isRef && isDir && dir.isWithSnapshot()) {
         //if the path is a non-snapshot path, update the latest snapshot.
         if (!existing.isSnapshot()) {
-          existing.updateLatestSnapshot(dir.getDirectoryWithSnapshotFeature()
+          existing.updateLatestSnapshotId(dir.getDirectoryWithSnapshotFeature()
-              .getLastSnapshot());
+              .getLastSnapshotId());
         }
       } else if (isRef && isDir && !lastComp) {
         // If the curNode is a reference node, need to check its dstSnapshot:
@@ -151,16 +151,17 @@ public class INodesInPath {
         // recordModification method.
         if (!existing.isSnapshot()) {
           int dstSnapshotId = curNode.asReference().getDstSnapshotId();
-          Snapshot latest = existing.getLatestSnapshot();
+          int latest = existing.getLatestSnapshotId();
-          if (latest == null ||  // no snapshot in dst tree of rename
+          if (latest == Snapshot.CURRENT_STATE_ID || // no snapshot in dst tree of rename
-              dstSnapshotId >= latest.getId()) { // the above scenario 
+              (dstSnapshotId != Snapshot.CURRENT_STATE_ID && 
-            Snapshot lastSnapshot = null;
+                dstSnapshotId >= latest)) { // the above scenario 
+            int lastSnapshot = Snapshot.CURRENT_STATE_ID;
             DirectoryWithSnapshotFeature sf = null;
             if (curNode.isDirectory() && 
                 (sf = curNode.asDirectory().getDirectoryWithSnapshotFeature()) != null) {
-              lastSnapshot = sf.getLastSnapshot();
+              lastSnapshot = sf.getLastSnapshotId();
             }
-            existing.setSnapshot(lastSnapshot);
+            existing.setSnapshotId(lastSnapshot);
           }
         }
       }
@@ -206,14 +207,14 @@ public class INodesInPath {
           curNode = null;
         } else {
           curNode = s.getRoot();
-          existing.setSnapshot(s);
+          existing.setSnapshotId(s.getId());
         }
         if (index >= -1) {
           existing.snapshotRootIndex = existing.numNonNull;
         }
       } else {
         // normal case, and also for resolving file/dir under snapshot root
-        curNode = dir.getChild(childName, existing.getPathSnapshot());
+        curNode = dir.getChild(childName, existing.getPathSnapshotId());
       }
       count++;
       index++;
@@ -245,11 +246,12 @@ public class INodesInPath {
    */
   private int snapshotRootIndex;
   /**
-   * For snapshot paths, it is the reference to the snapshot; or null if the
+   * For snapshot paths, it is the id of the snapshot; or 
-   * snapshot does not exist. For non-snapshot paths, it is the reference to
+   * {@link Snapshot#CURRENT_STATE_ID} if the snapshot does not exist. For 
-   * the latest snapshot found in the path; or null if no snapshot is found.
+   * non-snapshot paths, it is the id of the latest snapshot found in the path;
+   * or {@link Snapshot#CURRENT_STATE_ID} if no snapshot is found.
    */
-  private Snapshot snapshot = null; 
+  private int snapshotId = Snapshot.CURRENT_STATE_ID; 
 
   private INodesInPath(byte[][] path, int number) {
     this.path = path;
@@ -262,29 +264,30 @@ public class INodesInPath {
   }
 
   /**
-   * For non-snapshot paths, return the latest snapshot found in the path.
+   * For non-snapshot paths, return the latest snapshot id found in the path.
-   * For snapshot paths, return null.
    */
-  public Snapshot getLatestSnapshot() {
+  public int getLatestSnapshotId() {
-    return isSnapshot? null: snapshot;
+    Preconditions.checkState(!isSnapshot);
+    return snapshotId;
   }
   
   /**
-   * For snapshot paths, return the snapshot specified in the path.
+   * For snapshot paths, return the id of the snapshot specified in the path.
-   * For non-snapshot paths, return null.
+   * For non-snapshot paths, return {@link Snapshot#CURRENT_STATE_ID}.
    */
-  public Snapshot getPathSnapshot() {
+  public int getPathSnapshotId() {
-    return isSnapshot? snapshot: null;
+    return isSnapshot ? snapshotId : Snapshot.CURRENT_STATE_ID;
   }
 
-  private void setSnapshot(Snapshot s) {
+  private void setSnapshotId(int sid) {
-    snapshot = s;
+    snapshotId = sid;
   }
   
-  private void updateLatestSnapshot(Snapshot s) {
+  private void updateLatestSnapshotId(int sid) {
-    if (snapshot == null
+    if (snapshotId == Snapshot.CURRENT_STATE_ID
-        || (s != null && Snapshot.ID_COMPARATOR.compare(snapshot, s) < 0)) {
+        || (sid != Snapshot.CURRENT_STATE_ID && Snapshot.ID_INTEGER_COMPARATOR
-      snapshot = s;
+            .compare(snapshotId, sid) < 0)) {
+      snapshotId = sid;
     }
   }
 
@@ -386,7 +389,7 @@ public class INodesInPath {
      .append("\n  capacity   = ").append(capacity)
      .append("\n  isSnapshot        = ").append(isSnapshot)
      .append("\n  snapshotRootIndex = ").append(snapshotRootIndex)
-     .append("\n  snapshot          = ").append(snapshot);
+     .append("\n  snapshotId        = ").append(snapshotId);
     return b.toString();
   }
 

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

@@ -22,8 +22,8 @@ import java.io.IOException;
 import java.util.List;
 
 import org.apache.hadoop.hdfs.server.namenode.INode;
-import org.apache.hadoop.hdfs.server.namenode.INodeAttributes;
 import org.apache.hadoop.hdfs.server.namenode.INode.BlocksMapUpdateInfo;
+import org.apache.hadoop.hdfs.server.namenode.INodeAttributes;
 import org.apache.hadoop.hdfs.server.namenode.Quota;
 import org.apache.hadoop.hdfs.server.namenode.snapshot.SnapshotFSImageFormat.ReferenceMap;
 
@@ -52,8 +52,8 @@ abstract class AbstractINodeDiff<N extends INode,
                                  D extends AbstractINodeDiff<N, A, D>>
     implements Comparable<Integer> {
 
-  /** The snapshot will be obtained after this diff is applied. */
+  /** The id of the corresponding snapshot. */
-  Snapshot snapshot;
+  private int snapshotId;
   /** The snapshot inode data.  It is null when there is no change. */
   A snapshotINode;
   /**
@@ -64,10 +64,8 @@ abstract class AbstractINodeDiff<N extends INode,
    */
   private D posteriorDiff;
 
-  AbstractINodeDiff(Snapshot snapshot, A snapshotINode, D posteriorDiff) {
+  AbstractINodeDiff(int snapshotId, A snapshotINode, D posteriorDiff) {
-    Preconditions.checkNotNull(snapshot, "snapshot is null");
+    this.snapshotId = snapshotId;
-
-    this.snapshot = snapshot;
     this.snapshotINode = snapshotINode;
     this.posteriorDiff = posteriorDiff;
   }
@@ -75,16 +73,16 @@ abstract class AbstractINodeDiff<N extends INode,
   /** Compare diffs with snapshot ID. */
   @Override
   public final int compareTo(final Integer that) {
-    return Snapshot.ID_INTEGER_COMPARATOR.compare(this.snapshot.getId(), that);
+    return Snapshot.ID_INTEGER_COMPARATOR.compare(this.snapshotId, that);
   }
 
   /** @return the snapshot object of this diff. */
-  public final Snapshot getSnapshot() {
+  public final int getSnapshotId() {
-    return snapshot;
+    return snapshotId;
   }
   
-  final void setSnapshot(Snapshot snapshot) {
+  final void setSnapshotId(int snapshot) {
-    this.snapshot = snapshot;
+    this.snapshotId = snapshot;
   }
 
   /** @return the posterior diff. */
@@ -132,13 +130,12 @@ abstract class AbstractINodeDiff<N extends INode,
 
   @Override
   public String toString() {
-    return getClass().getSimpleName() + ": " + snapshot + " (post="
+    return getClass().getSimpleName() + ": " + this.getSnapshotId() + " (post="
-        + (posteriorDiff == null? null: posteriorDiff.snapshot) + ")";
+        + (posteriorDiff == null? null: posteriorDiff.getSnapshotId()) + ")";
   }
 
   void writeSnapshot(DataOutput out) throws IOException {
-    // Assume the snapshot is recorded before, write id only.
+    out.writeInt(snapshotId);
-    out.writeInt(snapshot.getId());
   }
   
   abstract void write(DataOutput out, ReferenceMap referenceMap

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

@@ -53,7 +53,7 @@ abstract class AbstractINodeDiffList<N extends INode,
   }
 
   /** @return an {@link AbstractINodeDiff}. */
-  abstract D createDiff(Snapshot snapshot, N currentINode);
+  abstract D createDiff(int snapshotId, N currentINode);
 
   /** @return a snapshot copy of the current inode. */  
   abstract A createSnapshotCopy(N currentINode);
@@ -63,25 +63,25 @@ abstract class AbstractINodeDiffList<N extends INode,
    * outside. If the diff to remove is not the first one in the diff list, we 
    * need to combine the diff with its previous one.
    * 
-   * @param snapshot The snapshot to be deleted
+   * @param snapshot The id of the snapshot to be deleted
-   * @param prior The snapshot taken before the to-be-deleted snapshot
+   * @param prior The id of the snapshot taken before the to-be-deleted snapshot
    * @param collectedBlocks Used to collect information for blocksMap update
    * @return delta in namespace. 
    */
-  public final Quota.Counts deleteSnapshotDiff(final Snapshot snapshot,
+  public final Quota.Counts deleteSnapshotDiff(final int snapshot,
-      Snapshot prior, final N currentINode,
+      final int prior, final N currentINode,
       final BlocksMapUpdateInfo collectedBlocks,
       final List<INode> removedINodes, boolean countDiffChange) 
       throws QuotaExceededException {
-    int snapshotIndex = Collections.binarySearch(diffs, snapshot.getId());
+    int snapshotIndex = Collections.binarySearch(diffs, snapshot);
     
     Quota.Counts counts = Quota.Counts.newInstance();
     D removed = null;
     if (snapshotIndex == 0) {
-      if (prior != null) {
+      if (prior != Snapshot.NO_SNAPSHOT_ID) { // there is still snapshot before
         // set the snapshot to latestBefore
-        diffs.get(snapshotIndex).setSnapshot(prior);
+        diffs.get(snapshotIndex).setSnapshotId(prior);
-      } else {
+      } else { // there is no snapshot before
         removed = diffs.remove(0);
         if (countDiffChange) {
           counts.add(Quota.NAMESPACE, 1);
@@ -96,8 +96,8 @@ abstract class AbstractINodeDiffList<N extends INode,
       }
     } else if (snapshotIndex > 0) {
       final AbstractINodeDiff<N, A, D> previous = diffs.get(snapshotIndex - 1);
-      if (!previous.getSnapshot().equals(prior)) {
+      if (previous.getSnapshotId() != prior) {
-        diffs.get(snapshotIndex).setSnapshot(prior);
+        diffs.get(snapshotIndex).setSnapshotId(prior);
       } else {
         // combine the to-be-removed diff with its previous diff
         removed = diffs.remove(snapshotIndex);
@@ -120,10 +120,10 @@ abstract class AbstractINodeDiffList<N extends INode,
   }
 
   /** Add an {@link AbstractINodeDiff} for the given snapshot. */
-  final D addDiff(Snapshot latest, N currentINode)
+  final D addDiff(int latestSnapshotId, N currentINode)
       throws QuotaExceededException {
     currentINode.addSpaceConsumed(1, 0, true);
-    return addLast(createDiff(latest, currentINode));
+    return addLast(createDiff(latestSnapshotId, currentINode));
   }
 
   /** Append the diff at the end of the list. */
@@ -149,10 +149,10 @@ abstract class AbstractINodeDiffList<N extends INode,
     return n == 0? null: diffs.get(n - 1);
   }
 
-  /** @return the last snapshot. */
+  /** @return the id of the last snapshot. */
-  public final Snapshot getLastSnapshot() {
+  public final int getLastSnapshotId() {
     final AbstractINodeDiff<N, A, D> last = getLast();
-    return last == null? null: last.getSnapshot();
+    return last == null ? Snapshot.CURRENT_STATE_ID : last.getSnapshotId();
   }
   
   /**
@@ -161,60 +161,49 @@ abstract class AbstractINodeDiffList<N extends INode,
    *                 snapshot id.
    * @param exclusive True means the returned snapshot's id must be < the given
    *                  id, otherwise <=.
-   * @return The latest snapshot before the given snapshot.
+   * @return The id of the latest snapshot before the given snapshot.
    */
-  private final Snapshot getPrior(int anchorId, boolean exclusive) {
+  private final int getPrior(int anchorId, boolean exclusive) {
-    if (anchorId == Snapshot.INVALID_ID) {
+    if (anchorId == Snapshot.CURRENT_STATE_ID) {
-      return getLastSnapshot();
+      return getLastSnapshotId();
     }
     final int i = Collections.binarySearch(diffs, anchorId);
     if (exclusive) { // must be the one before
       if (i == -1 || i == 0) {
-        return null;
+        return Snapshot.NO_SNAPSHOT_ID;
       } else {
         int priorIndex = i > 0 ? i - 1 : -i - 2;
-        return diffs.get(priorIndex).getSnapshot();
+        return diffs.get(priorIndex).getSnapshotId();
       }
     } else { // the one, or the one before if not existing
       if (i >= 0) {
-        return diffs.get(i).getSnapshot();
+        return diffs.get(i).getSnapshotId();
       } else if (i < -1) {
-        return diffs.get(-i - 2).getSnapshot();
+        return diffs.get(-i - 2).getSnapshotId();
       } else { // i == -1
-        return null;
+        return Snapshot.NO_SNAPSHOT_ID;
       }
     }
   }
   
-  public final Snapshot getPrior(int snapshotId) {
+  public final int getPrior(int snapshotId) {
     return getPrior(snapshotId, false);
   }
   
   /**
    * Update the prior snapshot.
    */
-  final Snapshot updatePrior(Snapshot snapshot, Snapshot prior) {
+  final int updatePrior(int snapshot, int prior) {
-    int id = snapshot == null ? Snapshot.INVALID_ID : snapshot.getId();
+    int p = getPrior(snapshot, true);
-    Snapshot s = getPrior(id, true);
+    if (p != Snapshot.CURRENT_STATE_ID
-    if (s != null && 
+        && Snapshot.ID_INTEGER_COMPARATOR.compare(p, prior) > 0) {
-        (prior == null || Snapshot.ID_COMPARATOR.compare(s, prior) > 0)) {
+      return p;
-      return s;
     }
     return prior;
   }
-
-  /**
-   * @return the diff corresponding to the given snapshot.
-   *         When the diff is null, it means that the current state and
-   *         the corresponding snapshot state are the same. 
-   */
-  public final D getDiff(Snapshot snapshot) {
-    return getDiffById(snapshot == null ? 
-        Snapshot.INVALID_ID : snapshot.getId());
-  }
   
-  private final D getDiffById(final int snapshotId) {
+  public final D getDiffById(final int snapshotId) {
-    if (snapshotId == Snapshot.INVALID_ID) {
+    if (snapshotId == Snapshot.CURRENT_STATE_ID) {
       return null;
     }
     final int i = Collections.binarySearch(diffs, snapshotId);
@@ -234,9 +223,9 @@ abstract class AbstractINodeDiffList<N extends INode,
    * Search for the snapshot whose id is 1) no less than the given id, 
    * and 2) most close to the given id.
    */
-  public final Snapshot getSnapshotById(final int snapshotId) {
+  public final int getSnapshotById(final int snapshotId) {
     D diff = getDiffById(snapshotId);
-    return diff == null ? null : diff.getSnapshot();
+    return diff == null ? Snapshot.CURRENT_STATE_ID : diff.getSnapshotId();
   }
   
   /**
@@ -271,8 +260,8 @@ abstract class AbstractINodeDiffList<N extends INode,
    *         Note that the current inode is returned if there is no change
    *         between the given snapshot and the current state. 
    */
-  public A getSnapshotINode(final Snapshot snapshot, final A currentINode) {
+  public A getSnapshotINode(final int snapshotId, final A currentINode) {
-    final D diff = getDiff(snapshot);
+    final D diff = getDiffById(snapshotId);
     final A inode = diff == null? null: diff.getSnapshotINode();
     return inode == null? currentINode: inode;
   }
@@ -281,15 +270,16 @@ abstract class AbstractINodeDiffList<N extends INode,
    * Check if the latest snapshot diff exists.  If not, add it.
    * @return the latest snapshot diff, which is never null.
    */
-  final D checkAndAddLatestSnapshotDiff(Snapshot latest, N currentINode)
+  final D checkAndAddLatestSnapshotDiff(int latestSnapshotId, N currentINode)
       throws QuotaExceededException {
     final D last = getLast();
     if (last != null
-        && Snapshot.ID_COMPARATOR.compare(last.getSnapshot(), latest) >= 0) {
+        && Snapshot.ID_INTEGER_COMPARATOR.compare(last.getSnapshotId(),
+            latestSnapshotId) >= 0) {
       return last;
     } else {
       try {
-        return addDiff(latest, currentINode);
+        return addDiff(latestSnapshotId, currentINode);
       } catch(NSQuotaExceededException e) {
         e.setMessagePrefix("Failed to record modification for snapshot");
         throw e;
@@ -298,10 +288,10 @@ abstract class AbstractINodeDiffList<N extends INode,
   }
 
   /** Save the snapshot copy to the latest snapshot. */
-  public void saveSelf2Snapshot(Snapshot latest, N currentINode, A snapshotCopy)
+  public void saveSelf2Snapshot(int latestSnapshotId, N currentINode,
-      throws QuotaExceededException {
+      A snapshotCopy) throws QuotaExceededException {
-    if (latest != null) {
+    if (latestSnapshotId != Snapshot.CURRENT_STATE_ID) {
-      D diff = checkAndAddLatestSnapshotDiff(latest, currentINode);
+      D diff = checkAndAddLatestSnapshotDiff(latestSnapshotId, currentINode);
       if (diff.snapshotINode == null) {
         if (snapshotCopy == null) {
           snapshotCopy = createSnapshotCopy(currentINode);

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

@@ -225,30 +225,36 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
     private final int childrenSize;
     /** The children list diff. */
     private final ChildrenDiff diff;
+    private boolean isSnapshotRoot = false;
+    
+    private DirectoryDiff(int snapshotId, INodeDirectory dir) {
+      super(snapshotId, null, null);
 
-    private DirectoryDiff(Snapshot snapshot, INodeDirectory dir) {
+      this.childrenSize = dir.getChildrenList(Snapshot.CURRENT_STATE_ID).size();
-      super(snapshot, null, null);
-
-      this.childrenSize = dir.getChildrenList(null).size();
       this.diff = new ChildrenDiff();
     }
 
     /** Constructor used by FSImage loading */
-    DirectoryDiff(Snapshot snapshot, INodeDirectoryAttributes snapshotINode,
+    DirectoryDiff(int snapshotId, INodeDirectoryAttributes snapshotINode,
-        DirectoryDiff posteriorDiff, int childrenSize,
+        DirectoryDiff posteriorDiff, int childrenSize, List<INode> createdList,
-        List<INode> createdList, List<INode> deletedList) {
+        List<INode> deletedList, boolean isSnapshotRoot) {
-      super(snapshot, snapshotINode, posteriorDiff);
+      super(snapshotId, snapshotINode, posteriorDiff);
       this.childrenSize = childrenSize;
       this.diff = new ChildrenDiff(createdList, deletedList);
+      this.isSnapshotRoot = isSnapshotRoot;
     }
 
     ChildrenDiff getChildrenDiff() {
       return diff;
     }
-
+    
-    /** Is the inode the root of the snapshot? */
+    void setSnapshotRoot(INodeDirectoryAttributes root) {
+      this.snapshotINode = root;
+      this.isSnapshotRoot = true;
+    }
+    
     boolean isSnapshotRoot() {
-      return snapshotINode == snapshot.getRoot();
+      return isSnapshotRoot;
     }
 
     @Override
@@ -287,7 +293,7 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
               combined.combinePosterior(d.diff, null);
             }
             children = combined.apply2Current(ReadOnlyList.Util.asList(
-                currentDir.getChildrenList(null)));
+                currentDir.getChildrenList(Snapshot.CURRENT_STATE_ID)));
           }
           return children;
         }
@@ -327,7 +333,7 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
           return null;
         } else if (d.getPosterior() == null) {
           // no more posterior diff, get from current inode.
-          return currentDir.getChild(name, null);
+          return currentDir.getChild(name, Snapshot.CURRENT_STATE_ID);
         }
       }
     }
@@ -342,11 +348,9 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
       writeSnapshot(out);
       out.writeInt(childrenSize);
 
-      // write snapshotINode
+      // Write snapshotINode
-      if (isSnapshotRoot()) {
+      out.writeBoolean(isSnapshotRoot);
-        out.writeBoolean(true);
+      if (!isSnapshotRoot) {
-      } else {
-        out.writeBoolean(false);
         if (snapshotINode != null) {
           out.writeBoolean(true);
           FSImageSerialization.writeINodeDirectoryAttributes(snapshotINode, out);
@@ -373,7 +377,7 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
       extends AbstractINodeDiffList<INodeDirectory, INodeDirectoryAttributes, DirectoryDiff> {
 
     @Override
-    DirectoryDiff createDiff(Snapshot snapshot, INodeDirectory currentDir) {
+    DirectoryDiff createDiff(int snapshot, INodeDirectory currentDir) {
       return new DirectoryDiff(snapshot, currentDir);
     }
 
@@ -424,12 +428,13 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
   /**
    * Destroy a subtree under a DstReference node.
    */
-  public static void destroyDstSubtree(INode inode, final Snapshot snapshot,
+  public static void destroyDstSubtree(INode inode, final int snapshot,
-      final Snapshot prior, final BlocksMapUpdateInfo collectedBlocks,
+      final int prior, final BlocksMapUpdateInfo collectedBlocks,
       final List<INode> removedINodes) throws QuotaExceededException {
-    Preconditions.checkArgument(prior != null);
+    Preconditions.checkArgument(prior != Snapshot.NO_SNAPSHOT_ID);
     if (inode.isReference()) {
-      if (inode instanceof INodeReference.WithName && snapshot != null) {
+      if (inode instanceof INodeReference.WithName
+          && snapshot != Snapshot.CURRENT_STATE_ID) {
         // this inode has been renamed before the deletion of the DstReference
         // subtree
         inode.cleanSubtree(snapshot, prior, collectedBlocks, removedINodes,
@@ -447,18 +452,18 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
       DirectoryWithSnapshotFeature sf = dir.getDirectoryWithSnapshotFeature();
       if (sf != null) {
         DirectoryDiffList diffList = sf.getDiffs();
-        DirectoryDiff priorDiff = diffList.getDiff(prior);
+        DirectoryDiff priorDiff = diffList.getDiffById(prior);
-        if (priorDiff != null && priorDiff.getSnapshot().equals(prior)) {
+        if (priorDiff != null && priorDiff.getSnapshotId() == prior) {
           List<INode> dList = priorDiff.diff.getList(ListType.DELETED);
           excludedNodes = cloneDiffList(dList);
         }
         
