HDFS-6995. Block should be placed in the client's 'rack-local' node if 'client-local' node is not available (vinayakumarb)

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

@@ -1054,6 +1054,9 @@ Release 2.6.0 - UNRELEASED
     HDFS-7131. During HA upgrade, JournalNode should create a new committedTxnId
     file in the current directory. (jing9)
 
+    HDFS-6995. Block should be placed in the client's 'rack-local' node
+    if 'client-local' node is not available (vinayakumarb)
+
   BREAKDOWN OF HDFS-6584 ARCHIVAL STORAGE
 
     HDFS-6677. Change INodeFile and FSImage to support storage policy ID.

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

@@ -1493,7 +1493,7 @@ public class BlockManager {
   /** Choose target for getting additional datanodes for an existing pipeline. */
   public DatanodeStorageInfo[] chooseTarget4AdditionalDatanode(String src,
       int numAdditionalNodes,
-      DatanodeDescriptor clientnode,
+      Node clientnode,
       List<DatanodeStorageInfo> chosen,
       Set<Node> excludes,
       long blocksize,
@@ -1513,7 +1513,7 @@ public class BlockManager {
    *      Set, long, List, BlockStoragePolicy)
    */
   public DatanodeStorageInfo[] chooseTarget4NewBlock(final String src,
-      final int numOfReplicas, final DatanodeDescriptor client,
+      final int numOfReplicas, final Node client,
       final Set<Node> excludedNodes,
       final long blocksize,
       final List<String> favoredNodes,

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

@@ -204,11 +204,7 @@ public class BlockPlacementPolicyDefault extends BlockPlacementPolicy {
       // add localMachine and related nodes to excludedNodes
       addToExcludedNodes(storage.getDatanodeDescriptor(), excludedNodes);
     }
-      
-    if (!clusterMap.contains(writer)) {
-      writer = null;
-    }
-      
+
     boolean avoidStaleNodes = (stats != null
         && stats.isAvoidingStaleDataNodesForWrite());
     final Node localNode = chooseTarget(numOfReplicas, writer, excludedNodes,
@@ -219,8 +215,10 @@ public class BlockPlacementPolicyDefault extends BlockPlacementPolicy {
     }
       
     // sorting nodes to form a pipeline
-    return getPipeline((writer==null)?localNode:writer,
-                       results.toArray(new DatanodeStorageInfo[results.size()]));
+    return getPipeline(
+        (writer != null && writer instanceof DatanodeDescriptor) ? writer
+            : localNode,
+        results.toArray(new DatanodeStorageInfo[results.size()]));
   }
 
   private int[] getMaxNodesPerRack(int numOfChosen, int numOfReplicas) {
@@ -271,7 +269,7 @@ public class BlockPlacementPolicyDefault extends BlockPlacementPolicy {
                             final EnumSet<StorageType> unavailableStorages,
                             final boolean newBlock) {
     if (numOfReplicas == 0 || clusterMap.getNumOfLeaves()==0) {
-      return writer;
+      return (writer instanceof DatanodeDescriptor) ? writer : null;
     }
     final int numOfResults = results.size();
     final int totalReplicasExpected = numOfReplicas + numOfResults;

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

@@ -691,8 +691,7 @@ public class DatanodeManager {
       names.add(node.getHostName());
     }
     
-    // resolve its network location
-    List<String> rName = dnsToSwitchMapping.resolve(names);
+    List<String> rName = resolveNetworkLocation(names);
     String networkLocation;
     if (rName == null) {
       LOG.error("The resolve call returned null!");
@@ -704,6 +703,18 @@ public class DatanodeManager {
     return networkLocation;
   }
 
+  /**
+   * Resolve network locations for specified hosts
+   *
+   * @param names
+   * @return Network locations if available, Else returns null
+   */
+  public List<String> resolveNetworkLocation(List<String> names) {
+    // resolve its network location
+    List<String> rName = dnsToSwitchMapping.resolve(names);
+    return rName;
+  }
+
   /**
    * Resolve a node's dependencies in the network. If the DNS to switch 
    * mapping fails then this method returns empty list of dependencies 

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

@@ -266,6 +266,7 @@ import org.apache.hadoop.metrics2.lib.DefaultMetricsSystem;
 import org.apache.hadoop.metrics2.util.MBeans;
 import org.apache.hadoop.net.NetworkTopology;
 import org.apache.hadoop.net.Node;
+import org.apache.hadoop.net.NodeBase;
 import org.apache.hadoop.security.AccessControlException