-        if (snapshot != null) {
+        if (snapshot != Snapshot.CURRENT_STATE_ID) {
           diffList.deleteSnapshotDiff(snapshot, prior, dir, collectedBlocks,
               removedINodes, true);
         }
-        priorDiff = diffList.getDiff(prior);
+        priorDiff = diffList.getDiffById(prior);
-        if (priorDiff != null && priorDiff.getSnapshot().equals(prior)) {
+        if (priorDiff != null && priorDiff.getSnapshotId() == prior) {
           priorDiff.diff.destroyCreatedList(dir, collectedBlocks,
               removedINodes);
         }
@@ -478,14 +483,14 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
    * deleted list of prior.
    * @param inode The inode to clean.
    * @param post The post snapshot.
-   * @param prior The prior snapshot.
+   * @param prior The id of the prior snapshot.
    * @param collectedBlocks Used to collect blocks for later deletion.
    * @return Quota usage update.
    */
   private static Quota.Counts cleanDeletedINode(INode inode,
-      final Snapshot post, final Snapshot prior,
+      final int post, final int prior,
       final BlocksMapUpdateInfo collectedBlocks,
-      final List<INode> removedINodes, final boolean countDiffChange) 
+      final List<INode> removedINodes, final boolean countDiffChange)
       throws QuotaExceededException {
     Quota.Counts counts = Quota.Counts.newInstance();
     Deque<INode> queue = new ArrayDeque<INode>();
@@ -494,7 +499,7 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
       INode topNode = queue.pollFirst();
       if (topNode instanceof INodeReference.WithName) {
         INodeReference.WithName wn = (INodeReference.WithName) topNode;
-        if (wn.getLastSnapshotId() >= post.getId()) {
+        if (wn.getLastSnapshotId() >= post) {
           wn.cleanSubtree(post, prior, collectedBlocks, removedINodes,
               countDiffChange);
         }
@@ -511,8 +516,8 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
         if (sf != null) {
           // delete files/dirs created after prior. Note that these
           // files/dirs, along with inode, were deleted right after post.
-          DirectoryDiff priorDiff = sf.getDiffs().getDiff(prior);
+          DirectoryDiff priorDiff = sf.getDiffs().getDiffById(prior);
-          if (priorDiff != null && priorDiff.getSnapshot().equals(prior)) {
+          if (priorDiff != null && priorDiff.getSnapshotId() == prior) {
             priorChildrenDiff = priorDiff.getChildrenDiff();
             counts.add(priorChildrenDiff.destroyCreatedList(dir,
                 collectedBlocks, removedINodes));
@@ -540,8 +545,8 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
   }
 
   /** @return the last snapshot. */
-  public Snapshot getLastSnapshot() {
+  public int getLastSnapshotId() {
-    return diffs.getLastSnapshot();
+    return diffs.getLastSnapshotId();
   }
 
   /** @return the snapshot diff list. */
@@ -565,11 +570,13 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
    * to make sure that parent is in the given snapshot "latest".
    */
   public boolean addChild(INodeDirectory parent, INode inode,
-      boolean setModTime, Snapshot latest) throws QuotaExceededException {
+      boolean setModTime, int latestSnapshotId) throws QuotaExceededException {
-    ChildrenDiff diff = diffs.checkAndAddLatestSnapshotDiff(latest, parent).diff;
+    ChildrenDiff diff = diffs.checkAndAddLatestSnapshotDiff(latestSnapshotId,
+        parent).diff;
     int undoInfo = diff.create(inode);
 
-    final boolean added = parent.addChild(inode, setModTime, null);
+    final boolean added = parent.addChild(inode, setModTime,
+        Snapshot.CURRENT_STATE_ID);
     if (!added) {
       diff.undoCreate(inode, undoInfo);
     }
@@ -581,7 +588,7 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
    * needs to make sure that parent is in the given snapshot "latest".
    */
   public boolean removeChild(INodeDirectory parent, INode child,
-      Snapshot latest) throws QuotaExceededException {
+      int latestSnapshotId) throws QuotaExceededException {
     // For a directory that is not a renamed node, if isInLatestSnapshot returns
     // false, the directory is not in the latest snapshot, thus we do not need
     // to record the removed child in any snapshot.
@@ -593,7 +600,8 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
     // directory node cannot be in any snapshot (not in current tree, nor in
     // previous src tree). Thus we do not need to record the removed child in
     // any snapshot.
-    ChildrenDiff diff = diffs.checkAndAddLatestSnapshotDiff(latest, parent).diff;
+    ChildrenDiff diff = diffs.checkAndAddLatestSnapshotDiff(latestSnapshotId,
+        parent).diff;
     UndoInfo<INode> undoInfo = diff.delete(child);
 
     final boolean removed = parent.removeChild(child);
@@ -611,29 +619,29 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
    *         for the snapshot and return it. 
    */
   public ReadOnlyList<INode> getChildrenList(INodeDirectory currentINode,
-      final Snapshot snapshot) {
+      final int snapshotId) {
-    final DirectoryDiff diff = diffs.getDiff(snapshot);
+    final DirectoryDiff diff = diffs.getDiffById(snapshotId);
     return diff != null ? diff.getChildrenList(currentINode) : currentINode
-        .getChildrenList(null);
+        .getChildrenList(Snapshot.CURRENT_STATE_ID);
   }
   
   public INode getChild(INodeDirectory currentINode, byte[] name,
-      Snapshot snapshot) {
+      int snapshotId) {
-    final DirectoryDiff diff = diffs.getDiff(snapshot);
+    final DirectoryDiff diff = diffs.getDiffById(snapshotId);
     return diff != null ? diff.getChild(name, true, currentINode)
-        : currentINode.getChild(name, null);
+        : currentINode.getChild(name, Snapshot.CURRENT_STATE_ID);
   }
   
   /** Used to record the modification of a symlink node */
   public INode saveChild2Snapshot(INodeDirectory currentINode,
-      final INode child, final Snapshot latest, final INode snapshotCopy)
+      final INode child, final int latestSnapshotId, final INode snapshotCopy)
       throws QuotaExceededException {
     Preconditions.checkArgument(!child.isDirectory(),
         "child is a directory, child=%s", child);
-    Preconditions.checkArgument(latest != null);
+    Preconditions.checkArgument(latestSnapshotId != Snapshot.CURRENT_STATE_ID);
     
-    final DirectoryDiff diff = diffs.checkAndAddLatestSnapshotDiff(latest,
+    final DirectoryDiff diff = diffs.checkAndAddLatestSnapshotDiff(
-        currentINode);
+        latestSnapshotId, currentINode);
     if (diff.getChild(child.getLocalNameBytes(), false, currentINode) != null) {
       // it was already saved in the latest snapshot earlier.  
       return child;
@@ -656,7 +664,7 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
   public Quota.Counts computeQuotaUsage4CurrentDirectory(Quota.Counts counts) {
     for(DirectoryDiff d : diffs) {
       for(INode deleted : d.getChildrenDiff().getList(ListType.DELETED)) {
-        deleted.computeQuotaUsage(counts, false, Snapshot.INVALID_ID);
+        deleted.computeQuotaUsage(counts, false, Snapshot.CURRENT_STATE_ID);
       }
     }
     counts.add(Quota.NAMESPACE, diffs.asList().size());
@@ -744,14 +752,14 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
   }
 
   public Quota.Counts cleanDirectory(final INodeDirectory currentINode,
-      final Snapshot snapshot, Snapshot prior,
+      final int snapshot, int prior,
       final BlocksMapUpdateInfo collectedBlocks,
       final List<INode> removedINodes, final boolean countDiffChange)
       throws QuotaExceededException {
     Quota.Counts counts = Quota.Counts.newInstance();
     Map<INode, INode> priorCreated = null;
     Map<INode, INode> priorDeleted = null;
-    if (snapshot == null) { // delete the current directory
+    if (snapshot == Snapshot.CURRENT_STATE_ID) { // delete the current directory
       currentINode.recordModification(prior);
       // delete everything in created list
       DirectoryDiff lastDiff = diffs.getLast();
@@ -764,9 +772,9 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
       prior = getDiffs().updatePrior(snapshot, prior);
       // if there is a snapshot diff associated with prior, we need to record
       // its original created and deleted list before deleting post
-      if (prior != null) {
+      if (prior != Snapshot.NO_SNAPSHOT_ID) {
-        DirectoryDiff priorDiff = this.getDiffs().getDiff(prior);
+        DirectoryDiff priorDiff = this.getDiffs().getDiffById(prior);
-        if (priorDiff != null && priorDiff.getSnapshot().equals(prior)) {
+        if (priorDiff != null && priorDiff.getSnapshotId() == prior) {
           List<INode> cList = priorDiff.diff.getList(ListType.CREATED);
           List<INode> dList = priorDiff.diff.getList(ListType.DELETED);
           priorCreated = cloneDiffList(cList);
@@ -774,13 +782,13 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
         }
       }
       
-      counts.add(getDiffs().deleteSnapshotDiff(snapshot, prior, currentINode, 
+      counts.add(getDiffs().deleteSnapshotDiff(snapshot, prior,
-          collectedBlocks, removedINodes, countDiffChange));
+          currentINode, collectedBlocks, removedINodes, countDiffChange));
       
       // check priorDiff again since it may be created during the diff deletion
-      if (prior != null) {
+      if (prior != Snapshot.NO_SNAPSHOT_ID) {
-        DirectoryDiff priorDiff = this.getDiffs().getDiff(prior);
+        DirectoryDiff priorDiff = this.getDiffs().getDiffById(prior);
-        if (priorDiff != null && priorDiff.getSnapshot().equals(prior)) {
+        if (priorDiff != null && priorDiff.getSnapshotId() == prior) {
           // For files/directories created between "prior" and "snapshot", 
           // we need to clear snapshot copies for "snapshot". Note that we must
           // use null as prior in the cleanSubtree call. Files/directories that
@@ -791,8 +799,8 @@ public class DirectoryWithSnapshotFeature implements INode.Feature {
             for (INode cNode : priorDiff.getChildrenDiff().getList(
                 ListType.CREATED)) {
               if (priorCreated.containsKey(cNode)) {
-                counts.add(cNode.cleanSubtree(snapshot, null, collectedBlocks,
+                counts.add(cNode.cleanSubtree(snapshot, Snapshot.NO_SNAPSHOT_ID,
-                    removedINodes, countDiffChange));
+                    collectedBlocks, removedINodes, countDiffChange));
               }
             }
           }

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

@@ -38,15 +38,15 @@ public class FileDiff extends
   /** The file size at snapshot creation time. */
   private final long fileSize;
 
-  FileDiff(Snapshot snapshot, INodeFile file) {
+  FileDiff(int snapshotId, INodeFile file) {
-    super(snapshot, null, null);
+    super(snapshotId, null, null);
     fileSize = file.computeFileSize();
   }
 
   /** Constructor used by FSImage loading */
-  FileDiff(Snapshot snapshot, INodeFileAttributes snapshotINode,
+  FileDiff(int snapshotId, INodeFileAttributes snapshotINode,
       FileDiff posteriorDiff, long fileSize) {
-    super(snapshot, snapshotINode, posteriorDiff);
+    super(snapshotId, snapshotINode, posteriorDiff);
     this.fileSize = fileSize;
   }
 

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

@@ -25,8 +25,8 @@ public class FileDiffList extends
     AbstractINodeDiffList<INodeFile, INodeFileAttributes, FileDiff> {
   
   @Override
-  FileDiff createDiff(Snapshot snapshot, INodeFile file) {
+  FileDiff createDiff(int snapshotId, INodeFile file) {
-    return new FileDiff(snapshot, file);
+    return new FileDiff(snapshotId, file);
   }
   
   @Override

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

@@ -78,22 +78,22 @@ public class FileWithSnapshotFeature implements INode.Feature {
     return (isCurrentFileDeleted()? "(DELETED), ": ", ") + diffs;
   }
   
-  public Quota.Counts cleanFile(final INodeFile file, final Snapshot snapshot,
+  public Quota.Counts cleanFile(final INodeFile file, final int snapshotId,
-      Snapshot prior, final BlocksMapUpdateInfo collectedBlocks,
+      int priorSnapshotId, final BlocksMapUpdateInfo collectedBlocks,
       final List<INode> removedINodes, final boolean countDiffChange)
       throws QuotaExceededException {
-    if (snapshot == null) {
+    if (snapshotId == Snapshot.CURRENT_STATE_ID) {
       // delete the current file while the file has snapshot feature
       if (!isCurrentFileDeleted()) {
-        file.recordModification(prior);
+        file.recordModification(priorSnapshotId);
         deleteCurrentFile();
       }
       collectBlocksAndClear(file, collectedBlocks, removedINodes);
       return Quota.Counts.newInstance();
     } else { // delete the snapshot
-      prior = getDiffs().updatePrior(snapshot, prior);
+      priorSnapshotId = getDiffs().updatePrior(snapshotId, priorSnapshotId);
-      return diffs.deleteSnapshotDiff(snapshot, prior, file, collectedBlocks,
+      return diffs.deleteSnapshotDiff(snapshotId, priorSnapshotId, file,
-          removedINodes, countDiffChange);
+          collectedBlocks, removedINodes, countDiffChange);
     }
   }
   

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

@@ -206,6 +206,15 @@ public class INodeDirectorySnapshottable extends INodeDirectory {
     return i < 0? null: snapshotsByNames.get(i);
   }
   
+  Snapshot getSnapshotById(int sid) {
+    for (Snapshot s : snapshotsByNames) {
+      if (s.getId() == sid) {
+        return s;
+      }
+    }
+    return null;
+  }
+  
   /** @return {@link #snapshotsByNames} as a {@link ReadOnlyList} */
   public ReadOnlyList<Snapshot> getSnapshotList() {
     return ReadOnlyList.Util.asReadOnlyList(snapshotsByNames);
@@ -297,13 +306,14 @@ public class INodeDirectorySnapshottable extends INodeDirectory {
           + "snapshot with the same name \"" + Snapshot.getSnapshotName(s) + "\".");
     }
 
-    final DirectoryDiff d = getDiffs().addDiff(s, this);
+    final DirectoryDiff d = getDiffs().addDiff(id, this);
-    d.snapshotINode = s.getRoot();
+    d.setSnapshotRoot(s.getRoot());
     snapshotsByNames.add(-i - 1, s);
 
     //set modification time
-    updateModificationTime(Time.now(), null);
+    updateModificationTime(Time.now(), Snapshot.CURRENT_STATE_ID);
-    s.getRoot().setModificationTime(getModificationTime(), null);
+    s.getRoot().setModificationTime(getModificationTime(),
+        Snapshot.CURRENT_STATE_ID);
     return s;
   }
   
@@ -326,10 +336,10 @@ public class INodeDirectorySnapshottable extends INodeDirectory {
           + ": the snapshot does not exist.");
     } else {
       final Snapshot snapshot = snapshotsByNames.get(i);
-      Snapshot prior = Snapshot.findLatestSnapshot(this, snapshot);
+      int prior = Snapshot.findLatestSnapshot(this, snapshot.getId());
       try {
-        Quota.Counts counts = cleanSubtree(snapshot, prior, collectedBlocks,
+        Quota.Counts counts = cleanSubtree(snapshot.getId(), prior,
-            removedINodes, true);
+            collectedBlocks, removedINodes, true);
         INodeDirectory parent = getParent();
         if (parent != null) {
           // there will not be any WithName node corresponding to the deleted 
@@ -425,8 +435,9 @@ public class INodeDirectorySnapshottable extends INodeDirectory {
           diffReport.addDirDiff(dir, relativePath, diff);
         }
       }
-      ReadOnlyList<INode> children = dir.getChildrenList(diffReport
+      ReadOnlyList<INode> children = dir.getChildrenList(
-          .isFromEarlier() ? diffReport.to : diffReport.from);
+          diffReport.isFromEarlier() ? Snapshot.getSnapshotId(diffReport.to) : 
+            Snapshot.getSnapshotId(diffReport.from));
       for (INode child : children) {
         final byte[] name = child.getLocalNameBytes();
         if (diff.searchIndex(ListType.CREATED, name) < 0
@@ -454,16 +465,15 @@ public class INodeDirectorySnapshottable extends INodeDirectory {
    * Replace itself with {@link INodeDirectoryWithSnapshot} or
    * {@link INodeDirectory} depending on the latest snapshot.
    */
-  INodeDirectory replaceSelf(final Snapshot latest, final INodeMap inodeMap)
+  INodeDirectory replaceSelf(final int latestSnapshotId, final INodeMap inodeMap)
       throws QuotaExceededException {
-    if (latest == null) {
+    if (latestSnapshotId == Snapshot.CURRENT_STATE_ID) {
-      Preconditions.checkState(
+      Preconditions.checkState(getDirectoryWithSnapshotFeature()
-          getDirectoryWithSnapshotFeature().getLastSnapshot() == null,
+          .getLastSnapshotId() == Snapshot.CURRENT_STATE_ID, "this=%s", this);
-          "latest == null but getLastSnapshot() != null, this=%s", this);
     }
     INodeDirectory dir = replaceSelf4INodeDirectory(inodeMap);
-    if (latest != null) {
+    if (latestSnapshotId != Snapshot.CURRENT_STATE_ID) {
-      dir.recordModification(latest);
+      dir.recordModification(latestSnapshotId);
     }
     return dir;
   }
@@ -475,10 +485,10 @@ public class INodeDirectorySnapshottable extends INodeDirectory {
 
   @Override
   public void dumpTreeRecursively(PrintWriter out, StringBuilder prefix,
-      Snapshot snapshot) {
+      int snapshot) {
     super.dumpTreeRecursively(out, prefix, snapshot);
 
-    if (snapshot == null) {
+    if (snapshot == Snapshot.CURRENT_STATE_ID) {
       out.println();
       out.print(prefix);
 
@@ -494,7 +504,8 @@ public class INodeDirectorySnapshottable extends INodeDirectory {
           n++;
         }
       }
-      Preconditions.checkState(n == snapshotsByNames.size());
+      Preconditions.checkState(n == snapshotsByNames.size(), "#n=" + n
+          + ", snapshotsByNames.size()=" + snapshotsByNames.size());
       out.print(", #snapshot=");
       out.println(n);
 
@@ -522,8 +533,9 @@ public class INodeDirectorySnapshottable extends INodeDirectory {
   
             @Override
             public SnapshotAndINode next() {
-              final Snapshot s = next.snapshot;
+              final SnapshotAndINode pair = new SnapshotAndINode(next
-              final SnapshotAndINode pair = new SnapshotAndINode(s);
+                  .getSnapshotId(), getSnapshotById(next.getSnapshotId())
+                  .getRoot());
               next = findNext();
               return pair;
             }

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

@@ -37,7 +37,11 @@ import org.apache.hadoop.hdfs.util.ReadOnlyList;
 /** Snapshot of a sub-tree in the namesystem. */
 @InterfaceAudience.Private
 public class Snapshot implements Comparable<byte[]> {
-  public static final int INVALID_ID = -1;
+  /**
+   * This id is used to indicate the current state (vs. snapshots)
+   */
+  public static final int CURRENT_STATE_ID = Integer.MAX_VALUE - 1;
+  public static final int NO_SNAPSHOT_ID = -1;
   
   /**
    * The pattern for generating the default snapshot name.
@@ -61,14 +65,18 @@ public class Snapshot implements Comparable<byte[]> {
         .toString();
   }
   
-  /** 
+  /**
-   * Get the name of the given snapshot. 
+   * Get the name of the given snapshot.
    * @param s The given snapshot.
    * @return The name of the snapshot, or an empty string if {@code s} is null
    */
   static String getSnapshotName(Snapshot s) {
     return s != null ? s.getRoot().getLocalName() : "";
   }
+  
+  public static int getSnapshotId(Snapshot s) {
+    return s == null ? CURRENT_STATE_ID : s.getId();
+  }
 
   /**
    * Compare snapshot with IDs, where null indicates the current status thus
@@ -78,9 +86,8 @@ public class Snapshot implements Comparable<byte[]> {
       = new Comparator<Snapshot>() {
     @Override
     public int compare(Snapshot left, Snapshot right) {
-      return ID_INTEGER_COMPARATOR.compare(
+      return ID_INTEGER_COMPARATOR.compare(Snapshot.getSnapshotId(left),
-          left == null? null: left.getId(),
+          Snapshot.getSnapshotId(right));
-          right == null? null: right.getId());
     }
   };
 
@@ -92,12 +99,9 @@ public class Snapshot implements Comparable<byte[]> {
       = new Comparator<Integer>() {
     @Override
     public int compare(Integer left, Integer right) {
-      // null means the current state, thus should be the largest
+      // Snapshot.CURRENT_STATE_ID means the current state, thus should be the 
-      if (left == null) {
+      // largest
-        return right == null? 0: 1;
+      return left - right;
-      } else {
-        return right == null? -1: left - right; 
-      }
     }
   };
 
@@ -108,12 +112,12 @@ public class Snapshot implements Comparable<byte[]> {
    * is not null).
    * 
    * @param inode the given inode that the returned snapshot needs to cover
-   * @param anchor the returned snapshot should be taken before this snapshot.
+   * @param anchor the returned snapshot should be taken before this given id.
-   * @return the latest snapshot covers the given inode and was taken before the
+   * @return id of the latest snapshot that covers the given inode and was taken 
-   *         the given snapshot (if it is not null).
+   *         before the the given snapshot (if it is not null).
    */
-  public static Snapshot findLatestSnapshot(INode inode, Snapshot anchor) {
+  public static int findLatestSnapshot(INode inode, final int anchor) {
-    Snapshot latest = null;
+    int latest = NO_SNAPSHOT_ID;
     for(; inode != null; inode = inode.getParent()) {
       if (inode.isDirectory()) {
         final INodeDirectory dir = inode.asDirectory();
@@ -139,13 +143,13 @@ public class Snapshot implements Comparable<byte[]> {
     }
 
     @Override
-    public ReadOnlyList<INode> getChildrenList(Snapshot snapshot) {
+    public ReadOnlyList<INode> getChildrenList(int snapshotId) {
-      return getParent().getChildrenList(snapshot);
+      return getParent().getChildrenList(snapshotId);
     }
 
     @Override
-    public INode getChild(byte[] name, Snapshot snapshot) {
+    public INode getChild(byte[] name, int snapshotId) {
-      return getParent().getChild(name, snapshot);
+      return getParent().getChild(name, snapshotId);
     }
     
     @Override

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

@@ -118,7 +118,7 @@ public class SnapshotFSImageFormat {
 
   private static FileDiff loadFileDiff(FileDiff posterior, DataInput in,
       FSImageFormat.Loader loader) throws IOException {
-    // 1. Read the full path of the Snapshot root to identify the Snapshot
+    // 1. Read the id of the Snapshot root to identify the Snapshot
     final Snapshot snapshot = loader.getSnapshot(in);
 
     // 2. Load file size
@@ -128,7 +128,7 @@ public class SnapshotFSImageFormat {
     final INodeFileAttributes snapshotINode = in.readBoolean()?
         loader.loadINodeFileAttributes(in): null;
     
-    return new FileDiff(snapshot, snapshotINode, posterior, fileSize);
+    return new FileDiff(snapshot.getId(), snapshotINode, posterior, fileSize);
   }
 
   /**
@@ -149,7 +149,8 @@ public class SnapshotFSImageFormat {
       } // else go to the next SnapshotDiff
     } 
     // use the current child
-    INode currentChild = parent.getChild(createdNodeName, null);
+    INode currentChild = parent.getChild(createdNodeName,
+        Snapshot.CURRENT_STATE_ID);
     if (currentChild == null) {
       throw new IOException("Cannot find an INode associated with the INode "
           + DFSUtil.bytes2String(createdNodeName)
@@ -295,9 +296,9 @@ public class SnapshotFSImageFormat {
     
     // 6. Compose the SnapshotDiff
     List<DirectoryDiff> diffs = parent.getDiffs().asList();
-    DirectoryDiff sdiff = new DirectoryDiff(snapshot, snapshotINode,
+    DirectoryDiff sdiff = new DirectoryDiff(snapshot.getId(), snapshotINode,
-        diffs.isEmpty() ? null : diffs.get(0),
+        diffs.isEmpty() ? null : diffs.get(0), childrenSize, createdList,
-        childrenSize, createdList, deletedList);
+        deletedList, snapshotINode == snapshot.getRoot());
     return sdiff;
   }
   

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

@@ -114,7 +114,7 @@ public class SnapshotManager implements SnapshotStats {
       s = (INodeDirectorySnapshottable)d; 
       s.setSnapshotQuota(INodeDirectorySnapshottable.SNAPSHOT_LIMIT);
     } else {
-      s = d.replaceSelf4INodeDirectorySnapshottable(iip.getLatestSnapshot(),
+      s = d.replaceSelf4INodeDirectorySnapshottable(iip.getLatestSnapshotId(),
           fsdir.getINodeMap());
     }
     addSnapshottable(s);
@@ -160,7 +160,7 @@ public class SnapshotManager implements SnapshotStats {
     if (s == fsdir.getRoot()) {
       s.setSnapshotQuota(0); 
     } else {
-      s.replaceSelf(iip.getLatestSnapshot(), fsdir.getINodeMap());
+      s.replaceSelf(iip.getLatestSnapshotId(), fsdir.getINodeMap());
     }
     removeSnapshottable(s);
   }
@@ -324,7 +324,8 @@ public class SnapshotManager implements SnapshotStats {
         SnapshottableDirectoryStatus status = new SnapshottableDirectoryStatus(
             dir.getModificationTime(), dir.getAccessTime(),
             dir.getFsPermission(), dir.getUserName(), dir.getGroupName(),
-            dir.getLocalNameBytes(), dir.getId(), dir.getChildrenNum(null),
+            dir.getLocalNameBytes(), dir.getId(), 
+            dir.getChildrenNum(Snapshot.CURRENT_STATE_ID),
             dir.getNumSnapshots(),
             dir.getSnapshotQuota(), dir.getParent() == null ? 
                 DFSUtil.EMPTY_BYTES : 

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

@@ -29,6 +29,7 @@ import org.apache.hadoop.fs.Path;
 import org.apache.hadoop.hdfs.DFSTestUtil;
 import org.apache.hadoop.hdfs.DistributedFileSystem;
 import org.apache.hadoop.hdfs.MiniDFSCluster;
+import org.apache.hadoop.hdfs.server.namenode.snapshot.Snapshot;
 import org.junit.After;
 import org.junit.Assert;
 import org.junit.Before;
@@ -119,7 +120,7 @@ public class TestFSDirectory {
     fsdir.reset();
     Assert.assertFalse(fsdir.isReady());
     final INodeDirectory root = (INodeDirectory) fsdir.getINode("/");
-    Assert.assertTrue(root.getChildrenList(null).isEmpty());
+    Assert.assertTrue(root.getChildrenList(Snapshot.CURRENT_STATE_ID).isEmpty());
     fsdir.imageLoadComplete();
     Assert.assertTrue(fsdir.isReady());
   }

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

@@ -34,6 +34,7 @@ import org.apache.hadoop.fs.FileStatus;
 import org.apache.hadoop.fs.Path;
 import org.apache.hadoop.fs.permission.FsPermission;
 import org.apache.hadoop.hdfs.DFSTestUtil;
+import org.apache.hadoop.hdfs.DFSUtil;
 import org.apache.hadoop.hdfs.DistributedFileSystem;
 import org.apache.hadoop.hdfs.MiniDFSCluster;
 import org.apache.hadoop.hdfs.client.HdfsDataOutputStream;
@@ -43,6 +44,7 @@ import org.apache.hadoop.hdfs.protocol.SnapshottableDirectoryStatus;
 import org.apache.hadoop.hdfs.server.namenode.NNStorage.NameNodeFile;
 import org.apache.hadoop.hdfs.server.namenode.snapshot.DirectoryWithSnapshotFeature.DirectoryDiff;
 import org.apache.hadoop.hdfs.server.namenode.snapshot.INodeDirectorySnapshottable;
+import org.apache.hadoop.hdfs.server.namenode.snapshot.Snapshot;
 import org.apache.hadoop.hdfs.server.namenode.snapshot.SnapshotTestHelper;
 import org.apache.hadoop.hdfs.util.Canceler;
 import org.apache.log4j.Level;
@@ -195,11 +197,12 @@ public class TestFSImageWithSnapshot {
     INodeDirectorySnapshottable rootNode = 
         (INodeDirectorySnapshottable) fsn.dir.getINode4Write(root.toString());
     assertTrue("The children list of root should be empty", 
-        rootNode.getChildrenList(null).isEmpty());
+        rootNode.getChildrenList(Snapshot.CURRENT_STATE_ID).isEmpty());
     // one snapshot on root: s1
     List<DirectoryDiff> diffList = rootNode.getDiffs().asList();
     assertEquals(1, diffList.size());
-    assertEquals("s1", diffList.get(0).getSnapshot().getRoot().getLocalName());
+    Snapshot s1 = rootNode.getSnapshot(DFSUtil.string2Bytes("s1"));
+    assertEquals(s1.getId(), diffList.get(0).getSnapshotId());
     
     // check SnapshotManager's snapshottable directory list
     assertEquals(1, fsn.getSnapshotManager().getNumSnapshottableDirs());

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

@@ -60,6 +60,7 @@ import org.apache.hadoop.hdfs.protocol.LocatedBlock;
 import org.apache.hadoop.hdfs.protocol.LocatedBlocks;
 import org.apache.hadoop.hdfs.protocol.QuotaExceededException;
 import org.apache.hadoop.hdfs.server.blockmanagement.BlockInfo;
+import org.apache.hadoop.hdfs.server.namenode.snapshot.Snapshot;
 import org.apache.hadoop.hdfs.server.protocol.NamenodeProtocols;
 import org.apache.hadoop.io.IOUtils;
 import org.apache.hadoop.util.Time;
@@ -781,7 +782,7 @@ public class TestINodeFile {
       }
       System.out.println("Adding component " + DFSUtil.bytes2String(component));
       dir = new INodeDirectory(++id, component, permstatus, 0);
-      prev.addChild(dir, false, null);
+      prev.addChild(dir, false, Snapshot.CURRENT_STATE_ID);
       prev = dir;
     }
     return dir; // Last Inode in the chain

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

@@ -123,8 +123,12 @@ public class TestSnapshotPathINodes {
       final Snapshot snapshot, int index) {
     assertEquals(isSnapshot, inodesInPath.isSnapshot());
     assertEquals(index, inodesInPath.getSnapshotRootIndex());
-    assertEquals(isSnapshot? snapshot: null, inodesInPath.getPathSnapshot());
+    assertEquals(Snapshot.getSnapshotId(isSnapshot ? snapshot : null),
-    assertEquals(isSnapshot? null: snapshot, inodesInPath.getLatestSnapshot());
+        inodesInPath.getPathSnapshotId());
+    if (!isSnapshot) {
+      assertEquals(Snapshot.getSnapshotId(snapshot),
+          inodesInPath.getLatestSnapshotId());
+    }
     if (isSnapshot && index >= 0) {
       assertEquals(Snapshot.Root.class, inodesInPath.getINodes()[index].getClass());
     }
@@ -424,7 +428,7 @@ public class TestSnapshotPathINodes {
     // The modification time of the snapshot INode should be the same with the
     // original INode before modification
     assertEquals(modTime,
-        snapshotFileNode.getModificationTime(ssNodesInPath.getPathSnapshot()));
+        snapshotFileNode.getModificationTime(ssNodesInPath.getPathSnapshotId()));
 
     // Check the INode for /TestSnapshot/sub1/file1 again
     names = INode.getPathNames(file1.toString());

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

@@ -270,7 +270,8 @@ public class SnapshotTestHelper {
 
   public static void dumpTree2File(FSDirectory fsdir, File f) throws IOException{
     final PrintWriter out = new PrintWriter(new FileWriter(f, false), true);
-    fsdir.getINode("/").dumpTreeRecursively(out, new StringBuilder(), null);
+    fsdir.getINode("/").dumpTreeRecursively(out, new StringBuilder(),
+        Snapshot.CURRENT_STATE_ID);
     out.close();
   }
 

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

@@ -156,7 +156,6 @@ public class TestINodeFileUnderConstructionWithSnapshot {
     INodeDirectorySnapshottable dirNode = (INodeDirectorySnapshottable) fsdir
         .getINode(dir.toString());
     DirectoryDiff last = dirNode.getDiffs().getLast();
-    Snapshot s0 = last.snapshot;
     
     // 2. append without closing stream
     out = appendFileWithoutClosing(file, BLOCKSIZE);
@@ -164,7 +163,7 @@ public class TestINodeFileUnderConstructionWithSnapshot {
     
     // re-check nodeInDeleted_S0
     dirNode = (INodeDirectorySnapshottable) fsdir.getINode(dir.toString());
-    assertEquals(BLOCKSIZE * 2, fileNode.computeFileSize(s0));
+    assertEquals(BLOCKSIZE * 2, fileNode.computeFileSize(last.getSnapshotId()));
     
     // 3. take snapshot --> close stream
     hdfs.createSnapshot(dir, "s1");
@@ -175,9 +174,8 @@ public class TestINodeFileUnderConstructionWithSnapshot {
     fileNode = (INodeFile) fsdir.getINode(file.toString());
     dirNode = (INodeDirectorySnapshottable) fsdir.getINode(dir.toString());
     last = dirNode.getDiffs().getLast();
-    Snapshot s1 = last.snapshot;
     assertTrue(fileNode.isWithSnapshot());
-    assertEquals(BLOCKSIZE * 3, fileNode.computeFileSize(s1));
+    assertEquals(BLOCKSIZE * 3, fileNode.computeFileSize(last.getSnapshotId()));
     
     // 4. modify file --> append without closing stream --> take snapshot -->
     // close stream
@@ -187,7 +185,7 @@ public class TestINodeFileUnderConstructionWithSnapshot {
     out.close();
     
     // re-check the size of nodeInDeleted_S1
-    assertEquals(BLOCKSIZE * 3, fileNode.computeFileSize(s1));
+    assertEquals(BLOCKSIZE * 3, fileNode.computeFileSize(last.getSnapshotId()));
   }
   
   /**

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

@@ -44,6 +44,7 @@ import org.apache.hadoop.fs.permission.FsPermission;
 import org.apache.hadoop.hdfs.DFSConfigKeys;
 import org.apache.hadoop.hdfs.DFSOutputStream;
 import org.apache.hadoop.hdfs.DFSTestUtil;
+import org.apache.hadoop.hdfs.DFSUtil;
 import org.apache.hadoop.hdfs.DistributedFileSystem;
 import org.apache.hadoop.hdfs.MiniDFSCluster;
 import org.apache.hadoop.hdfs.client.HdfsDataOutputStream.SyncFlag;
@@ -72,6 +73,7 @@ import org.junit.Assert;
 import org.junit.Before;
 import org.junit.Test;
 import org.mockito.Mockito;
+;
 
 /** Testing rename with snapshots. */
 public class TestRenameWithSnapshots {
@@ -402,9 +404,11 @@ public class TestRenameWithSnapshots {
     final Path foo_s3 = SnapshotTestHelper.getSnapshotPath(sdir1, "s3",
         "foo");
     assertFalse(hdfs.exists(foo_s3));
+    INodeDirectorySnapshottable sdir2Node = 
+        (INodeDirectorySnapshottable) fsdir.getINode(sdir2.toString());
+    Snapshot s2 = sdir2Node.getSnapshot(DFSUtil.string2Bytes("s2"));
     INodeFile sfoo = fsdir.getINode(newfoo.toString()).asFile();
-    assertEquals("s2", sfoo.getDiffs().getLastSnapshot().getRoot()
+    assertEquals(s2.getId(), sfoo.getDiffs().getLastSnapshotId());
-        .getLocalName());
   }
   
   /**
@@ -604,8 +608,10 @@ public class TestRenameWithSnapshots {
     
     INodeFile snode = fsdir.getINode(newfoo.toString()).asFile();
     assertEquals(1, snode.getDiffs().asList().size());
-    assertEquals("s2", snode.getDiffs().getLastSnapshot().getRoot()
+    INodeDirectorySnapshottable sdir2Node = 
-        .getLocalName());
+        (INodeDirectorySnapshottable) fsdir.getINode(sdir2.toString());
+    Snapshot s2 = sdir2Node.getSnapshot(DFSUtil.string2Bytes("s2"));
+    assertEquals(s2.getId(), snode.getDiffs().getLastSnapshotId());
     
     // restart cluster
     restartClusterAndCheckImage(true);
@@ -758,12 +764,14 @@ public class TestRenameWithSnapshots {
     assertEquals(2, fooWithCount.getReferenceCount());
     INodeDirectory foo = fooWithCount.asDirectory();
     assertEquals(1, foo.getDiffs().asList().size());
-    assertEquals("s1", foo.getDirectoryWithSnapshotFeature().getLastSnapshot()
+    INodeDirectorySnapshottable sdir1Node = 
-        .getRoot().getLocalName());
+        (INodeDirectorySnapshottable) fsdir.getINode(sdir1.toString());
+    Snapshot s1 = sdir1Node.getSnapshot(DFSUtil.string2Bytes("s1"));
+    assertEquals(s1.getId(), foo.getDirectoryWithSnapshotFeature()
+        .getLastSnapshotId());
     INodeFile bar1 = fsdir.getINode4Write(bar1_dir1.toString()).asFile();
     assertEquals(1, bar1.getDiffs().asList().size());
-    assertEquals("s1", bar1.getDiffs().getLastSnapshot().getRoot()
+    assertEquals(s1.getId(), bar1.getDiffs().getLastSnapshotId());
-        .getLocalName());
     
     INodeReference barRef = fsdir.getINode4Write(bar2_dir1.toString())
         .asReference();
@@ -772,8 +780,7 @@ public class TestRenameWithSnapshots {
     assertEquals(2, barWithCount.getReferenceCount());
     INodeFile bar = barWithCount.asFile();
     assertEquals(1, bar.getDiffs().asList().size());
-    assertEquals("s1", bar.getDiffs().getLastSnapshot().getRoot()
+    assertEquals(s1.getId(), bar.getDiffs().getLastSnapshotId());
-        .getLocalName());
     
     // restart the cluster and check fsimage
     restartClusterAndCheckImage(true);
@@ -967,6 +974,13 @@ public class TestRenameWithSnapshots {
     hdfs.rename(bar_dir2, bar_dir1);
     
     // check the internal details
+    INodeDirectorySnapshottable sdir1Node = 
+        (INodeDirectorySnapshottable) fsdir.getINode(sdir1.toString());
+    INodeDirectorySnapshottable sdir2Node = 
+        (INodeDirectorySnapshottable) fsdir.getINode(sdir2.toString());
+    INodeDirectorySnapshottable sdir3Node = 
+        (INodeDirectorySnapshottable) fsdir.getINode(sdir3.toString());
+    
     INodeReference fooRef = fsdir.getINode4Write(foo_dir1.toString())
         .asReference();
     INodeReference.WithCount fooWithCount = (WithCount) fooRef.getReferredINode();
@@ -975,16 +989,22 @@ public class TestRenameWithSnapshots {
     INodeDirectory foo = fooWithCount.asDirectory();
     List<DirectoryDiff> fooDiffs = foo.getDiffs().asList();
     assertEquals(4, fooDiffs.size());
-    assertEquals("s2222", fooDiffs.get(3).snapshot.getRoot().getLocalName());
+    
-    assertEquals("s333", fooDiffs.get(2).snapshot.getRoot().getLocalName());
+    Snapshot s2222 = sdir2Node.getSnapshot(DFSUtil.string2Bytes("s2222"));
-    assertEquals("s22", fooDiffs.get(1).snapshot.getRoot().getLocalName());
+    Snapshot s333 = sdir3Node.getSnapshot(DFSUtil.string2Bytes("s333"));
-    assertEquals("s1", fooDiffs.get(0).snapshot.getRoot().getLocalName());
+    Snapshot s22 = sdir2Node.getSnapshot(DFSUtil.string2Bytes("s22"));
+    Snapshot s1 = sdir1Node.getSnapshot(DFSUtil.string2Bytes("s1"));
+    
+    assertEquals(s2222.getId(), fooDiffs.get(3).getSnapshotId());
+    assertEquals(s333.getId(), fooDiffs.get(2).getSnapshotId());
+    assertEquals(s22.getId(), fooDiffs.get(1).getSnapshotId());
+    assertEquals(s1.getId(), fooDiffs.get(0).getSnapshotId());
     INodeFile bar1 = fsdir.getINode4Write(bar1_dir1.toString()).asFile();
     List<FileDiff> bar1Diffs = bar1.getDiffs().asList();
     assertEquals(3, bar1Diffs.size());
-    assertEquals("s333", bar1Diffs.get(2).snapshot.getRoot().getLocalName());
+    assertEquals(s333.getId(), bar1Diffs.get(2).getSnapshotId());
-    assertEquals("s22", bar1Diffs.get(1).snapshot.getRoot().getLocalName());
+    assertEquals(s22.getId(), bar1Diffs.get(1).getSnapshotId());
-    assertEquals("s1", bar1Diffs.get(0).snapshot.getRoot().getLocalName());
+    assertEquals(s1.getId(), bar1Diffs.get(0).getSnapshotId());
     
     INodeReference barRef = fsdir.getINode4Write(bar_dir1.toString())
         .asReference();
@@ -994,10 +1014,10 @@ public class TestRenameWithSnapshots {
     INodeFile bar = barWithCount.asFile();
     List<FileDiff> barDiffs = bar.getDiffs().asList();
     assertEquals(4, barDiffs.size());
-    assertEquals("s2222", barDiffs.get(3).snapshot.getRoot().getLocalName());
+    assertEquals(s2222.getId(), barDiffs.get(3).getSnapshotId());
-    assertEquals("s333", barDiffs.get(2).snapshot.getRoot().getLocalName());
+    assertEquals(s333.getId(), barDiffs.get(2).getSnapshotId());
-    assertEquals("s22", barDiffs.get(1).snapshot.getRoot().getLocalName());
+    assertEquals(s22.getId(), barDiffs.get(1).getSnapshotId());
-    assertEquals("s1", barDiffs.get(0).snapshot.getRoot().getLocalName());
+    assertEquals(s1.getId(), barDiffs.get(0).getSnapshotId());
     
     // restart the cluster and check fsimage
     restartClusterAndCheckImage(true);
@@ -1033,10 +1053,10 @@ public class TestRenameWithSnapshots {
     foo = fooWithCount.asDirectory();
     fooDiffs = foo.getDiffs().asList();
     assertEquals(4, fooDiffs.size());
-    assertEquals("s2222", fooDiffs.get(3).snapshot.getRoot().getLocalName());
+    assertEquals(s2222.getId(), fooDiffs.get(3).getSnapshotId());
     bar1Diffs = bar1.getDiffs().asList();
     assertEquals(3, bar1Diffs.size());
-    assertEquals("s333", bar1Diffs.get(2).snapshot.getRoot().getLocalName());
+    assertEquals(s333.getId(), bar1Diffs.get(2).getSnapshotId());
     
     barRef = fsdir.getINode(bar_s2222.toString()).asReference();
     barWithCount = (WithCount) barRef.getReferredINode();
@@ -1044,7 +1064,7 @@ public class TestRenameWithSnapshots {
     bar = barWithCount.asFile();
     barDiffs = bar.getDiffs().asList();
     assertEquals(4, barDiffs.size());
-    assertEquals("s2222", barDiffs.get(3).snapshot.getRoot().getLocalName());
+    assertEquals(s2222.getId(), barDiffs.get(3).getSnapshotId());
   }
   
   /**
@@ -1164,6 +1184,9 @@ public class TestRenameWithSnapshots {
     assertTrue(hdfs.exists(bar_s2));
     
     // check internal details
+    INodeDirectorySnapshottable sdir2Node = 
+        (INodeDirectorySnapshottable) fsdir.getINode(sdir2.toString());
+    Snapshot s2 = sdir2Node.getSnapshot(DFSUtil.string2Bytes("s2"));
     final Path foo_s2 = SnapshotTestHelper.getSnapshotPath(sdir2, "s2", "foo");
     INodeReference fooRef = fsdir.getINode(foo_s2.toString()).asReference();
     assertTrue(fooRef instanceof INodeReference.WithName);
@@ -1172,7 +1195,7 @@ public class TestRenameWithSnapshots {
     INodeDirectory fooDir = fooWC.getReferredINode().asDirectory();
     List<DirectoryDiff> diffs = fooDir.getDiffs().asList();
     assertEquals(1, diffs.size());
-    assertEquals("s2", diffs.get(0).snapshot.getRoot().getLocalName());
+    assertEquals(s2.getId(), diffs.get(0).getSnapshotId());
     
     // restart the cluster and check fsimage
     restartClusterAndCheckImage(true);
@@ -1260,7 +1283,7 @@ public class TestRenameWithSnapshots {
     INodeDirectory dir2 = fsdir.getINode4Write(sdir2.toString()).asDirectory();
     INodeDirectory mockDir2 = spy(dir2);
     doReturn(false).when(mockDir2).addChild((INode) anyObject(), anyBoolean(),
-            (Snapshot) anyObject());
+           Mockito.anyInt());
     INodeDirectory root = fsdir.getINode4Write("/").asDirectory();
     root.replaceChild(dir2, mockDir2, fsdir.getINodeMap());
     
@@ -1271,12 +1294,14 @@ public class TestRenameWithSnapshots {
     // check the current internal details
     INodeDirectorySnapshottable dir1Node = (INodeDirectorySnapshottable) fsdir
         .getINode4Write(sdir1.toString());
-    ReadOnlyList<INode> dir1Children = dir1Node.getChildrenList(null);
+    Snapshot s1 = dir1Node.getSnapshot(DFSUtil.string2Bytes("s1"));
+    ReadOnlyList<INode> dir1Children = dir1Node
+        .getChildrenList(Snapshot.CURRENT_STATE_ID);
     assertEquals(1, dir1Children.size());
     assertEquals(foo.getName(), dir1Children.get(0).getLocalName());
     List<DirectoryDiff> dir1Diffs = dir1Node.getDiffs().asList();
     assertEquals(1, dir1Diffs.size());
-    assertEquals("s1", dir1Diffs.get(0).snapshot.getRoot().getLocalName());
+    assertEquals(s1.getId(), dir1Diffs.get(0).getSnapshotId());
     
     // after the undo of rename, both the created and deleted list of sdir1
     // should be empty
@@ -1288,7 +1313,7 @@ public class TestRenameWithSnapshots {
     assertTrue(fooNode.isDirectory() && fooNode.asDirectory().isWithSnapshot());
     List<DirectoryDiff> fooDiffs = fooNode.asDirectory().getDiffs().asList();
     assertEquals(1, fooDiffs.size());
-    assertEquals("s1", fooDiffs.get(0).snapshot.getRoot().getLocalName());
+    assertEquals(s1.getId(), fooDiffs.get(0).getSnapshotId());
     
     final Path foo_s1 = SnapshotTestHelper.getSnapshotPath(sdir1, "s1", "foo");
     INode fooNode_s1 = fsdir.getINode(foo_s1.toString());
@@ -1299,7 +1324,8 @@ public class TestRenameWithSnapshots {
     INodeDirectory dir2Node = fsdir.getINode4Write(sdir2.toString())
         .asDirectory();
     assertFalse(dir2Node.isWithSnapshot());
-    ReadOnlyList<INode> dir2Children = dir2Node.getChildrenList(null);
+    ReadOnlyList<INode> dir2Children = dir2Node
+        .getChildrenList(Snapshot.CURRENT_STATE_ID);
     assertEquals(1, dir2Children.size());
     assertEquals(dir2file.getName(), dir2Children.get(0).getLocalName());
   }
@@ -1327,7 +1353,7 @@ public class TestRenameWithSnapshots {
     INodeDirectory dir2 = fsdir.getINode4Write(sdir2.toString()).asDirectory();
     INodeDirectory mockDir2 = spy(dir2);
     doReturn(false).when(mockDir2).addChild((INode) anyObject(), anyBoolean(),
-            (Snapshot) anyObject());
+            Mockito.anyInt());
     INodeDirectory root = fsdir.getINode4Write("/").asDirectory();
     root.replaceChild(dir2, mockDir2, fsdir.getINodeMap());
     
@@ -1338,12 +1364,14 @@ public class TestRenameWithSnapshots {
     // check the current internal details
     INodeDirectorySnapshottable dir1Node = (INodeDirectorySnapshottable) fsdir
         .getINode4Write(sdir1.toString());
-    ReadOnlyList<INode> dir1Children = dir1Node.getChildrenList(null);
+    Snapshot s1 = dir1Node.getSnapshot(DFSUtil.string2Bytes("s1"));
+    ReadOnlyList<INode> dir1Children = dir1Node
+        .getChildrenList(Snapshot.CURRENT_STATE_ID);
     assertEquals(1, dir1Children.size());
     assertEquals(foo.getName(), dir1Children.get(0).getLocalName());
     List<DirectoryDiff> dir1Diffs = dir1Node.getDiffs().asList();
     assertEquals(1, dir1Diffs.size());
-    assertEquals("s1", dir1Diffs.get(0).snapshot.getRoot().getLocalName());
+    assertEquals(s1.getId(), dir1Diffs.get(0).getSnapshotId());
     
     // after the undo of rename, the created list of sdir1 should contain 
     // 1 element
@@ -1363,7 +1391,8 @@ public class TestRenameWithSnapshots {
     INodeDirectory dir2Node = fsdir.getINode4Write(sdir2.toString())
         .asDirectory();
     assertFalse(dir2Node.isWithSnapshot());
-    ReadOnlyList<INode> dir2Children = dir2Node.getChildrenList(null);
+    ReadOnlyList<INode> dir2Children = dir2Node
+        .getChildrenList(Snapshot.CURRENT_STATE_ID);
     assertEquals(1, dir2Children.size());
     assertEquals(dir2file.getName(), dir2Children.get(0).getLocalName());
   }
@@ -1389,7 +1418,7 @@ public class TestRenameWithSnapshots {
     INodeDirectory dir3 = fsdir.getINode4Write(sdir3.toString()).asDirectory();
     INodeDirectory mockDir3 = spy(dir3);
     doReturn(false).when(mockDir3).addChild((INode) anyObject(), anyBoolean(),
-            (Snapshot) anyObject());
+            Mockito.anyInt());
     INodeDirectory root = fsdir.getINode4Write("/").asDirectory();
     root.replaceChild(dir3, mockDir3, fsdir.getINodeMap());
     
@@ -1400,13 +1429,18 @@ public class TestRenameWithSnapshots {
     assertFalse(result);
     
     // check the current internal details
+    INodeDirectorySnapshottable dir1Node = (INodeDirectorySnapshottable) fsdir
+        .getINode4Write(sdir1.toString());
+    Snapshot s1 = dir1Node.getSnapshot(DFSUtil.string2Bytes("s1"));
     INodeDirectorySnapshottable dir2Node = (INodeDirectorySnapshottable) fsdir
         .getINode4Write(sdir2.toString());
-    ReadOnlyList<INode> dir2Children = dir2Node.getChildrenList(null);
+    Snapshot s2 = dir2Node.getSnapshot(DFSUtil.string2Bytes("s2"));
+    ReadOnlyList<INode> dir2Children = dir2Node
+        .getChildrenList(Snapshot.CURRENT_STATE_ID);
     assertEquals(1, dir2Children.size());
     List<DirectoryDiff> dir2Diffs = dir2Node.getDiffs().asList();
     assertEquals(1, dir2Diffs.size());
-    assertEquals("s2", Snapshot.getSnapshotName(dir2Diffs.get(0).snapshot));
+    assertEquals(s2.getId(), dir2Diffs.get(0).getSnapshotId());
     ChildrenDiff childrenDiff = dir2Diffs.get(0).getChildrenDiff();
     assertEquals(0, childrenDiff.getList(ListType.DELETED).size());
     assertEquals(1, childrenDiff.getList(ListType.CREATED).size());
@@ -1418,7 +1452,7 @@ public class TestRenameWithSnapshots {
     assertTrue(fooNode instanceof INodeReference.DstReference);
     List<DirectoryDiff> fooDiffs = fooNode.asDirectory().getDiffs().asList();
     assertEquals(1, fooDiffs.size());
-    assertEquals("s1", fooDiffs.get(0).snapshot.getRoot().getLocalName());
+    assertEquals(s1.getId(), fooDiffs.get(0).getSnapshotId());
     
     // create snapshot on sdir2 and rename again
     hdfs.createSnapshot(sdir2, "s3");
@@ -1428,13 +1462,14 @@ public class TestRenameWithSnapshots {
     // check internal details again
     dir2Node = (INodeDirectorySnapshottable) fsdir.getINode4Write(sdir2
         .toString());
+    Snapshot s3 = dir2Node.getSnapshot(DFSUtil.string2Bytes("s3"));
     fooNode = fsdir.getINode4Write(foo_dir2.toString());
-    dir2Children = dir2Node.getChildrenList(null);
+    dir2Children = dir2Node.getChildrenList(Snapshot.CURRENT_STATE_ID);
     assertEquals(1, dir2Children.size());
     dir2Diffs = dir2Node.getDiffs().asList();
     assertEquals(2, dir2Diffs.size());
-    assertEquals("s2", Snapshot.getSnapshotName(dir2Diffs.get(0).snapshot));
+    assertEquals(s2.getId(), dir2Diffs.get(0).getSnapshotId());
-    assertEquals("s3", Snapshot.getSnapshotName(dir2Diffs.get(1).snapshot));
+    assertEquals(s3.getId(), dir2Diffs.get(1).getSnapshotId());
     
     childrenDiff = dir2Diffs.get(0).getChildrenDiff();
     assertEquals(0, childrenDiff.getList(ListType.DELETED).size());
@@ -1452,8 +1487,8 @@ public class TestRenameWithSnapshots {
     assertTrue(fooNode instanceof INodeReference.DstReference);
     fooDiffs = fooNode.asDirectory().getDiffs().asList();
     assertEquals(2, fooDiffs.size());
-    assertEquals("s1", fooDiffs.get(0).snapshot.getRoot().getLocalName());
+    assertEquals(s1.getId(), fooDiffs.get(0).getSnapshotId());
-    assertEquals("s3", fooDiffs.get(1).snapshot.getRoot().getLocalName());
+    assertEquals(s3.getId(), fooDiffs.get(1).getSnapshotId());
   }
   
   /**
@@ -1489,9 +1524,9 @@ public class TestRenameWithSnapshots {
     INodeDirectory mockDir3 = spy(dir3);
     // fail the rename but succeed in undo
     doReturn(false).when(mockDir3).addChild((INode) Mockito.isNull(),
-        anyBoolean(), (Snapshot) anyObject());
+        anyBoolean(), Mockito.anyInt());
     Mockito.when(mockDir3.addChild((INode) Mockito.isNotNull(), anyBoolean(), 
-        (Snapshot) anyObject())).thenReturn(false).thenCallRealMethod();
+        Mockito.anyInt())).thenReturn(false).thenCallRealMethod();
     INodeDirectory root = fsdir.getINode4Write("/").asDirectory();
     root.replaceChild(dir3, mockDir3, fsdir.getINodeMap());
     foo3Node.setParent(mockDir3);
@@ -1551,7 +1586,7 @@ public class TestRenameWithSnapshots {
     INodeDirectory dir1Node = fsdir.getINode4Write(dir1.toString())
         .asDirectory();
     List<INode> childrenList = ReadOnlyList.Util.asList(dir1Node
-        .getChildrenList(null));
+        .getChildrenList(Snapshot.CURRENT_STATE_ID));
     assertEquals(1, childrenList.size());
     INode fooNode = childrenList.get(0);
     assertTrue(fooNode.asDirectory().isWithSnapshot());
@@ -1572,7 +1607,7 @@ public class TestRenameWithSnapshots {
     assertEquals(3, counts.get(Quota.NAMESPACE));
     assertEquals(0, counts.get(Quota.DISKSPACE));
     childrenList = ReadOnlyList.Util.asList(dir2Node.asDirectory()
-        .getChildrenList(null));
+        .getChildrenList(Snapshot.CURRENT_STATE_ID));
     assertEquals(1, childrenList.size());
     INode subdir2Node = childrenList.get(0);
     assertSame(dir2Node, subdir2Node.getParent());
@@ -1627,7 +1662,7 @@ public class TestRenameWithSnapshots {
     INodeDirectory dir1Node = fsdir.getINode4Write(dir1.toString())
         .asDirectory();
     List<INode> childrenList = ReadOnlyList.Util.asList(dir1Node
-        .getChildrenList(null));
+        .getChildrenList(Snapshot.CURRENT_STATE_ID));
     assertEquals(1, childrenList.size());
     INode fooNode = childrenList.get(0);
     assertTrue(fooNode.asDirectory().isWithSnapshot());
@@ -1646,7 +1681,7 @@ public class TestRenameWithSnapshots {
     assertEquals(4, counts.get(Quota.NAMESPACE));
     assertEquals(0, counts.get(Quota.DISKSPACE));
     childrenList = ReadOnlyList.Util.asList(dir2Node.asDirectory()
-        .getChildrenList(null));
+        .getChildrenList(Snapshot.CURRENT_STATE_ID));
     assertEquals(1, childrenList.size());
     INode subdir2Node = childrenList.get(0);
     assertTrue(subdir2Node.asDirectory().isWithSnapshot());
@@ -1690,14 +1725,18 @@ public class TestRenameWithSnapshots {
     }
     
     // check
+    INodeDirectorySnapshottable rootNode = (INodeDirectorySnapshottable) fsdir
+        .getINode4Write(root.toString());
     INodeDirectory fooNode = fsdir.getINode4Write(foo.toString()).asDirectory();
-    ReadOnlyList<INode> children = fooNode.getChildrenList(null);
+    ReadOnlyList<INode> children = fooNode
+        .getChildrenList(Snapshot.CURRENT_STATE_ID);
     assertEquals(1, children.size());
     List<DirectoryDiff> diffList = fooNode.getDiffs().asList();
     assertEquals(1, diffList.size());
     DirectoryDiff diff = diffList.get(0);
     // this diff is generated while renaming
-    assertEquals(snap1, Snapshot.getSnapshotName(diff.snapshot));
+    Snapshot s1 = rootNode.getSnapshot(DFSUtil.string2Bytes(snap1));
+    assertEquals(s1.getId(), diff.getSnapshotId());
     // after undo, the diff should be empty
     assertTrue(diff.getChildrenDiff().getList(ListType.DELETED).isEmpty());
     assertTrue(diff.getChildrenDiff().getList(ListType.CREATED).isEmpty());
@@ -1709,7 +1748,7 @@ public class TestRenameWithSnapshots {
     List<FileDiff> barDiffList = barNode.getDiffs().asList();
     assertEquals(1, barDiffList.size());
     FileDiff barDiff = barDiffList.get(0);
-    assertEquals(snap1, Snapshot.getSnapshotName(barDiff.snapshot));
+    assertEquals(s1.getId(), barDiff.getSnapshotId());
     
     // restart cluster multiple times to make sure the fsimage and edits log are
     // correct. Note that when loading fsimage, foo and bar will be converted 
@@ -1941,12 +1980,14 @@ public class TestRenameWithSnapshots {
         (WithCount) fooRef.asReference().getReferredINode();
     assertEquals(1, wc.getReferenceCount());
     INodeDirectory fooNode = wc.getReferredINode().asDirectory();
-    ReadOnlyList<INode> children = fooNode.getChildrenList(null);
+    ReadOnlyList<INode> children = fooNode
+        .getChildrenList(Snapshot.CURRENT_STATE_ID);
     assertEquals(1, children.size());
     assertEquals(bar.getName(), children.get(0).getLocalName());
     List<DirectoryDiff> diffList = fooNode.getDiffs().asList();
     assertEquals(1, diffList.size());
-    assertEquals("s1", Snapshot.getSnapshotName(diffList.get(0).snapshot));
+    Snapshot s1 = dir1Node.getSnapshot(DFSUtil.string2Bytes("s1"));
+    assertEquals(s1.getId(), diffList.get(0).getSnapshotId());
     ChildrenDiff diff = diffList.get(0).getChildrenDiff();
     assertEquals(0, diff.getList(ListType.CREATED).size());
     assertEquals(0, diff.getList(ListType.DELETED).size());
@@ -2009,14 +2050,16 @@ public class TestRenameWithSnapshots {
         (WithCount) fooRef.asReference().getReferredINode();
     assertEquals(2, wc.getReferenceCount());
     INodeDirectory fooNode = wc.getReferredINode().asDirectory();
-    ReadOnlyList<INode> children = fooNode.getChildrenList(null);
+    ReadOnlyList<INode> children = fooNode
+        .getChildrenList(Snapshot.CURRENT_STATE_ID);
     assertEquals(3, children.size());
     assertEquals(bar.getName(), children.get(0).getLocalName());
     assertEquals(bar2.getName(), children.get(1).getLocalName());
     assertEquals(bar3.getName(), children.get(2).getLocalName());
     List<DirectoryDiff> diffList = fooNode.getDiffs().asList();
     assertEquals(1, diffList.size());
-    assertEquals("s1", Snapshot.getSnapshotName(diffList.get(0).snapshot));
+    Snapshot s1 = dir1Node.getSnapshot(DFSUtil.string2Bytes("s1"));
+    assertEquals(s1.getId(), diffList.get(0).getSnapshotId());
     ChildrenDiff diff = diffList.get(0).getChildrenDiff();
     // bar2 and bar3 in the created list
     assertEquals(2, diff.getList(ListType.CREATED).size());
@@ -2134,11 +2177,12 @@ public class TestRenameWithSnapshots {
     // recordModification before the rename
     assertTrue(fooNode.isWithSnapshot());
     assertTrue(fooNode.getDiffs().asList().isEmpty());
-    INodeDirectory barNode = fooNode.getChildrenList(null).get(0).asDirectory();
+    INodeDirectory barNode = fooNode.getChildrenList(Snapshot.CURRENT_STATE_ID)
+        .get(0).asDirectory();
     // bar should also be INodeDirectory (With Snapshot), and both of its diff 
     // list and children list are empty 
     assertTrue(barNode.getDiffs().asList().isEmpty());
-    assertTrue(barNode.getChildrenList(null).isEmpty());
+    assertTrue(barNode.getChildrenList(Snapshot.CURRENT_STATE_ID).isEmpty());
     
     restartClusterAndCheckImage(true);
   }
@@ -2210,7 +2254,10 @@ public class TestRenameWithSnapshots {
     List<DirectoryDiff> barDiffList = barNode.getDiffs().asList();
     assertEquals(1, barDiffList.size());
     DirectoryDiff diff = barDiffList.get(0);
-    assertEquals("s0", Snapshot.getSnapshotName(diff.snapshot));
+    INodeDirectorySnapshottable testNode = 
+        (INodeDirectorySnapshottable) fsdir.getINode4Write(test.toString());
+    Snapshot s0 = testNode.getSnapshot(DFSUtil.string2Bytes("s0"));
+    assertEquals(s0.getId(), diff.getSnapshotId());
     // and file should be stored in the deleted list of this snapshot diff
     assertEquals("file", diff.getChildrenDiff().getList(ListType.DELETED)
         .get(0).getLocalName());
@@ -2276,7 +2323,7 @@ public class TestRenameWithSnapshots {
     final Path barInS0 = SnapshotTestHelper.getSnapshotPath(test, "s0",
         "foo/bar");
     INodeDirectory barNode = fsdir.getINode(barInS0.toString()).asDirectory();
-    assertEquals(0, barNode.getChildrenList(null).size());
+    assertEquals(0, barNode.getChildrenList(Snapshot.CURRENT_STATE_ID).size());
     List<DirectoryDiff> diffList = barNode.getDiffs().asList();
     assertEquals(1, diffList.size());
     DirectoryDiff diff = diffList.get(0);

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

@@ -28,6 +28,7 @@ import org.apache.hadoop.conf.Configuration;
 import org.apache.hadoop.fs.Path;
 import org.apache.hadoop.hdfs.DFSConfigKeys;
 import org.apache.hadoop.hdfs.DFSTestUtil;
+import org.apache.hadoop.hdfs.DFSUtil;
 import org.apache.hadoop.hdfs.DistributedFileSystem;
 import org.apache.hadoop.hdfs.MiniDFSCluster;
 import org.apache.hadoop.hdfs.protocol.HdfsConstants;
@@ -153,7 +154,9 @@ public class TestSetQuotaWithSnapshot {
     assertTrue(subNode.asDirectory().isWithSnapshot());
     List<DirectoryDiff> diffList = subNode.asDirectory().getDiffs().asList();
     assertEquals(1, diffList.size());
-    assertEquals("s2", Snapshot.getSnapshotName(diffList.get(0).snapshot));
+    Snapshot s2 = ((INodeDirectorySnapshottable) dirNode).getSnapshot(DFSUtil
+        .string2Bytes("s2"));
+    assertEquals(s2.getId(), diffList.get(0).getSnapshotId());
     List<INode> createdList = diffList.get(0).getChildrenDiff().getList(ListType.CREATED);
     assertEquals(1, createdList.size());
     assertSame(fsdir.getINode4Write(file.toString()), createdList.get(0));

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

@@ -268,7 +268,8 @@ public class TestSnapshotDeletion {
         (INodeDirectory) fsdir.getINode(snapshotNoChangeDir.toString());
     // should still be an INodeDirectory
     assertEquals(INodeDirectory.class, snapshotNode.getClass());
-    ReadOnlyList<INode> children = snapshotNode.getChildrenList(null);
+    ReadOnlyList<INode> children = snapshotNode
+        .getChildrenList(Snapshot.CURRENT_STATE_ID);
     // check 2 children: noChangeFile and metaChangeFile2
     assertEquals(2, children.size());
     INode noChangeFileSCopy = children.get(1);
@@ -286,11 +287,11 @@ public class TestSnapshotDeletion {
     
     // check the replication factor of metaChangeFile2SCopy
     assertEquals(REPLICATION_1,
-        metaChangeFile2SCopy.getFileReplication(null));
+        metaChangeFile2SCopy.getFileReplication(Snapshot.CURRENT_STATE_ID));
     assertEquals(REPLICATION_1,
-        metaChangeFile2SCopy.getFileReplication(snapshot1));
+        metaChangeFile2SCopy.getFileReplication(snapshot1.getId()));
     assertEquals(REPLICATION,
-        metaChangeFile2SCopy.getFileReplication(snapshot0));
+        metaChangeFile2SCopy.getFileReplication(snapshot0.getId()));
     
     // Case 4: delete directory sub
     // before deleting sub, we first create a new file under sub
@@ -316,23 +317,25 @@ public class TestSnapshotDeletion {
     assertTrue(snapshotNode4Sub.isWithSnapshot());
     // the snapshot copy of sub has only one child subsub.
     // newFile should have been destroyed
-    assertEquals(1, snapshotNode4Sub.getChildrenList(null).size());
+    assertEquals(1, snapshotNode4Sub.getChildrenList(Snapshot.CURRENT_STATE_ID)
+        .size());
     // but should have two children, subsub and noChangeDir, when s1 was taken  
-    assertEquals(2, snapshotNode4Sub.getChildrenList(snapshot1).size());
+    assertEquals(2, snapshotNode4Sub.getChildrenList(snapshot1.getId()).size());
     
     // check the snapshot copy of subsub, which is contained in the subtree of
     // sub's snapshot copy
-    INode snapshotNode4Subsub = snapshotNode4Sub.getChildrenList(null).get(0);
+    INode snapshotNode4Subsub = snapshotNode4Sub.getChildrenList(
+        Snapshot.CURRENT_STATE_ID).get(0);
     assertTrue(snapshotNode4Subsub.asDirectory().isWithSnapshot());
     assertTrue(snapshotNode4Sub == snapshotNode4Subsub.getParent());
     // check the children of subsub
     INodeDirectory snapshotSubsubDir = (INodeDirectory) snapshotNode4Subsub;
-    children = snapshotSubsubDir.getChildrenList(null);
+    children = snapshotSubsubDir.getChildrenList(Snapshot.CURRENT_STATE_ID);
     assertEquals(2, children.size());
     assertEquals(children.get(0).getLocalName(), metaChangeFile1.getName());
     assertEquals(children.get(1).getLocalName(), newFileAfterS0.getName());
     // only one child before snapshot s0 
-    children = snapshotSubsubDir.getChildrenList(snapshot0);
+    children = snapshotSubsubDir.getChildrenList(snapshot0.getId());
     assertEquals(1, children.size());
     INode child = children.get(0);
     assertEquals(child.getLocalName(), metaChangeFile1.getName());
@@ -341,11 +344,11 @@ public class TestSnapshotDeletion {
     assertTrue(metaChangeFile1SCopy.isWithSnapshot());
     assertFalse(metaChangeFile1SCopy.isUnderConstruction());
     assertEquals(REPLICATION_1,
-        metaChangeFile1SCopy.getFileReplication(null));
+        metaChangeFile1SCopy.getFileReplication(Snapshot.CURRENT_STATE_ID));
     assertEquals(REPLICATION_1,
-        metaChangeFile1SCopy.getFileReplication(snapshot1));
+        metaChangeFile1SCopy.getFileReplication(snapshot1.getId()));
     assertEquals(REPLICATION,
-        metaChangeFile1SCopy.getFileReplication(snapshot0));
+        metaChangeFile1SCopy.getFileReplication(snapshot0.getId()));
   }
   
   /**
@@ -474,9 +477,10 @@ public class TestSnapshotDeletion {
         (INodeDirectorySnapshottable) fsdir.getINode(dir.toString());
     Snapshot snapshot0 = dirNode.getSnapshot(DFSUtil.string2Bytes("s0"));
     assertNull(snapshot0);
+    Snapshot snapshot1 = dirNode.getSnapshot(DFSUtil.string2Bytes("s1"));
     DirectoryDiffList diffList = dirNode.getDiffs();
     assertEquals(1, diffList.asList().size());
-    assertEquals("s1", diffList.getLast().snapshot.getRoot().getLocalName());
+    assertEquals(snapshot1.getId(), diffList.getLast().getSnapshotId());
     diffList = fsdir.getINode(metaChangeDir.toString()).asDirectory()
         .getDiffs();
     assertEquals(0, diffList.asList().size());

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

@@ -101,7 +101,8 @@ public class TestSnapshotRename {
     List<DirectoryDiff> listByTime = srcRoot.getDiffs().asList();
     assertEquals(names.length, listByTime.size());
     for (int i = 0; i < listByTime.size(); i++) {
-      assertEquals(names[i], listByTime.get(i).getSnapshot().getRoot().getLocalName());
+      Snapshot s = srcRoot.getSnapshotById(listByTime.get(i).getSnapshotId());
+      assertEquals(names[i], s.getRoot().getLocalName());
     }
   }
   

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

@@ -150,7 +150,7 @@ public class TestSnapshotReplication {
       assertEquals(expectedBlockRep, ssInode.getBlockReplication());
       // Also check the number derived from INodeFile#getFileReplication
       assertEquals(snapshotRepMap.get(ss).shortValue(),
-          ssInode.getFileReplication(iip.getPathSnapshot()));
+          ssInode.getFileReplication(iip.getPathSnapshotId()));
     }
   }
   

hadoop-yarn-project/CHANGES.txt

@@ -55,6 +55,9 @@ Release 2.4.0 - UNRELEASED
     YARN-1028. Added FailoverProxyProvider capability to ResourceManager to help
     with RM failover. (Karthik Kambatla via vinodkv)
 
+    YARN-1029. Added embedded leader election in the ResourceManager. (Karthik
+    Kambatla via vinodkv)
+
   IMPROVEMENTS
 
     YARN-7. Support CPU resource for DistributedShell. (Junping Du via llu)
@@ -197,6 +200,11 @@ Release 2.4.0 - UNRELEASED
     YARN-1493. Changed ResourceManager and Scheduler interfacing to recognize
     app-attempts separately from apps. (Jian He via vinodkv)
 
+    YARN-1482. Modified WebApplicationProxy to make it work across ResourceManager
+    fail-over. (Xuan Gong via vinodkv)
+
+    YARN-1568. Rename clusterid to clusterId in ActiveRMInfoProto (kasha)
+
   OPTIMIZATIONS
 
   BUG FIXES
@@ -292,6 +300,12 @@ Release 2.4.0 - UNRELEASED
     YARN-1560. Fixed TestYarnClient#testAMMRTokens failure with null AMRM token.
     (Ted Yu via jianhe)
 
+    YARN-1409. NonAggregatingLogHandler can throw RejectedExecutionException
+    (Tsuyoshi OZAWA via jlowe)
+
+    YARN-1293. Fixed TestContainerLaunch#testInvalidEnvSyntaxDiagnostics failure
+    caused by non-English system locale. (Tsuyoshi OZAWA via jianhe)
+
 Release 2.3.0 - UNRELEASED
 
   INCOMPATIBLE CHANGES

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api/src/main/java/org/apache/hadoop/yarn/conf/HAUtil.java

@@ -51,6 +51,22 @@ public class HAUtil {
         YarnConfiguration.DEFAULT_RM_HA_ENABLED);
   }
 
+  public static boolean isAutomaticFailoverEnabled(Configuration conf) {
+    return conf.getBoolean(YarnConfiguration.AUTO_FAILOVER_ENABLED,
+        YarnConfiguration.DEFAULT_AUTO_FAILOVER_ENABLED);
+  }
+
+  public static boolean isAutomaticFailoverEnabledAndEmbedded(
+      Configuration conf) {
+    return isAutomaticFailoverEnabled(conf) &&
+        isAutomaticFailoverEmbedded(conf);
+  }
+
+  public static boolean isAutomaticFailoverEmbedded(Configuration conf) {
+    return conf.getBoolean(YarnConfiguration.AUTO_FAILOVER_EMBEDDED,
+        YarnConfiguration.DEFAULT_AUTO_FAILOVER_EMBEDDED);
+  }
+
   /**
    * Verify configuration for Resource Manager HA.
    * @param conf Configuration
@@ -162,8 +178,7 @@ public class HAUtil {
    * @param conf Configuration. Please use verifyAndSetRMHAId to check.
    * @return RM Id on success
    */
-  @VisibleForTesting
+  public static String getRMHAId(Configuration conf) {
-  static String getRMHAId(Configuration conf) {
     return conf.get(YarnConfiguration.RM_HA_ID);
   }
 

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api/src/main/java/org/apache/hadoop/yarn/conf/YarnConfiguration.java

@@ -59,7 +59,7 @@ public class YarnConfiguration extends Configuration {
   public static final String IPC_PREFIX = YARN_PREFIX + "ipc.";
 
   /** Factory to create client IPC classes.*/
-  public static final String IPC_CLIENT_FACTORY_CLASS = 
+  public static final String IPC_CLIENT_FACTORY_CLASS =
     IPC_PREFIX + "client.factory.class";
   public static final String DEFAULT_IPC_CLIENT_FACTORY_CLASS = 
       "org.apache.hadoop.yarn.factories.impl.pb.RpcClientFactoryPBImpl";
@@ -87,6 +87,8 @@ public class YarnConfiguration extends Configuration {
   ////////////////////////////////
   public static final String RM_PREFIX = "yarn.resourcemanager.";
 
+  public static final String RM_CLUSTER_ID = RM_PREFIX + "cluster-id";
+
   /** The address of the applications manager interface in the RM.*/
   public static final String RM_ADDRESS = 
     RM_PREFIX + "address";
@@ -278,6 +280,36 @@ public class YarnConfiguration extends Configuration {
   public static final String RECOVERY_ENABLED = RM_PREFIX + "recovery.enabled";
   public static final boolean DEFAULT_RM_RECOVERY_ENABLED = false;
 
+  /** Zookeeper interaction configs */
+  public static final String RM_ZK_PREFIX = RM_PREFIX + "zk-";
+
+  public static final String RM_ZK_ADDRESS = RM_ZK_PREFIX + "address";
+
+  public static final String RM_ZK_NUM_RETRIES = RM_ZK_PREFIX + "num-retries";
+  public static final int DEFAULT_ZK_RM_NUM_RETRIES = 500;
+
+  public static final String RM_ZK_RETRY_INTERVAL_MS =
+      RM_ZK_PREFIX + "retry-interval-ms";
+  public static final long DEFAULT_RM_ZK_RETRY_INTERVAL_MS = 2000;
+
+  public static final String RM_ZK_TIMEOUT_MS = RM_ZK_PREFIX + "timeout-ms";
+  public static final int DEFAULT_RM_ZK_TIMEOUT_MS = 10000;
+
+  public static final String RM_ZK_ACL = RM_ZK_PREFIX + "acl";
+  public static final String DEFAULT_RM_ZK_ACL = "world:anyone:rwcda";
+
+  public static final String ZK_STATE_STORE_PREFIX =
+      RM_PREFIX + "zk-state-store.";
+
+  /** Parent znode path under which ZKRMStateStore will create znodes */
+  public static final String ZK_RM_STATE_STORE_PARENT_PATH =
+      ZK_STATE_STORE_PREFIX + "parent-path";
+  public static final String DEFAULT_ZK_RM_STATE_STORE_PARENT_PATH = "/rmstore";
+
+  /** Root node ACLs for fencing */
+  public static final String ZK_RM_STATE_STORE_ROOT_NODE_ACL =
+      ZK_STATE_STORE_PREFIX + "root-node.acl";
+
   /** HA related configs */
   public static final String RM_HA_PREFIX = RM_PREFIX + "ha.";
   public static final String RM_HA_ENABLED = RM_HA_PREFIX + "enabled";
@@ -296,6 +328,22 @@ public class YarnConfiguration extends Configuration {
           HttpConfig.isSecure() ? RM_WEBAPP_HTTPS_ADDRESS
               : RM_WEBAPP_ADDRESS));
 
+  public static final String AUTO_FAILOVER_PREFIX =
+      RM_HA_PREFIX + "automatic-failover.";
+
+  public static final String AUTO_FAILOVER_ENABLED =
+      AUTO_FAILOVER_PREFIX + "enabled";
+  public static final boolean DEFAULT_AUTO_FAILOVER_ENABLED = false;
+
+  public static final String AUTO_FAILOVER_EMBEDDED =
+      AUTO_FAILOVER_PREFIX + "embedded";
+  public static final boolean DEFAULT_AUTO_FAILOVER_EMBEDDED = false;
+
+  public static final String AUTO_FAILOVER_ZK_BASE_PATH =
+      AUTO_FAILOVER_PREFIX + "zk-base-path";
+  public static final String DEFAULT_AUTO_FAILOVER_ZK_BASE_PATH =
+      "/yarn-leader-election";
+
   public static final String CLIENT_FAILOVER_PREFIX =
       YARN_PREFIX + "client.failover-";
   public static final String CLIENT_FAILOVER_PROXY_PROVIDER =
@@ -334,36 +382,6 @@ public class YarnConfiguration extends Configuration {
       + "fs.state-store.retry-policy-spec";
   public static final String DEFAULT_FS_RM_STATE_STORE_RETRY_POLICY_SPEC =
       "2000, 500";
-  /**
-   * Comma separated host:port pairs, each corresponding to a ZK server for
-   * ZKRMStateStore
-   */
-  public static final String ZK_STATE_STORE_PREFIX =
-      RM_PREFIX + "zk-state-store.";
-  public static final String ZK_RM_STATE_STORE_NUM_RETRIES =
-      ZK_STATE_STORE_PREFIX + "num-retries";
-  public static final int DEFAULT_ZK_RM_STATE_STORE_NUM_RETRIES = 500;
-  /** retry interval when connecting to zookeeper*/
-  public static final String ZK_RM_STATE_STORE_RETRY_INTERVAL_MS =
-      ZK_STATE_STORE_PREFIX + "retry-interval-ms";
-  public static final long DEFAULT_ZK_RM_STATE_STORE_RETRY_INTERVAL_MS = 2000;
-  public static final String ZK_RM_STATE_STORE_ADDRESS =
-      ZK_STATE_STORE_PREFIX + "address";
-  /** Timeout in millisec for ZK server connection for ZKRMStateStore */
-  public static final String ZK_RM_STATE_STORE_TIMEOUT_MS =
-      ZK_STATE_STORE_PREFIX + "timeout-ms";
-  public static final int DEFAULT_ZK_RM_STATE_STORE_TIMEOUT_MS = 60000;
-  /** Parent znode path under which ZKRMStateStore will create znodes */
-  public static final String ZK_RM_STATE_STORE_PARENT_PATH =
-      ZK_STATE_STORE_PREFIX + "parent-path";
-  public static final String DEFAULT_ZK_RM_STATE_STORE_PARENT_PATH = "/rmstore";
-  /** ACL for znodes in ZKRMStateStore */
-  public static final String ZK_RM_STATE_STORE_ACL =
-      ZK_STATE_STORE_PREFIX + "acl";
-  public static final String DEFAULT_ZK_RM_STATE_STORE_ACL =
-      "world:anyone:rwcda";
-  public static final String ZK_RM_STATE_STORE_ROOT_NODE_ACL =
-      ZK_STATE_STORE_PREFIX + "root-node.acl";
 
   /** The maximum number of completed applications RM keeps. */ 
   public static final String RM_MAX_COMPLETED_APPLICATIONS =

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api/src/main/proto/server/yarn_server_resourcemanager_service_protos.proto

@@ -133,3 +133,11 @@ message RMStateVersionProto {
   optional int32 major_version = 1;
   optional int32 minor_version = 2;
 }
+
+//////////////////////////////////////////////////////////////////
+///////////// RM Failover related records ////////////////////////
+//////////////////////////////////////////////////////////////////
+message ActiveRMInfoProto {
+  required string clusterId = 1;
+  required string rmId = 2;
+}

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-client/pom.xml

@@ -30,7 +30,13 @@
   </properties>
 
   <dependencies>
-  	<dependency>
+    <dependency>
+      <groupId>org.apache.zookeeper</groupId>
+      <artifactId>zookeeper</artifactId>
+      <type>test-jar</type>
+      <scope>test</scope>
+    </dependency>
+    <dependency>
   		<groupId>org.apache.hadoop</groupId>
   		<artifactId>hadoop-yarn-api</artifactId>
   	</dependency>

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-client/src/test/java/org/apache/hadoop/yarn/client/TestRMFailover.java

@@ -24,41 +24,52 @@ import static org.junit.Assert.assertTrue;
 import static org.junit.Assert.fail;
 
 import java.io.IOException;
+import java.net.HttpURLConnection;
+import java.net.URL;
 
 import org.apache.commons.logging.Log;
 import org.apache.commons.logging.LogFactory;
 import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.ha.ClientBaseWithFixes;
 import org.apache.hadoop.ha.HAServiceProtocol;
+import org.apache.hadoop.ha.proto.HAServiceProtocolProtos;
+import org.apache.hadoop.service.Service.STATE;
+import org.apache.hadoop.yarn.api.records.ApplicationId;
 import org.apache.hadoop.yarn.client.api.YarnClient;
 import org.apache.hadoop.yarn.conf.HAUtil;
 import org.apache.hadoop.yarn.conf.YarnConfiguration;
+import org.apache.hadoop.yarn.exceptions.ApplicationNotFoundException;
 import org.apache.hadoop.yarn.exceptions.YarnException;
 import org.apache.hadoop.yarn.server.MiniYARNCluster;
 import org.apache.hadoop.yarn.server.resourcemanager.AdminService;
-import org.junit.AfterClass;
+import org.apache.hadoop.yarn.server.webproxy.WebAppProxyServer;
-import org.junit.BeforeClass;
+import org.junit.After;
+import org.junit.Assert;
+import org.junit.Before;
 import org.junit.Test;
 
-public class TestRMFailover {
+public class TestRMFailover extends ClientBaseWithFixes {
   private static final Log LOG =
       LogFactory.getLog(TestRMFailover.class.getName());
+  private static final HAServiceProtocol.StateChangeRequestInfo req =
+      new HAServiceProtocol.StateChangeRequestInfo(
+          HAServiceProtocol.RequestSource.REQUEST_BY_USER);
 
   private static final String RM1_NODE_ID = "rm1";
   private static final int RM1_PORT_BASE = 10000;
   private static final String RM2_NODE_ID = "rm2";
   private static final int RM2_PORT_BASE = 20000;
-  private static final HAServiceProtocol.StateChangeRequestInfo req =
-      new HAServiceProtocol.StateChangeRequestInfo(
-          HAServiceProtocol.RequestSource.REQUEST_BY_USER_FORCED);
 
-  private static Configuration conf;
+  private Configuration conf;
-  private static MiniYARNCluster cluster;
+  private MiniYARNCluster cluster;
+  private ApplicationId fakeAppId;
 
-  private static void setConfForRM(String rmId, String prefix, String value) {
+
+  private void setConfForRM(String rmId, String prefix, String value) {
     conf.set(HAUtil.addSuffix(prefix, rmId), value);
   }
 
-  private static void setRpcAddressForRM(String rmId, int base) {
+  private void setRpcAddressForRM(String rmId, int base) {
     setConfForRM(rmId, YarnConfiguration.RM_ADDRESS, "0.0.0.0:" +
         (base + YarnConfiguration.DEFAULT_RM_PORT));
     setConfForRM(rmId, YarnConfiguration.RM_SCHEDULER_ADDRESS, "0.0.0.0:" +
@@ -73,13 +84,9 @@ public class TestRMFailover {
         (base + YarnConfiguration.DEFAULT_RM_WEBAPP_HTTPS_PORT));
   }
 
-  private static AdminService getRMAdminService(int index) {
+  @Before
-    return
+  public void setup() throws IOException {
-        cluster.getResourceManager(index).getRMContext().getRMAdminService();
+    fakeAppId = ApplicationId.newInstance(System.currentTimeMillis(), 0);
-  }
-
-  @BeforeClass
-  public static void setup() throws IOException {
     conf = new YarnConfiguration();
     conf.setBoolean(YarnConfiguration.RM_HA_ENABLED, true);
     conf.set(YarnConfiguration.RM_HA_IDS, RM1_NODE_ID + "," + RM2_NODE_ID);
@@ -87,27 +94,22 @@ public class TestRMFailover {
     setRpcAddressForRM(RM2_NODE_ID, RM2_PORT_BASE);
 
     conf.setLong(YarnConfiguration.CLIENT_FAILOVER_SLEEPTIME_BASE_MS, 100L);
+
     conf.setBoolean(YarnConfiguration.YARN_MINICLUSTER_FIXED_PORTS, true);
     conf.setBoolean(YarnConfiguration.YARN_MINICLUSTER_USE_RPC, true);
 
     cluster = new MiniYARNCluster(TestRMFailover.class.getName(), 2, 1, 1, 1);
-    cluster.init(conf);
-    cluster.start();
-
-    cluster.getResourceManager(0).getRMContext().getRMAdminService()
-        .transitionToActive(req);
-    assertFalse("RM never turned active", -1 == cluster.getActiveRMIndex());
   }
 
-  @AfterClass
+  @After
-  public static void teardown() {
+  public void teardown() {
     cluster.stop();
   }
 
   private void verifyClientConnection() {
     int numRetries = 3;
     while(numRetries-- > 0) {
-      Configuration conf = new YarnConfiguration(TestRMFailover.conf);
+      Configuration conf = new YarnConfiguration(this.conf);
       YarnClient client = YarnClient.createYarnClient();
       client.init(conf);
       client.start();
@@ -123,31 +125,131 @@ public class TestRMFailover {
     fail("Client couldn't connect to the Active RM");
   }
 
+  private void verifyConnections() throws InterruptedException, YarnException {
+    assertTrue("NMs failed to connect to the RM",
+        cluster.waitForNodeManagersToConnect(20000));
+    verifyClientConnection();
+  }
+
+  private AdminService getAdminService(int index) {
+    return cluster.getResourceManager(index).getRMContext().getRMAdminService();
+  }
+
+  private void explicitFailover() throws IOException {
+    int activeRMIndex = cluster.getActiveRMIndex();
+    int newActiveRMIndex = (activeRMIndex + 1) % 2;
+    getAdminService(activeRMIndex).transitionToStandby(req);
+    getAdminService(newActiveRMIndex).transitionToActive(req);
+    assertEquals("Failover failed", newActiveRMIndex, cluster.getActiveRMIndex());
+  }
+
+  private void failover()
+      throws IOException, InterruptedException, YarnException {
+    int activeRMIndex = cluster.getActiveRMIndex();
+    cluster.stopResourceManager(activeRMIndex);
+    assertEquals("Failover failed",
+        (activeRMIndex + 1) % 2, cluster.getActiveRMIndex());
+    cluster.restartResourceManager(activeRMIndex);
+  }
+
   @Test
   public void testExplicitFailover()
       throws YarnException, InterruptedException, IOException {
-    assertTrue("NMs failed to connect to the RM",
+    conf.setBoolean(YarnConfiguration.AUTO_FAILOVER_ENABLED, false);
-        cluster.waitForNodeManagersToConnect(5000));
+    cluster.init(conf);
-    verifyClientConnection();
+    cluster.start();
+    getAdminService(0).transitionToActive(req);
+    assertFalse("RM never turned active", -1 == cluster.getActiveRMIndex());
+    verifyConnections();
 
-    // Failover to the second RM
+    explicitFailover();
-    getRMAdminService(0).transitionToStandby(req);
+    verifyConnections();
-    getRMAdminService(1).transitionToActive(req);
-    assertEquals("Wrong ResourceManager is active",
-        HAServiceProtocol.HAServiceState.ACTIVE,
-        getRMAdminService(1).getServiceStatus().getState());
-    assertTrue("NMs failed to connect to the RM",
-        cluster.waitForNodeManagersToConnect(5000));
-    verifyClientConnection();
 
-    // Failover back to the first RM
+    explicitFailover();
-    getRMAdminService(1).transitionToStandby(req);
+    verifyConnections();
-    getRMAdminService(0).transitionToActive(req);
+  }
-    assertEquals("Wrong ResourceManager is active",
+
-        HAServiceProtocol.HAServiceState.ACTIVE,
+  @Test
-        getRMAdminService(0).getServiceStatus().getState());
+  public void testAutomaticFailover()
-    assertTrue("NMs failed to connect to the RM",
+      throws YarnException, InterruptedException, IOException {
-        cluster.waitForNodeManagersToConnect(5000));
+    conf.setBoolean(YarnConfiguration.AUTO_FAILOVER_ENABLED, true);
-    verifyClientConnection();
+    conf.setBoolean(YarnConfiguration.AUTO_FAILOVER_EMBEDDED, true);
+    conf.set(YarnConfiguration.RM_CLUSTER_ID, "yarn-test-cluster");
+    conf.set(YarnConfiguration.RM_ZK_ADDRESS, hostPort);
+    conf.setInt(YarnConfiguration.RM_ZK_TIMEOUT_MS, 2000);
+
+    cluster.init(conf);
+    cluster.start();
+    assertFalse("RM never turned active", -1 == cluster.getActiveRMIndex());
+    verifyConnections();
+
+    failover();
+    verifyConnections();
+
+    failover();
+    verifyConnections();
+  }
+
+  @Test
+  public void testWebAppProxyInStandAloneMode() throws YarnException,
+      InterruptedException, IOException {
+    WebAppProxyServer webAppProxyServer = new WebAppProxyServer();
+    try {
+      conf.set(YarnConfiguration.PROXY_ADDRESS, "0.0.0.0:9099");
+      cluster.init(conf);
+      cluster.start();
+      getAdminService(0).transitionToActive(req);
+      assertFalse("RM never turned active", -1 == cluster.getActiveRMIndex());
+      verifyConnections();
+      webAppProxyServer.init(conf);
+
+      // Start webAppProxyServer
+      Assert.assertEquals(STATE.INITED, webAppProxyServer.getServiceState());
+      webAppProxyServer.start();
+      Assert.assertEquals(STATE.STARTED, webAppProxyServer.getServiceState());
+
+      URL wrongUrl = new URL("http://0.0.0.0:9099/proxy/" + fakeAppId);
+      HttpURLConnection proxyConn = (HttpURLConnection) wrongUrl
+          .openConnection();
+
+      proxyConn.connect();
+      verifyExpectedException(proxyConn.getResponseMessage());
+
+      explicitFailover();
+      verifyConnections();
+      proxyConn.connect();
+      verifyExpectedException(proxyConn.getResponseMessage());
+    } finally {
+      webAppProxyServer.stop();
+    }
+  }
+
+  @Test
+  public void testEmbeddedWebAppProxy() throws YarnException,
+      InterruptedException, IOException {
+    cluster.init(conf);
+    cluster.start();
+    getAdminService(0).transitionToActive(req);
+    assertFalse("RM never turned active", -1 == cluster.getActiveRMIndex());
+    verifyConnections();
+    URL wrongUrl = new URL("http://0.0.0.0:18088/proxy/" + fakeAppId);
+    HttpURLConnection proxyConn = (HttpURLConnection) wrongUrl
+        .openConnection();
+
+    proxyConn.connect();
+    verifyExpectedException(proxyConn.getResponseMessage());
+
+    explicitFailover();
+    verifyConnections();
+    proxyConn.connect();
+    verifyExpectedException(proxyConn.getResponseMessage());
+  }
+
+  private void verifyExpectedException(String exceptionMessage){
+    assertTrue(exceptionMessage.contains(ApplicationNotFoundException.class
+        .getName()));
+    assertTrue(exceptionMessage
+        .contains("Application with id '" + fakeAppId + "' " +
+            "doesn't exist in RM."));
   }
 }

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common/src/main/java/org/apache/hadoop/yarn/client/RMHAServiceTarget.java

@@ -21,16 +21,19 @@ package org.apache.hadoop.yarn.client;
 import org.apache.hadoop.ha.BadFencingConfigurationException;
 import org.apache.hadoop.ha.HAServiceTarget;
 import org.apache.hadoop.ha.NodeFencer;
+import org.apache.hadoop.yarn.conf.HAUtil;
 import org.apache.hadoop.yarn.conf.YarnConfiguration;
 
 import java.io.IOException;
 import java.net.InetSocketAddress;
 
 public class RMHAServiceTarget extends HAServiceTarget {
-  private InetSocketAddress haAdminServiceAddress;
+  private final boolean autoFailoverEnabled;
+  private final InetSocketAddress haAdminServiceAddress;
 
   public RMHAServiceTarget(YarnConfiguration conf)
       throws IOException {
+    autoFailoverEnabled = HAUtil.isAutomaticFailoverEnabled(conf);
     haAdminServiceAddress = conf.getSocketAddr(
         YarnConfiguration.RM_ADMIN_ADDRESS,
         YarnConfiguration.DEFAULT_RM_ADMIN_ADDRESS,
@@ -44,19 +47,23 @@ public class RMHAServiceTarget extends HAServiceTarget {
 
   @Override
   public InetSocketAddress getZKFCAddress() {
-    // TODO (YARN-1177): Hook up ZKFC information
+    // TODO (YARN-1177): ZKFC implementation
-    return null;
+    throw new UnsupportedOperationException("RMHAServiceTarget doesn't have " +
+        "a corresponding ZKFC address");
   }
 
   @Override
   public NodeFencer getFencer() {
-    // TODO (YARN-1026): Hook up fencing implementation
     return null;
   }
 
   @Override
-  public void checkFencingConfigured()
+  public void checkFencingConfigured() throws BadFencingConfigurationException {
-      throws BadFencingConfigurationException {
+    throw new BadFencingConfigurationException("Fencer not configured");
-    // TODO (YARN-1026): Based on fencing implementation
+  }
+
+  @Override
+  public boolean isAutoFailoverEnabled() {
+    return autoFailoverEnabled;
   }
 }

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common/src/main/resources/yarn-default.xml

@@ -264,13 +264,20 @@
 
   <property>
     <description>Enable RM to recover state after starting. If true, then 
-    yarn.resourcemanager.store.class must be specified</description>
+      yarn.resourcemanager.store.class must be specified. </description>
     <name>yarn.resourcemanager.recovery.enabled</name>
     <value>false</value>
   </property>
 
   <property>
-    <description>The class to use as the persistent store.</description>
+    <description>The class to use as the persistent store.
+
+      If org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore
+      is used, the store is implicitly fenced; meaning a single ResourceManager
+      is able to use the store at any point in time. More details on this
+      implicit fencing, along with setting up appropriate ACLs is discussed
+      under yarn.resourcemanager.zk-state-store.root-node.acl.
+    </description>
     <name>yarn.resourcemanager.store.class</name>
     <value>org.apache.hadoop.yarn.server.resourcemanager.recovery.FileSystemRMStateStore</value>
   </property>
@@ -291,31 +298,24 @@
   </property>
 
   <property>
-    <description>Host:Port of the ZooKeeper server where RM state will 
+    <description>Host:Port of the ZooKeeper server to be used by the RM. This
-    be stored. This must be supplied when using
+      must be supplied when using the ZooKeeper based implementation of the
-    org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore
+      RM state store and/or embedded automatic failover in a HA setting.
-    as the value for yarn.resourcemanager.store.class. ZKRMStateStore
+    </description>
-    is implicitly fenced, meaning a single ResourceManager is
+    <name>yarn.resourcemanager.zk-address</name>
-    able to use the store at any point in time. More details on this, along
-    with setting up appropriate ACLs is discussed under the description for
-    yarn.resourcemanager.zk-state-store.root-node.acl.</description>
-    <name>yarn.resourcemanager.zk-state-store.address</name>
     <!--value>127.0.0.1:2181</value-->
   </property>
 
   <property>
-    <description>Number of times ZKRMStateStore tries to connect to
+    <description>Number of times RM tries to connect to ZooKeeper.</description>
-    ZooKeeper. This may be supplied when using
+    <name>yarn.resourcemanager.zk-num-retries</name>
-    org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore
-    as the value for yarn.resourcemanager.store.class</description>
-    <name>yarn.resourcemanager.zk-state-store.num-retries</name>
     <value>500</value>
   </property>
 
   <property>
-    <description>Retry interval in milliseconds when ZKRMStateStore tries to
+    <description>Retry interval in milliseconds when connecting to ZooKeeper.
-    connect to ZooKeeper.</description>
+    </description>
-    <name>yarn.resourcemanager.zk-state-store.retry-interval-ms</name>
+    <name>yarn.resourcemanager.zk-retry-interval-ms</name>
     <value>2000</value>
   </property>
 
@@ -333,20 +333,14 @@
     is managed by the ZooKeeper cluster itself, not by the client. This value is
     used by the cluster to determine when the client's session expires.
     Expirations happens when the cluster does not hear from the client within
-    the specified session timeout period (i.e. no heartbeat).
+    the specified session timeout period (i.e. no heartbeat).</description>
-    This may be supplied when using
+    <name>yarn.resourcemanager.zk-timeout-ms</name>
-    org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore
-    as the value for yarn.resourcemanager.store.class</description>
-    <name>yarn.resourcemanager.zk-state-store.timeout-ms</name>
     <value>60000</value>
   </property>
 
   <property>
-    <description>ACL's to be used for ZooKeeper znodes.
+    <description>ACL's to be used for ZooKeeper znodes.</description>
-    This may be supplied when using
+    <name>yarn.resourcemanager.zk-acl</name>
-    org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore
-    as the value for yarn.resourcemanager.store.class</description>
-    <name>yarn.resourcemanager.zk-state-store.acl</name>
     <value>world:anyone:rwcda</value>
   </property>
 
@@ -362,7 +356,7 @@
       permissions.
 
       By default, when this property is not set, we use the ACLs from
-      yarn.resourcemanager.zk-state-store.acl for shared admin access and
+      yarn.resourcemanager.zk-acl for shared admin access and
       rm-address:cluster-timestamp for username-based exclusive create-delete
       access.
 
@@ -408,6 +402,36 @@
     <value>false</value>
   </property>
 
+  <property>
+    <description>Enable automatic failover.</description>
+    <name>yarn.resourcemanager.ha.automatic-failover.enabled</name>
+    <value>false</value>
+  </property>
+
+  <property>
+    <description>Enable embedded automatic failover. The embedded elector
+      relies on the RM state store to handle fencing, and is primarily intended
+      to be used in conjunction with ZKRMStateStore.</description>
+    <name>yarn.resourcemanager.ha.automatic-failover.embedded</name>
+    <value>false</value>
+  </property>
+
+  <property>
+    <description>The base znode path to use for storing leader information,
+      when using ZooKeeper based leader election.</description>
+    <name>yarn.resourcemanager.ha.automatic-failover.zk-base-path</name>
+    <value>/yarn-leader-election</value>
+  </property>
+
+  <property>
+    <description>Name of the cluster. In a HA setting,
+      this is used to ensure the RM participates in leader
+      election fo this cluster and ensures it does not affect
+      other clusters</description>
+    <name>yarn.resourcemanager.cluster-id</name>
+    <!--value>yarn-cluster</value-->
+  </property>
+
   <property>
     <description>The list of RM nodes in the cluster when HA is
       enabled. See description of yarn.resourcemanager.ha

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/containermanager/loghandler/NonAggregatingLogHandler.java

@@ -23,6 +23,7 @@ import java.util.concurrent.ConcurrentHashMap;
 import java.util.concurrent.ScheduledThreadPoolExecutor;
 import java.util.concurrent.ThreadFactory;
 import java.util.concurrent.TimeUnit;
+import java.util.concurrent.RejectedExecutionException;
 
 import org.apache.commons.logging.Log;
 import org.apache.commons.logging.LogFactory;
@@ -118,10 +119,17 @@ public class NonAggregatingLogHandler extends AbstractService implements
         LOG.info("Scheduling Log Deletion for application: "
             + appFinishedEvent.getApplicationId() + ", with delay of "
             + this.deleteDelaySeconds + " seconds");
-        sched.schedule(
+        LogDeleterRunnable logDeleter =
             new LogDeleterRunnable(appOwners.remove(appFinishedEvent
-                .getApplicationId()), appFinishedEvent.getApplicationId()),
+                  .getApplicationId()), appFinishedEvent.getApplicationId());
-            this.deleteDelaySeconds, TimeUnit.SECONDS);
+        try {
+          sched.schedule(logDeleter, this.deleteDelaySeconds,
+              TimeUnit.SECONDS);
+        } catch (RejectedExecutionException e) {
+          // Handling this event in local thread before starting threads
+          // or after calling sched.shutdownNow().
+          logDeleter.run();
+        }
         break;
       default:
         ; // Ignore

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/test/java/org/apache/hadoop/yarn/server/nodemanager/containermanager/launcher/TestContainerLaunch.java

@@ -260,8 +260,12 @@ public class TestContainerLaunch extends BaseContainerManagerTest {
       fos.flush();
       fos.close();
 
+      // It is supposed that LANG is set as C.
+      Map<String, String> cmdEnv = new HashMap<String, String>();
+      cmdEnv.put("LANG", "C");
       Shell.ShellCommandExecutor shexc
-      = new Shell.ShellCommandExecutor(new String[]{shellFile.getAbsolutePath()}, tmpDir);
+      = new Shell.ShellCommandExecutor(new String[]{shellFile.getAbsolutePath()},
+        tmpDir, cmdEnv);
       String diagnostics = null;
       try {
         shexc.execute();

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/test/java/org/apache/hadoop/yarn/server/nodemanager/containermanager/loghandler/TestNonAggregatingLogHandler.java

@@ -35,6 +35,7 @@ import org.apache.hadoop.yarn.conf.YarnConfiguration;
 import org.apache.hadoop.yarn.event.Dispatcher;
 import org.apache.hadoop.yarn.event.DrainDispatcher;
 import org.apache.hadoop.yarn.event.EventHandler;
+import org.apache.hadoop.yarn.event.InlineDispatcher;
 import org.apache.hadoop.yarn.logaggregation.ContainerLogsRetentionPolicy;
 import org.apache.hadoop.yarn.server.nodemanager.DeletionService;
 import org.apache.hadoop.yarn.server.nodemanager.LocalDirsHandlerService;
@@ -202,6 +203,32 @@ public class TestNonAggregatingLogHandler {
     verify(logHandler.mockSched).shutdownNow();
   }
 
+  @Test
+  public void testHandlingApplicationFinishedEvent() {
+    Configuration conf = new Configuration();
+    LocalDirsHandlerService dirsService  = new LocalDirsHandlerService();
+    DeletionService delService = new DeletionService(null);
+    NonAggregatingLogHandler aggregatingLogHandler =
+        new NonAggregatingLogHandler(new InlineDispatcher(),
+            delService,
+            dirsService);
+
+    dirsService.init(conf);
+    dirsService.start();
+    delService.init(conf);
+    delService.start();
+    aggregatingLogHandler.init(conf);
+    aggregatingLogHandler.start();
+    ApplicationId appId = BuilderUtils.newApplicationId(1234, 1);
+    // It should NOT throw RejectedExecutionException
+    aggregatingLogHandler.handle(new LogHandlerAppFinishedEvent(appId));
+    aggregatingLogHandler.stop();
+
+    // It should NOT throw RejectedExecutionException after stopping
+    // handler service.
+    aggregatingLogHandler.handle(new LogHandlerAppFinishedEvent(appId));
+  }
+
   private class NonAggregatingLogHandlerWithMockExecutor extends
       NonAggregatingLogHandler {
 

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/AdminService.java

@@ -20,11 +20,8 @@ package org.apache.hadoop.yarn.server.resourcemanager;
 
 import java.io.IOException;
 import java.net.InetSocketAddress;
-import java.util.ArrayList;
-import java.util.List;
 import java.util.Map;
 import java.util.Set;
-import java.util.concurrent.ConcurrentMap;
 
 import org.apache.commons.logging.Log;
 import org.apache.commons.logging.LogFactory;
@@ -41,14 +38,16 @@ import org.apache.hadoop.ipc.ProtobufRpcEngine;
 import org.apache.hadoop.ipc.RPC;
 import org.apache.hadoop.ipc.RPC.Server;
 import org.apache.hadoop.ipc.StandbyException;
+import org.apache.hadoop.security.AccessControlException;
 import org.apache.hadoop.security.Groups;
 import org.apache.hadoop.security.UserGroupInformation;
 import org.apache.hadoop.security.authorize.AccessControlList;
 import org.apache.hadoop.security.authorize.PolicyProvider;
 import org.apache.hadoop.security.authorize.ProxyUsers;
-import org.apache.hadoop.service.AbstractService;
+import org.apache.hadoop.service.CompositeService;
 import org.apache.hadoop.yarn.api.records.NodeId;
 import org.apache.hadoop.yarn.api.records.ResourceOption;
+import org.apache.hadoop.yarn.conf.HAUtil;
 import org.apache.hadoop.yarn.conf.YarnConfiguration;
 import org.apache.hadoop.yarn.exceptions.YarnException;
 import org.apache.hadoop.yarn.factories.RecordFactory;
@@ -75,7 +74,7 @@ import org.apache.hadoop.yarn.server.resourcemanager.security.authorize.RMPolicy
 
 import com.google.protobuf.BlockingService;
 
-public class AdminService extends AbstractService implements
+public class AdminService extends CompositeService implements
     HAServiceProtocol, ResourceManagerAdministrationProtocol {
 
   private static final Log LOG = LogFactory.getLog(AdminService.class);
@@ -84,6 +83,8 @@ public class AdminService extends AbstractService implements
   private final ResourceManager rm;
   private String rmId;
 
+  private boolean autoFailoverEnabled;
+
   private Server server;
   private InetSocketAddress masterServiceAddress;
   private AccessControlList adminAcl;
@@ -99,6 +100,15 @@ public class AdminService extends AbstractService implements
 
   @Override
   public synchronized void serviceInit(Configuration conf) throws Exception {
+    if (rmContext.isHAEnabled()) {
+      autoFailoverEnabled = HAUtil.isAutomaticFailoverEnabled(conf);
+      if (autoFailoverEnabled) {
+        if (HAUtil.isAutomaticFailoverEmbedded(conf)) {
+          addIfService(createEmbeddedElectorService());
+        }
+      }
+    }
+
     masterServiceAddress = conf.getSocketAddr(
         YarnConfiguration.RM_ADMIN_ADDRESS,
         YarnConfiguration.DEFAULT_RM_ADMIN_ADDRESS,
@@ -162,6 +172,10 @@ public class AdminService extends AbstractService implements
     }
   }
 
+  protected EmbeddedElectorService createEmbeddedElectorService() {
+    return new EmbeddedElectorService(rmContext);
+  }
+
   private UserGroupInformation checkAccess(String method) throws IOException {
     return RMServerUtils.verifyAccess(adminAcl, method, LOG);
   }
@@ -174,6 +188,43 @@ public class AdminService extends AbstractService implements
     }
   }
 
+  /**
+   * Check that a request to change this node's HA state is valid.
+   * In particular, verifies that, if auto failover is enabled, non-forced
+   * requests from the HAAdmin CLI are rejected, and vice versa.
+   *
+   * @param req the request to check
+   * @throws AccessControlException if the request is disallowed
+   */
+  private void checkHaStateChange(StateChangeRequestInfo req)
+      throws AccessControlException {
+    switch (req.getSource()) {
+      case REQUEST_BY_USER:
+        if (autoFailoverEnabled) {
+          throw new AccessControlException(
+              "Manual failover for this ResourceManager is disallowed, " +
+                  "because automatic failover is enabled.");
+        }
+        break;
+      case REQUEST_BY_USER_FORCED:
+        if (autoFailoverEnabled) {
+          LOG.warn("Allowing manual failover from " +
+              org.apache.hadoop.ipc.Server.getRemoteAddress() +
+              " even though automatic failover is enabled, because the user " +
+              "specified the force flag");
+        }
+        break;
+      case REQUEST_BY_ZKFC:
+        if (!autoFailoverEnabled) {
+          throw new AccessControlException(
+              "Request from ZK failover controller at " +
+                  org.apache.hadoop.ipc.Server.getRemoteAddress() + " denied " +
+                  "since automatic failover is not enabled");
+        }
+        break;
+    }
+  }
+
   private synchronized boolean isRMActive() {
     return HAServiceState.ACTIVE == rmContext.getHAServiceState();
   }
@@ -196,8 +247,7 @@ public class AdminService extends AbstractService implements
   public synchronized void transitionToActive(
       HAServiceProtocol.StateChangeRequestInfo reqInfo) throws IOException {
     UserGroupInformation user = checkAccess("transitionToActive");
-    // TODO (YARN-1177): When automatic failover is enabled,
+    checkHaStateChange(reqInfo);
-    // check if transition should be allowed for this request
     try {
       rm.transitionToActive();
       RMAuditLogger.logSuccess(user.getShortUserName(),
@@ -215,8 +265,7 @@ public class AdminService extends AbstractService implements
   public synchronized void transitionToStandby(
       HAServiceProtocol.StateChangeRequestInfo reqInfo) throws IOException {
     UserGroupInformation user = checkAccess("transitionToStandby");
-    // TODO (YARN-1177): When automatic failover is enabled,
+    checkHaStateChange(reqInfo);
-    // check if transition should be allowed for this request
     try {
       rm.transitionToStandby(true);
       RMAuditLogger.logSuccess(user.getShortUserName(),
@@ -394,7 +443,7 @@ public class AdminService extends AbstractService implements
   public String[] getGroupsForUser(String user) throws IOException {
     return UserGroupInformation.createRemoteUser(user).getGroupNames();
   }
-  
+
   @Override
   public UpdateNodeResourceResponse updateNodeResource(
       UpdateNodeResourceRequest request) throws YarnException, IOException {

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/EmbeddedElectorService.java

@@ -0,0 +1,209 @@
+/**
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package org.apache.hadoop.yarn.server.resourcemanager;
+
+import com.google.protobuf.InvalidProtocolBufferException;
+import org.apache.commons.logging.Log;
+import org.apache.commons.logging.LogFactory;
+import org.apache.hadoop.classification.InterfaceAudience;
+import org.apache.hadoop.classification.InterfaceStability;
+import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.ha.ActiveStandbyElector;
+import org.apache.hadoop.ha.HAServiceProtocol;
+import org.apache.hadoop.ha.ServiceFailedException;
+import org.apache.hadoop.service.AbstractService;
+import org.apache.hadoop.util.StringUtils;
+import org.apache.hadoop.util.ZKUtil;
+import org.apache.hadoop.yarn.conf.HAUtil;
+import org.apache.hadoop.yarn.conf.YarnConfiguration;
+import org.apache.hadoop.yarn.event.Dispatcher;
+import org.apache.hadoop.yarn.exceptions.YarnRuntimeException;
+import org.apache.hadoop.yarn.proto.YarnServerResourceManagerServiceProtos;
+import org.apache.zookeeper.KeeperException;
+import org.apache.zookeeper.data.ACL;
+
+import java.io.IOException;
+import java.util.Collections;
+import java.util.List;
+
+@InterfaceAudience.Private
+@InterfaceStability.Unstable
+public class EmbeddedElectorService extends AbstractService
+    implements ActiveStandbyElector.ActiveStandbyElectorCallback {
+  private static final Log LOG =
+      LogFactory.getLog(EmbeddedElectorService.class.getName());
+  private static final HAServiceProtocol.StateChangeRequestInfo req =
+      new HAServiceProtocol.StateChangeRequestInfo(
+          HAServiceProtocol.RequestSource.REQUEST_BY_ZKFC);
+
+  private RMContext rmContext;
+
+  private byte[] localActiveNodeInfo;
+  private ActiveStandbyElector elector;
+
+  EmbeddedElectorService(RMContext rmContext) {
+    super(EmbeddedElectorService.class.getName());
+    this.rmContext = rmContext;
+  }
+
+  @Override
+  protected synchronized void serviceInit(Configuration conf)
+      throws Exception {
+    conf = conf instanceof YarnConfiguration ? conf : new YarnConfiguration(conf);
+
+    String zkQuorum = conf.get(YarnConfiguration.RM_ZK_ADDRESS);
+    if (zkQuorum == null) {
+     throw new YarnRuntimeException("Embedded automatic failover " +
+          "is enabled, but " + YarnConfiguration.RM_ZK_ADDRESS +
+          " is not set");
+    }
+
+    String rmId = HAUtil.getRMHAId(conf);
+    String clusterId = conf.get(YarnConfiguration.RM_CLUSTER_ID);
+    if (clusterId == null) {
+      throw new YarnRuntimeException(YarnConfiguration.RM_CLUSTER_ID +
+          " is not specified!");
+    }
+    localActiveNodeInfo = createActiveNodeInfo(clusterId, rmId);
+
+    String zkBasePath = conf.get(YarnConfiguration.AUTO_FAILOVER_ZK_BASE_PATH,
+        YarnConfiguration.DEFAULT_AUTO_FAILOVER_ZK_BASE_PATH);
+    String electionZNode = zkBasePath + "/" + clusterId;
+
+    long zkSessionTimeout = conf.getLong(YarnConfiguration.RM_ZK_TIMEOUT_MS,
+        YarnConfiguration.DEFAULT_RM_ZK_TIMEOUT_MS);
+
+    String zkAclConf = conf.get(YarnConfiguration.RM_ZK_ACL,
+        YarnConfiguration.DEFAULT_RM_ZK_ACL);
+    List<ACL> zkAcls;
+    try {
+      zkAcls = ZKUtil.parseACLs(ZKUtil.resolveConfIndirection(zkAclConf));
+    } catch (ZKUtil.BadAclFormatException bafe) {
+      throw new YarnRuntimeException(
+          YarnConfiguration.RM_ZK_ACL + "has ill-formatted ACLs");
+    }
+
+    // TODO (YARN-1528): ZKAuthInfo to be set for rm-store and elector
+    List<ZKUtil.ZKAuthInfo> zkAuths = Collections.emptyList();
+
+    elector = new ActiveStandbyElector(zkQuorum, (int) zkSessionTimeout,
+        electionZNode, zkAcls, zkAuths, this);
+
+    elector.ensureParentZNode();
+    if (!isParentZnodeSafe(clusterId)) {
+      notifyFatalError(electionZNode + " znode has invalid data! "+
+          "Might need formatting!");
+    }
+
+    super.serviceInit(conf);
+  }
+
+  @Override
+  protected synchronized void serviceStart() throws Exception {
+    elector.joinElection(localActiveNodeInfo);
+    super.serviceStart();
+  }
+
+  @Override
+  protected synchronized void serviceStop() throws Exception {
+    elector.quitElection(false);
+    elector.terminateConnection();
+    super.serviceStop();
+  }
+
+  @Override
+  public synchronized void becomeActive() throws ServiceFailedException {
+    try {
+      rmContext.getRMAdminService().transitionToActive(req);
+    } catch (Exception e) {
+      throw new ServiceFailedException("RM could not transition to Active", e);
+    }
+  }
+
+  @Override
+  public synchronized void becomeStandby() {
+    try {
+      rmContext.getRMAdminService().transitionToStandby(req);
+    } catch (Exception e) {
+      LOG.error("RM could not transition to Standby", e);
+    }
+  }
+
+  @Override
+  public void enterNeutralMode() {
+    /**
+     * Possibly due to transient connection issues. Do nothing.
+     * TODO: Might want to keep track of how long in this state and transition
+     * to standby.
+     */
+  }
+
+  @SuppressWarnings(value = "unchecked")
+  @Override
+  public synchronized void notifyFatalError(String errorMessage) {
+    rmContext.getDispatcher().getEventHandler().handle(
+        new RMFatalEvent(RMFatalEventType.EMBEDDED_ELECTOR_FAILED, errorMessage));
+  }
+
+  @Override
+  public synchronized void fenceOldActive(byte[] oldActiveData) {
+    if (LOG.isDebugEnabled()) {
+      LOG.debug("Request to fence old active being ignored, " +
+          "as embedded leader election doesn't support fencing");
+    }
+  }
+
+  private static byte[] createActiveNodeInfo(String clusterId, String rmId)
+      throws IOException {
+    return YarnServerResourceManagerServiceProtos.ActiveRMInfoProto
+        .newBuilder()
+        .setClusterId(clusterId)
+        .setRmId(rmId)
+        .build()
+        .toByteArray();
+  }
+
+  private synchronized boolean isParentZnodeSafe(String clusterId)
+      throws InterruptedException, IOException, KeeperException {
+    byte[] data;
+    try {
+      data = elector.getActiveData();
+    } catch (ActiveStandbyElector.ActiveNotFoundException e) {
+      // no active found, parent znode is safe
+      return true;
+    }
+
+    YarnServerResourceManagerServiceProtos.ActiveRMInfoProto proto;
+    try {
+      proto = YarnServerResourceManagerServiceProtos.ActiveRMInfoProto
+          .parseFrom(data);
+    } catch (InvalidProtocolBufferException e) {
+      LOG.error("Invalid data in ZK: " + StringUtils.byteToHexString(data));
+      return false;
+    }
+
+    // Check if the passed proto corresponds to an RM in the same cluster
+    if (!proto.getClusterId().equals(clusterId)) {
+      LOG.error("Mismatched cluster! The other RM seems " +
+          "to be from a different cluster. Current cluster = " + clusterId +
+          "Other RM's cluster = " + proto.getClusterId());
+      return false;
+    }
+    return true;
+  }
+}

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/RMFatalEvent.java

@@ -0,0 +1,37 @@
+/**
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package org.apache.hadoop.yarn.server.resourcemanager;
+
+import org.apache.hadoop.util.StringUtils;
+import org.apache.hadoop.yarn.event.AbstractEvent;
+
+public class RMFatalEvent extends AbstractEvent<RMFatalEventType> {
+  private String cause;
+
+  public RMFatalEvent(RMFatalEventType rmFatalEventType, String cause) {
+    super(rmFatalEventType);
+    this.cause = cause;
+  }
+
+  public RMFatalEvent(RMFatalEventType rmFatalEventType, Exception cause) {
+    super(rmFatalEventType);
+    this.cause = StringUtils.stringifyException(cause);
+  }
+
+  public String getCause() {return this.cause;}
+}

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/RMFatalEventType.java

@@ -15,9 +15,17 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
-package org.apache.hadoop.yarn.server.resourcemanager.recovery;
 
-public enum RMStateStoreOperationFailedEventType {
+package org.apache.hadoop.yarn.server.resourcemanager;
-  FENCED, // Store operation failed because it was fenced
+
-  FAILED // Store operation failed for no known reason
+import org.apache.hadoop.classification.InterfaceAudience;
+
+@InterfaceAudience.Private
+public enum RMFatalEventType {
+  // Source <- Store
+  STATE_STORE_FENCED,
+  STATE_STORE_OP_FAILED,
+
+  // Source <- Embedded Elector
+  EMBEDDED_ELECTOR_FAILED
 }

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ResourceManager.java

@@ -59,8 +59,6 @@ import org.apache.hadoop.yarn.server.resourcemanager.recovery.NullRMStateStore;
 import org.apache.hadoop.yarn.server.resourcemanager.recovery.RMStateStore;
 import org.apache.hadoop.yarn.server.resourcemanager.recovery.RMStateStore.RMState;
 import org.apache.hadoop.yarn.server.resourcemanager.recovery.RMStateStoreFactory;
-import org.apache.hadoop.yarn.server.resourcemanager.recovery.RMStateStoreOperationFailedEvent;
-import org.apache.hadoop.yarn.server.resourcemanager.recovery.RMStateStoreOperationFailedEventType;
 import org.apache.hadoop.yarn.server.resourcemanager.recovery.Recoverable;
 import org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMApp;
 import org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMAppEvent;
@@ -121,6 +119,7 @@ public class ResourceManager extends CompositeService implements Recoverable {
    */
   @VisibleForTesting
   protected RMContextImpl rmContext;
+  private Dispatcher rmDispatcher;
   @VisibleForTesting
   protected AdminService adminService;
 
@@ -134,7 +133,6 @@ public class ResourceManager extends CompositeService implements Recoverable {
    */
   protected RMActiveServices activeServices;
   protected RMSecretManagerService rmSecretManagerService;
-  private Dispatcher rmDispatcher;
 
   protected ResourceScheduler scheduler;
   private ClientRMService clientRM;
@@ -149,9 +147,12 @@ public class ResourceManager extends CompositeService implements Recoverable {
   protected QueueACLsManager queueACLsManager;
   private DelegationTokenRenewer delegationTokenRenewer;
   private WebApp webApp;
+  private AppReportFetcher fetcher = null;
   protected ResourceTrackerService resourceTracker;
   private boolean recoveryEnabled;
 
+  private String webAppAddress;
+
   /** End of Active services */
 
   private Configuration conf;
@@ -179,6 +180,13 @@ public class ResourceManager extends CompositeService implements Recoverable {
     this.conf = conf;
     this.rmContext = new RMContextImpl();
 
+    rmDispatcher = createDispatcher();
+    addIfService(rmDispatcher);
+    rmContext.setDispatcher(rmDispatcher);
+
+    rmDispatcher.register(RMFatalEventType.class,
+        new ResourceManager.RMFatalEventDispatcher(this.rmContext, this));
+
     adminService = createAdminService();
     addService(adminService);
     rmContext.setRMAdminService(adminService);
@@ -189,6 +197,8 @@ public class ResourceManager extends CompositeService implements Recoverable {
     }
     createAndInitActiveServices();
 
+    webAppAddress = WebAppUtils.getRMWebAppURLWithoutScheme(conf);
+
     super.serviceInit(conf);
   }
   
@@ -207,11 +217,6 @@ public class ResourceManager extends CompositeService implements Recoverable {
     return new SchedulerEventDispatcher(this.scheduler);
   }
 
-  protected RMStateStoreOperationFailedEventDispatcher
-      createRMStateStoreOperationFailedEventDispatcher() {
-    return new RMStateStoreOperationFailedEventDispatcher(rmContext, this);
-  }
-
   protected Dispatcher createDispatcher() {
     return new AsyncDispatcher();
   }
@@ -297,10 +302,6 @@ public class ResourceManager extends CompositeService implements Recoverable {
     protected void serviceInit(Configuration configuration) throws Exception {
       conf.setBoolean(Dispatcher.DISPATCHER_EXIT_ON_ERROR_KEY, true);
 
-      rmDispatcher = createDispatcher();
-      addIfService(rmDispatcher);
-      rmContext.setDispatcher(rmDispatcher);
-
       rmSecretManagerService = createRMSecretManagerService();
       addService(rmSecretManagerService);
 
@@ -332,8 +333,6 @@ public class ResourceManager extends CompositeService implements Recoverable {
       try {
         rmStore.init(conf);
         rmStore.setRMDispatcher(rmDispatcher);
-        rmDispatcher.register(RMStateStoreOperationFailedEventType.class,
-            createRMStateStoreOperationFailedEventDispatcher());
       } catch (Exception e) {
         // the Exception from stateStore.init() needs to be handled for
         // HA and we need to give up master status if we got fenced
@@ -443,22 +442,12 @@ public class ResourceManager extends CompositeService implements Recoverable {
           throw e;
         }
       }
-      startWepApp();
-
-      if (getConfig().getBoolean(YarnConfiguration.IS_MINI_YARN_CLUSTER, false)) {
-        int port = webApp.port();
-        WebAppUtils.setRMWebAppPort(conf, port);
-      }
 
       super.serviceStart();
     }
 
     @Override
     protected void serviceStop() throws Exception {
-      if (webApp != null) {
-        webApp.stop();
-      }
-
 
       DefaultMetricsSystem.shutdown();
 
@@ -605,26 +594,23 @@ public class ResourceManager extends CompositeService implements Recoverable {
   }
 
   @Private
-  public static class RMStateStoreOperationFailedEventDispatcher implements
+  public static class RMFatalEventDispatcher
-      EventHandler<RMStateStoreOperationFailedEvent> {
+      implements EventHandler<RMFatalEvent> {
-
     private final RMContext rmContext;
     private final ResourceManager rm;
 
-    public RMStateStoreOperationFailedEventDispatcher(RMContext rmContext,
+    public RMFatalEventDispatcher(
-        ResourceManager resourceManager) {
+        RMContext rmContext, ResourceManager resourceManager) {
       this.rmContext = rmContext;
       this.rm = resourceManager;
     }
 
     @Override
-    public void handle(RMStateStoreOperationFailedEvent event) {
+    public void handle(RMFatalEvent event) {
-      if (LOG.isDebugEnabled()) {
+      LOG.fatal("Received a " + RMFatalEvent.class.getName() + " of type " +
-        LOG.debug("Received a " +
+          event.getType().name());
-            RMStateStoreOperationFailedEvent.class.getName() + " of type " +
+
-            event.getType().name());
+      if (event.getType() == RMFatalEventType.STATE_STORE_FENCED) {
-      }
-      if (event.getType() == RMStateStoreOperationFailedEventType.FENCED) {
         LOG.info("RMStateStore has been fenced");
         if (rmContext.isHAEnabled()) {
           try {
@@ -633,14 +619,11 @@ public class ResourceManager extends CompositeService implements Recoverable {
             rm.transitionToStandby(true);
             return;
           } catch (Exception e) {
-            LOG.error("Failed to transition RM to Standby mode.");
+            LOG.fatal("Failed to transition RM to Standby mode.");
           }
         }
       }
 
-      LOG.error("Shutting down RM on receiving a " +
-          RMStateStoreOperationFailedEvent.class.getName() + " of type " +
-          event.getType().name());
       ExitUtil.terminate(1, event.getCause());
     }
   }
@@ -764,12 +747,16 @@ public class ResourceManager extends CompositeService implements Recoverable {
                 YarnConfiguration.RM_WEBAPP_SPNEGO_USER_NAME_KEY)
             .withHttpSpnegoKeytabKey(
                 YarnConfiguration.RM_WEBAPP_SPNEGO_KEYTAB_FILE_KEY)
-            .at(WebAppUtils.getRMWebAppURLWithoutScheme(conf)); 
+            .at(webAppAddress);
     String proxyHostAndPort = WebAppUtils.getProxyHostAndPort(conf);
     if(WebAppUtils.getResolvedRMWebAppURLWithoutScheme(conf).
         equals(proxyHostAndPort)) {
-      AppReportFetcher fetcher = new AppReportFetcher(conf, getClientRMService());
+      if (HAUtil.isHAEnabled(conf)) {
-      builder.withServlet(ProxyUriUtils.PROXY_SERVLET_NAME, 
+        fetcher = new AppReportFetcher(conf);
+      } else {
+        fetcher = new AppReportFetcher(conf, getClientRMService());
+      }
+      builder.withServlet(ProxyUriUtils.PROXY_SERVLET_NAME,
           ProxyUriUtils.PROXY_PATH_SPEC, WebAppProxyServlet.class);
       builder.withAttribute(WebAppProxy.FETCHER_ATTRIBUTE, fetcher);
       String[] proxyParts = proxyHostAndPort.split(":");
@@ -866,6 +853,11 @@ public class ResourceManager extends CompositeService implements Recoverable {
       transitionToActive();
     }
 
+    startWepApp();
+    if (getConfig().getBoolean(YarnConfiguration.IS_MINI_YARN_CLUSTER, false)) {
+      int port = webApp.port();
+      WebAppUtils.setRMWebAppPort(conf, port);
+    }
     super.serviceStart();
   }
   
@@ -876,6 +868,12 @@ public class ResourceManager extends CompositeService implements Recoverable {
 
   @Override
   protected void serviceStop() throws Exception {
+    if (webApp != null) {
+      webApp.stop();
+    }
+    if (fetcher != null) {
+      fetcher.stop();
+    }
     super.serviceStop();
     transitionToStandby(false);
     rmContext.setHAServiceState(HAServiceState.STOPPING);

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStore.java

@@ -48,6 +48,8 @@ import org.apache.hadoop.yarn.event.Dispatcher;
 import org.apache.hadoop.yarn.event.EventHandler;
 import org.apache.hadoop.yarn.security.AMRMTokenIdentifier;
 import org.apache.hadoop.yarn.security.client.RMDelegationTokenIdentifier;
+import org.apache.hadoop.yarn.server.resourcemanager.RMFatalEvent;
+import org.apache.hadoop.yarn.server.resourcemanager.RMFatalEventType;
 import org.apache.hadoop.yarn.server.resourcemanager.recovery.records.RMStateVersion;
 import org.apache.hadoop.yarn.server.resourcemanager.recovery.records.impl.pb.ApplicationAttemptStateDataPBImpl;
 import org.apache.hadoop.yarn.server.resourcemanager.recovery.records.impl.pb.ApplicationStateDataPBImpl;
@@ -679,15 +681,13 @@ public abstract class RMStateStore extends AbstractService {
    * @param failureCause the exception due to which the operation failed
    */
   private void notifyStoreOperationFailed(Exception failureCause) {
-    RMStateStoreOperationFailedEventType type;
+    RMFatalEventType type;
     if (failureCause instanceof StoreFencedException) {
-      type = RMStateStoreOperationFailedEventType.FENCED;
+      type = RMFatalEventType.STATE_STORE_FENCED;
     } else {
-      type = RMStateStoreOperationFailedEventType.FAILED;
+      type = RMFatalEventType.STATE_STORE_OP_FAILED;
     }
-
+    rmDispatcher.getEventHandler().handle(new RMFatalEvent(type, failureCause));
-    rmDispatcher.getEventHandler().handle(
-        new RMStateStoreOperationFailedEvent(type, failureCause));
   }
 
   @SuppressWarnings("unchecked")

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStoreOperationFailedEvent.java

@@ -1,36 +0,0 @@
-/**
-* Licensed to the Apache Software Foundation (ASF) under one
-* or more contributor license agreements.  See the NOTICE file
-* distributed with this work for additional information
-* regarding copyright ownership.  The ASF licenses this file
-* to you under the Apache License, Version 2.0 (the
-* "License"); you may not use this file except in compliance
-* with the License.  You may obtain a copy of the License at
-*
-*     http://www.apache.org/licenses/LICENSE-2.0
-*
-* Unless required by applicable law or agreed to in writing, software
-* distributed under the License is distributed on an "AS IS" BASIS,
-* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-* See the License for the specific language governing permissions and
-* limitations under the License.
-*/
-
-package org.apache.hadoop.yarn.server.resourcemanager.recovery;
-
-import org.apache.hadoop.yarn.event.AbstractEvent;
-
-public class RMStateStoreOperationFailedEvent
-    extends AbstractEvent<RMStateStoreOperationFailedEventType> {
-  private Exception cause;
-
-  RMStateStoreOperationFailedEvent(
-      RMStateStoreOperationFailedEventType type, Exception cause) {
-    super(type);
-    this.cause = cause;
-  }
-
-  public Exception getCause() {
-    return this.cause;
-  }
-}

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/ZKRMStateStore.java

@@ -182,34 +182,34 @@ public class ZKRMStateStore extends RMStateStore {
 
   @Override
   public synchronized void initInternal(Configuration conf) throws Exception {
-    zkHostPort = conf.get(YarnConfiguration.ZK_RM_STATE_STORE_ADDRESS);
+    zkHostPort = conf.get(YarnConfiguration.RM_ZK_ADDRESS);
     if (zkHostPort == null) {
       throw new YarnRuntimeException("No server address specified for " +
           "zookeeper state store for Resource Manager recovery. " +
-          YarnConfiguration.ZK_RM_STATE_STORE_ADDRESS + " is not configured.");
+          YarnConfiguration.RM_ZK_ADDRESS + " is not configured.");
     }
     numRetries =
-        conf.getInt(YarnConfiguration.ZK_RM_STATE_STORE_NUM_RETRIES,
+        conf.getInt(YarnConfiguration.RM_ZK_NUM_RETRIES,
-            YarnConfiguration.DEFAULT_ZK_RM_STATE_STORE_NUM_RETRIES);
+            YarnConfiguration.DEFAULT_ZK_RM_NUM_RETRIES);
     znodeWorkingPath =
         conf.get(YarnConfiguration.ZK_RM_STATE_STORE_PARENT_PATH,
             YarnConfiguration.DEFAULT_ZK_RM_STATE_STORE_PARENT_PATH);
     zkSessionTimeout =
-        conf.getInt(YarnConfiguration.ZK_RM_STATE_STORE_TIMEOUT_MS,
+        conf.getInt(YarnConfiguration.RM_ZK_TIMEOUT_MS,
-            YarnConfiguration.DEFAULT_ZK_RM_STATE_STORE_TIMEOUT_MS);
+            YarnConfiguration.DEFAULT_RM_ZK_TIMEOUT_MS);
     zkRetryInterval =
-        conf.getLong(YarnConfiguration.ZK_RM_STATE_STORE_RETRY_INTERVAL_MS,
+        conf.getLong(YarnConfiguration.RM_ZK_RETRY_INTERVAL_MS,
-          YarnConfiguration.DEFAULT_ZK_RM_STATE_STORE_RETRY_INTERVAL_MS);
+          YarnConfiguration.DEFAULT_RM_ZK_RETRY_INTERVAL_MS);
     // Parse authentication from configuration.
     String zkAclConf =
-        conf.get(YarnConfiguration.ZK_RM_STATE_STORE_ACL,
+        conf.get(YarnConfiguration.RM_ZK_ACL,
-            YarnConfiguration.DEFAULT_ZK_RM_STATE_STORE_ACL);
+            YarnConfiguration.DEFAULT_RM_ZK_ACL);
     zkAclConf = ZKUtil.resolveConfIndirection(zkAclConf);
 
     try {
       zkAcl = ZKUtil.parseACLs(zkAclConf);
     } catch (ZKUtil.BadAclFormatException bafe) {
-      LOG.error("Invalid format for " + YarnConfiguration.ZK_RM_STATE_STORE_ACL);
+      LOG.error("Invalid format for " + YarnConfiguration.RM_ZK_ACL);
       throw bafe;
     }
 

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/MockRM.java

@@ -396,6 +396,11 @@ public class MockRM extends ResourceManager {
       protected void stopServer() {
         // don't do anything
       }
+
+      @Override
+      protected EmbeddedElectorService createEmbeddedElectorService() {
+        return null;
+      }
     };
   }
 

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRMHA.java

@@ -25,6 +25,7 @@ import org.apache.hadoop.ha.HAServiceProtocol;
 import org.apache.hadoop.ha.HAServiceProtocol.HAServiceState;
 import org.apache.hadoop.ha.HAServiceProtocol.StateChangeRequestInfo;
 import org.apache.hadoop.ha.HealthCheckFailedException;
+import org.apache.hadoop.security.AccessControlException;
 import org.apache.hadoop.yarn.conf.YarnConfiguration;
 import org.apache.hadoop.yarn.conf.HAUtil;
 import org.junit.Before;
@@ -39,6 +40,7 @@ import static org.junit.Assert.fail;
 
 public class TestRMHA {
   private Log LOG = LogFactory.getLog(TestRMHA.class);
+  private final Configuration configuration = new YarnConfiguration();
   private MockRM rm = null;
   private static final String STATE_ERR =
       "ResourceManager is in wrong HA state";
@@ -51,17 +53,13 @@ public class TestRMHA {
 
   @Before
   public void setUp() throws Exception {
-    Configuration conf = new YarnConfiguration();
+    configuration.setBoolean(YarnConfiguration.RM_HA_ENABLED, true);
-    conf.setBoolean(YarnConfiguration.RM_HA_ENABLED, true);
+    configuration.set(YarnConfiguration.RM_HA_IDS, RM1_NODE_ID + "," + RM2_NODE_ID);
-    conf.set(YarnConfiguration.RM_HA_IDS, RM1_NODE_ID + "," + RM2_NODE_ID);
     for (String confKey : YarnConfiguration.RM_SERVICES_ADDRESS_CONF_KEYS) {
-      conf.set(HAUtil.addSuffix(confKey, RM1_NODE_ID), RM1_ADDRESS);
+      configuration.set(HAUtil.addSuffix(confKey, RM1_NODE_ID), RM1_ADDRESS);
-      conf.set(HAUtil.addSuffix(confKey, RM2_NODE_ID), RM2_ADDRESS);
+      configuration.set(HAUtil.addSuffix(confKey, RM2_NODE_ID), RM2_ADDRESS);
     }
-    conf.set(YarnConfiguration.RM_HA_ID, RM1_NODE_ID);
+    configuration.set(YarnConfiguration.RM_HA_ID, RM1_NODE_ID);
-
-    rm = new MockRM(conf);
-    rm.init(conf);
   }
 
   private void checkMonitorHealth() throws IOException {
@@ -113,6 +111,9 @@ public class TestRMHA {
    */
   @Test (timeout = 30000)
   public void testStartAndTransitions() throws IOException {
+    Configuration conf = new YarnConfiguration(configuration);
+    rm = new MockRM(conf);
+    rm.init(conf);
     StateChangeRequestInfo requestInfo = new StateChangeRequestInfo(
         HAServiceProtocol.RequestSource.REQUEST_BY_USER);
 
@@ -162,4 +163,63 @@ public class TestRMHA {
         rm.areActiveServicesRunning());
     checkMonitorHealth();
   }
+
+  @Test
+  public void testTransitionsWhenAutomaticFailoverEnabled() throws IOException {
+    final String ERR_UNFORCED_REQUEST = "User request succeeded even when " +
+        "automatic failover is enabled";
+
+    Configuration conf = new YarnConfiguration(configuration);
+    conf.setBoolean(YarnConfiguration.AUTO_FAILOVER_ENABLED, true);
+
+    rm = new MockRM(conf);
+    rm.init(conf);
+    rm.start();
+    StateChangeRequestInfo requestInfo = new StateChangeRequestInfo(
+        HAServiceProtocol.RequestSource.REQUEST_BY_USER);
+
+    // Transition to standby
+    try {
+      rm.adminService.transitionToStandby(requestInfo);
+      fail(ERR_UNFORCED_REQUEST);
+    } catch (AccessControlException e) {
+      // expected
+    }
+    checkMonitorHealth();
+    checkStandbyRMFunctionality();
+
+    // Transition to active
+    try {
+      rm.adminService.transitionToActive(requestInfo);
+      fail(ERR_UNFORCED_REQUEST);
+    } catch (AccessControlException e) {
+      // expected
+    }
+    checkMonitorHealth();
+    checkStandbyRMFunctionality();
+
+
+    final String ERR_FORCED_REQUEST = "Forced request by user should work " +
+        "even if automatic failover is enabled";
+    requestInfo = new StateChangeRequestInfo(
+        HAServiceProtocol.RequestSource.REQUEST_BY_USER_FORCED);
+
+    // Transition to standby
+    try {
+      rm.adminService.transitionToStandby(requestInfo);
+    } catch (AccessControlException e) {
+      fail(ERR_FORCED_REQUEST);
+    }
+    checkMonitorHealth();
+    checkStandbyRMFunctionality();
+
+    // Transition to active
+    try {
+      rm.adminService.transitionToActive(requestInfo);
+    } catch (AccessControlException e) {
+      fail(ERR_FORCED_REQUEST);
+    }
+    checkMonitorHealth();
+    checkActiveRMFunctionality();
+  }
 }

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/TestZKRMStateStore.java

@@ -92,7 +92,7 @@ public class TestZKRMStateStore extends RMStateStoreTestBase {
     public RMStateStore getRMStateStore() throws Exception {
       YarnConfiguration conf = new YarnConfiguration();
       workingZnode = "/Test";
-      conf.set(YarnConfiguration.ZK_RM_STATE_STORE_ADDRESS, hostPort);
+      conf.set(YarnConfiguration.RM_ZK_ADDRESS, hostPort);
       conf.set(YarnConfiguration.ZK_RM_STATE_STORE_PARENT_PATH, workingZnode);
       this.client = createClient();
       this.store = new TestZKRMStateStoreInternal(conf, workingZnode);
@@ -140,7 +140,7 @@ public class TestZKRMStateStore extends RMStateStoreTestBase {
     conf.set(YarnConfiguration.RM_HA_IDS, rmIds);
     conf.setBoolean(YarnConfiguration.RECOVERY_ENABLED, true);
     conf.set(YarnConfiguration.RM_STORE, ZKRMStateStore.class.getName());
-    conf.set(YarnConfiguration.ZK_RM_STATE_STORE_ADDRESS, hostPort);
+    conf.set(YarnConfiguration.RM_ZK_ADDRESS, hostPort);
     conf.set(YarnConfiguration.RM_HA_ID, rmId);
     for (String rpcAddress : YarnConfiguration.RM_SERVICES_ADDRESS_CONF_KEYS) {
       for (String id : HAUtil.getRMHAIds(conf)) {

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/TestZKRMStateStoreZKClientConnections.java

@@ -107,7 +107,7 @@ public class TestZKRMStateStoreZKClientConnections extends
 
     public RMStateStore getRMStateStore(Configuration conf) throws Exception {
       String workingZnode = "/Test";
-      conf.set(YarnConfiguration.ZK_RM_STATE_STORE_ADDRESS, hostPort);
+      conf.set(YarnConfiguration.RM_ZK_ADDRESS, hostPort);
       conf.set(YarnConfiguration.ZK_RM_STATE_STORE_PARENT_PATH, workingZnode);
       watcher = new TestForwardingWatcher();
       this.store = new TestZKRMStateStore(conf, workingZnode);
@@ -120,8 +120,8 @@ public class TestZKRMStateStoreZKClientConnections extends
     TestZKClient zkClientTester = new TestZKClient();
     final String path = "/test";
     YarnConfiguration conf = new YarnConfiguration();
-    conf.setInt(YarnConfiguration.ZK_RM_STATE_STORE_TIMEOUT_MS, 1000);
+    conf.setInt(YarnConfiguration.RM_ZK_TIMEOUT_MS, 1000);
-    conf.setLong(YarnConfiguration.ZK_RM_STATE_STORE_RETRY_INTERVAL_MS, 100);
+    conf.setLong(YarnConfiguration.RM_ZK_RETRY_INTERVAL_MS, 100);
     final ZKRMStateStore store =
         (ZKRMStateStore) zkClientTester.getRMStateStore(conf);
     TestDispatcher dispatcher = new TestDispatcher();
@@ -153,7 +153,7 @@ public class TestZKRMStateStoreZKClientConnections extends
     TestZKClient zkClientTester = new TestZKClient();
     String path = "/test";
     YarnConfiguration conf = new YarnConfiguration();
-    conf.setInt(YarnConfiguration.ZK_RM_STATE_STORE_TIMEOUT_MS, 100);
+    conf.setInt(YarnConfiguration.RM_ZK_TIMEOUT_MS, 100);
     ZKRMStateStore store =
         (ZKRMStateStore) zkClientTester.getRMStateStore(conf);
     TestDispatcher dispatcher = new TestDispatcher();
@@ -195,7 +195,7 @@ public class TestZKRMStateStoreZKClientConnections extends
     TestZKClient zkClientTester = new TestZKClient();
     String path = "/test";
     YarnConfiguration conf = new YarnConfiguration();
-    conf.setInt(YarnConfiguration.ZK_RM_STATE_STORE_TIMEOUT_MS, 100);
+    conf.setInt(YarnConfiguration.RM_ZK_TIMEOUT_MS, 100);
     ZKRMStateStore store =
         (ZKRMStateStore) zkClientTester.getRMStateStore(conf);
     TestDispatcher dispatcher = new TestDispatcher();
@@ -227,7 +227,7 @@ public class TestZKRMStateStoreZKClientConnections extends
   public void testSetZKAcl() {
     TestZKClient zkClientTester = new TestZKClient();
     YarnConfiguration conf = new YarnConfiguration();
-    conf.set(YarnConfiguration.ZK_RM_STATE_STORE_ACL, "world:anyone:rwca");
+    conf.set(YarnConfiguration.RM_ZK_ACL, "world:anyone:rwca");
     try {
       zkClientTester.store.zkClient.delete(zkClientTester.store
           .znodeWorkingPath, -1);
@@ -240,7 +240,7 @@ public class TestZKRMStateStoreZKClientConnections extends
   public void testInvalidZKAclConfiguration() {
     TestZKClient zkClientTester = new TestZKClient();
     YarnConfiguration conf = new YarnConfiguration();
-    conf.set(YarnConfiguration.ZK_RM_STATE_STORE_ACL, "randomstring&*");
+    conf.set(YarnConfiguration.RM_ZK_ACL, "randomstring&*");
     try {
       zkClientTester.getRMStateStore(conf);
       fail("ZKRMStateStore created with bad ACL");

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-tests/src/test/java/org/apache/hadoop/yarn/server/MiniYARNCluster.java

@@ -58,6 +58,7 @@ import org.apache.hadoop.yarn.server.nodemanager.NodeHealthCheckerService;
 import org.apache.hadoop.yarn.server.nodemanager.NodeManager;
 import org.apache.hadoop.yarn.server.nodemanager.NodeStatusUpdater;
 import org.apache.hadoop.yarn.server.nodemanager.NodeStatusUpdaterImpl;
+import org.apache.hadoop.yarn.server.resourcemanager.RMFatalEvent;
 import org.apache.hadoop.yarn.server.resourcemanager.ResourceManager;
 import org.apache.hadoop.yarn.server.resourcemanager.ResourceTrackerService;
 import org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptEvent;
@@ -98,6 +99,7 @@ public class MiniYARNCluster extends CompositeService {
 
   private boolean useFixedPorts;
   private boolean useRpc = false;
+  private int failoverTimeout;
 
   private ConcurrentMap<ApplicationAttemptId, Long> appMasters =
       new ConcurrentHashMap<ApplicationAttemptId, Long>(16, 0.75f, 2);
@@ -189,12 +191,15 @@ public class MiniYARNCluster extends CompositeService {
         YarnConfiguration.DEFAULT_YARN_MINICLUSTER_FIXED_PORTS);
     useRpc = conf.getBoolean(YarnConfiguration.YARN_MINICLUSTER_USE_RPC,
         YarnConfiguration.DEFAULT_YARN_MINICLUSTER_USE_RPC);
+    failoverTimeout = conf.getInt(YarnConfiguration.RM_ZK_TIMEOUT_MS,
+        YarnConfiguration.DEFAULT_RM_ZK_TIMEOUT_MS);
 
     if (useRpc && !useFixedPorts) {
       throw new YarnRuntimeException("Invalid configuration!" +
           " Minicluster can use rpc only when configured to use fixed ports");
     }
 
+    conf.setBoolean(YarnConfiguration.IS_MINI_YARN_CLUSTER, true);
     if (resourceManagers.length > 1) {
       conf.setBoolean(YarnConfiguration.RM_HA_ENABLED, true);
       if (conf.get(YarnConfiguration.RM_HA_IDS) == null) {
@@ -218,6 +223,13 @@ public class MiniYARNCluster extends CompositeService {
           // Don't try to login using keytab in the testcases.
         }
       };
+      if (!useFixedPorts) {
+        if (HAUtil.isHAEnabled(conf)) {
+          setHARMConfiguration(i, conf);
+        } else {
+          setNonHARMConfiguration(conf);
+        }
+      }
       addService(new ResourceManagerWrapper(i));
     }
     for(int index = 0; index < nodeManagers.length; index++) {
@@ -230,18 +242,103 @@ public class MiniYARNCluster extends CompositeService {
         conf instanceof YarnConfiguration ? conf : new YarnConfiguration(conf));
   }
 
+  private void setNonHARMConfiguration(Configuration conf) {
+    String hostname = MiniYARNCluster.getHostname();
+    conf.set(YarnConfiguration.RM_ADDRESS, hostname + ":0");
+    conf.set(YarnConfiguration.RM_ADMIN_ADDRESS, hostname + ":0");
+    conf.set(YarnConfiguration.RM_SCHEDULER_ADDRESS, hostname + ":0");
+    conf.set(YarnConfiguration.RM_RESOURCE_TRACKER_ADDRESS, hostname + ":0");
+    WebAppUtils.setRMWebAppHostnameAndPort(conf, hostname, 0);
+  }
+
+  private void setHARMConfiguration(final int index, Configuration conf) {
+    String hostname = MiniYARNCluster.getHostname();
+    for (String confKey : YarnConfiguration.RM_SERVICES_ADDRESS_CONF_KEYS) {
+      conf.set(HAUtil.addSuffix(confKey, rmIds[index]), hostname + ":0");
+    }
+  }
+
+  private synchronized void initResourceManager(int index, Configuration conf) {
+    if (HAUtil.isHAEnabled(conf)) {
+      conf.set(YarnConfiguration.RM_HA_ID, rmIds[index]);
+    }
+    resourceManagers[index].init(conf);
+    resourceManagers[index].getRMContext().getDispatcher().register(
+        RMAppAttemptEventType.class,
+        new EventHandler<RMAppAttemptEvent>() {
+          public void handle(RMAppAttemptEvent event) {
+            if (event instanceof RMAppAttemptRegistrationEvent) {
+              appMasters.put(event.getApplicationAttemptId(),
+                  event.getTimestamp());
+            } else if (event instanceof RMAppAttemptUnregistrationEvent) {
+              appMasters.remove(event.getApplicationAttemptId());
+            }
+          }
+        });
+  }
+
+  private synchronized void startResourceManager(final int index) {
+    try {
+      Thread rmThread = new Thread() {
+        public void run() {
+          resourceManagers[index].start();
+        }
+      };
+      rmThread.setName("RM-" + index);
+      rmThread.start();
+      int waitCount = 0;
+      while (resourceManagers[index].getServiceState() == STATE.INITED
+          && waitCount++ < 60) {
+        LOG.info("Waiting for RM to start...");
+        Thread.sleep(1500);
+      }
+      if (resourceManagers[index].getServiceState() != STATE.STARTED) {
+        // RM could have failed.
+        throw new IOException(
+            "ResourceManager failed to start. Final state is "
+                + resourceManagers[index].getServiceState());
+      }
+    } catch (Throwable t) {
+      throw new YarnRuntimeException(t);
+    }
+    LOG.info("MiniYARN ResourceManager address: " +
+        getConfig().get(YarnConfiguration.RM_ADDRESS));
+    LOG.info("MiniYARN ResourceManager web address: " +
+        WebAppUtils.getRMWebAppURLWithoutScheme(getConfig()));
+  }
+
+  @InterfaceAudience.Private
+  @VisibleForTesting
+  public synchronized void stopResourceManager(int index) {
+    if (resourceManagers[index] != null) {
+      resourceManagers[index].stop();
+      resourceManagers[index] = null;
+    }
+  }
+
+  @InterfaceAudience.Private
+  @VisibleForTesting
+  public synchronized void restartResourceManager(int index)
+      throws InterruptedException {
+    if (resourceManagers[index] != null) {
+      resourceManagers[index].stop();
+      resourceManagers[index] = null;
+    }
+    Configuration conf = getConfig();
+    resourceManagers[index] = new ResourceManager();
+    initResourceManager(index, getConfig());
+    startResourceManager(index);
+  }
+
   public File getTestWorkDir() {
     return testWorkDir;
   }
 
   /**
-   * In a HA cluster, go through all the RMs and find the Active RM. If none
+   * In a HA cluster, go through all the RMs and find the Active RM. In a
-   * of them are active, wait upto 5 seconds for them to transition to Active.
+   * non-HA cluster, return the index of the only RM.
    *
-   * In an non-HA cluster, return the index of the only RM.
+   * @return index of the active RM or -1 if none of them turn active
-   *
-   * @return index of the active RM or -1 if none of them transition to
-   * active even after 5 seconds of waiting
    */
   @InterfaceAudience.Private
   @VisibleForTesting
@@ -250,9 +347,12 @@ public class MiniYARNCluster extends CompositeService {
       return 0;
     }
 
-    int numRetriesForRMBecomingActive = 5;
+    int numRetriesForRMBecomingActive = failoverTimeout / 100;
     while (numRetriesForRMBecomingActive-- > 0) {
       for (int i = 0; i < resourceManagers.length; i++) {
+        if (resourceManagers[i] == null) {
+          continue;
+        }
         try {
           if (HAServiceProtocol.HAServiceState.ACTIVE ==
               resourceManagers[i].getRMContext().getRMAdminService()
@@ -265,7 +365,7 @@ public class MiniYARNCluster extends CompositeService {
         }
       }
       try {
-        Thread.sleep(1000);
+        Thread.sleep(100);
       } catch (InterruptedException e) {
         throw new YarnRuntimeException("Interrupted while waiting for one " +
             "of the ResourceManagers to become active");
@@ -282,7 +382,7 @@ public class MiniYARNCluster extends CompositeService {
     int activeRMIndex = getActiveRMIndex();
     return activeRMIndex == -1
         ? null
-        : this.resourceManagers[getActiveRMIndex()];
+        : this.resourceManagers[activeRMIndex];
   }
 
   public ResourceManager getResourceManager(int i) {
@@ -310,82 +410,21 @@ public class MiniYARNCluster extends CompositeService {
       index = i;
     }
 
-    private void setNonHARMConfiguration(Configuration conf) {
-      String hostname = MiniYARNCluster.getHostname();
-      conf.set(YarnConfiguration.RM_ADDRESS, hostname + ":0");
-      conf.set(YarnConfiguration.RM_ADMIN_ADDRESS, hostname + ":0");
-      conf.set(YarnConfiguration.RM_SCHEDULER_ADDRESS, hostname + ":0");
-      conf.set(YarnConfiguration.RM_RESOURCE_TRACKER_ADDRESS, hostname + ":0");
-      WebAppUtils.setRMWebAppHostnameAndPort(conf, hostname, 0);
-    }
-
-    private void setHARMConfiguration(Configuration conf) {
-      String hostname = MiniYARNCluster.getHostname();
-      for (String confKey : YarnConfiguration.RM_SERVICES_ADDRESS_CONF_KEYS) {
-        for (String id : HAUtil.getRMHAIds(conf)) {
-          conf.set(HAUtil.addSuffix(confKey, id), hostname + ":0");
-        }
-      }
-    }
-
     @Override
     protected synchronized void serviceInit(Configuration conf)
         throws Exception {
-      conf.setBoolean(YarnConfiguration.IS_MINI_YARN_CLUSTER, true);
+      initResourceManager(index, conf);
-
-      if (!useFixedPorts) {
-        if (HAUtil.isHAEnabled(conf)) {
-          setHARMConfiguration(conf);
-        } else {
-          setNonHARMConfiguration(conf);
-        }
-      }
-      if (HAUtil.isHAEnabled(conf)) {
-        conf.set(YarnConfiguration.RM_HA_ID, rmIds[index]);
-      }
-      resourceManagers[index].init(conf);
-      resourceManagers[index].getRMContext().getDispatcher().register
-          (RMAppAttemptEventType.class,
-          new EventHandler<RMAppAttemptEvent>() {
-            public void handle(RMAppAttemptEvent event) {
-              if (event instanceof RMAppAttemptRegistrationEvent) {
-                appMasters.put(event.getApplicationAttemptId(), event.getTimestamp());
-              } else if (event instanceof RMAppAttemptUnregistrationEvent) {
-                appMasters.remove(event.getApplicationAttemptId());
-              }
-            }
-          });
       super.serviceInit(conf);
     }
 
     @Override
     protected synchronized void serviceStart() throws Exception {
-      try {
+      startResourceManager(index);
-        new Thread() {
-          public void run() {
-            resourceManagers[index].start();
-          }
-        }.start();
-        int waitCount = 0;
-        while (resourceManagers[index].getServiceState() == STATE.INITED
-            && waitCount++ < 60) {
-          LOG.info("Waiting for RM to start...");
-          Thread.sleep(1500);
-        }
-        if (resourceManagers[index].getServiceState() != STATE.STARTED) {
-          // RM could have failed.
-          throw new IOException(
-              "ResourceManager failed to start. Final state is "
-                  + resourceManagers[index].getServiceState());
-        }
-        super.serviceStart();
-      } catch (Throwable t) {
-        throw new YarnRuntimeException(t);
-      }
       LOG.info("MiniYARN ResourceManager address: " +
                getConfig().get(YarnConfiguration.RM_ADDRESS));
       LOG.info("MiniYARN ResourceManager web address: " +
                WebAppUtils.getRMWebAppURLWithoutScheme(getConfig()));
+      super.serviceStart();
     }
 
     private void waitForAppMastersToFinish(long timeoutMillis) throws InterruptedException {
@@ -406,7 +445,6 @@ public class MiniYARNCluster extends CompositeService {
         waitForAppMastersToFinish(5000);
         resourceManagers[index].stop();
       }
-      super.serviceStop();
 
       if (Shell.WINDOWS) {
         // On Windows, clean up the short temporary symlink that was created to
@@ -420,6 +458,7 @@ public class MiniYARNCluster extends CompositeService {
             testWorkDir.getAbsolutePath());
         }
       }
+      super.serviceStop();
     }
   }
 

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-web-proxy/src/main/java/org/apache/hadoop/yarn/server/webproxy/AppReportFetcher.java

@@ -19,21 +19,20 @@
 package org.apache.hadoop.yarn.server.webproxy;
 
 import java.io.IOException;
-import java.net.InetSocketAddress;
-
 import org.apache.commons.logging.Log;
 import org.apache.commons.logging.LogFactory;
 import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.ipc.RPC;
 import org.apache.hadoop.yarn.api.ApplicationClientProtocol;
 import org.apache.hadoop.yarn.api.protocolrecords.GetApplicationReportRequest;
 import org.apache.hadoop.yarn.api.protocolrecords.GetApplicationReportResponse;
 import org.apache.hadoop.yarn.api.records.ApplicationId;
 import org.apache.hadoop.yarn.api.records.ApplicationReport;
-import org.apache.hadoop.yarn.conf.YarnConfiguration;
+import org.apache.hadoop.yarn.client.ClientRMProxy;
 import org.apache.hadoop.yarn.exceptions.YarnException;
+import org.apache.hadoop.yarn.exceptions.YarnRuntimeException;
 import org.apache.hadoop.yarn.factories.RecordFactory;
 import org.apache.hadoop.yarn.factory.providers.RecordFactoryProvider;
-import org.apache.hadoop.yarn.ipc.YarnRPC;
 
 /**
  * This class abstracts away how ApplicationReports are fetched.
@@ -50,16 +49,12 @@ public class AppReportFetcher {
    */
   public AppReportFetcher(Configuration conf) {
     this.conf = conf;
-    YarnRPC rpc = YarnRPC.create(this.conf);
+    try {
-    InetSocketAddress rmAddress = conf.getSocketAddr(
+      applicationsManager = ClientRMProxy.createRMProxy(conf,
-            YarnConfiguration.RM_ADDRESS,
+          ApplicationClientProtocol.class);
-            YarnConfiguration.DEFAULT_RM_ADDRESS,
+    } catch (IOException e) {
-            YarnConfiguration.DEFAULT_RM_PORT);
+      throw new YarnRuntimeException(e);
-    LOG.info("Connecting to ResourceManager at " + rmAddress);
+    }
-    applicationsManager =
-        (ApplicationClientProtocol) rpc.getProxy(ApplicationClientProtocol.class,
-            rmAddress, this.conf);
-    LOG.info("Connected to ResourceManager at " + rmAddress);  
   }
   
   /**
@@ -91,4 +86,10 @@ public class AppReportFetcher {
         .getApplicationReport(request);
     return response.getApplicationReport();
   }
+
+  public void stop() {
+    if (this.applicationsManager != null) {
+      RPC.stopProxy(this.applicationsManager);
+    }
+  }
 }

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-web-proxy/src/main/java/org/apache/hadoop/yarn/server/webproxy/WebAppProxy.java

@@ -117,6 +117,9 @@ public class WebAppProxy extends AbstractService {
         throw new YarnRuntimeException("Error stopping proxy web server",e);
       }
     }
+    if(this.fetcher != null) {
+      this.fetcher.stop();
+    }
     super.serviceStop();
   }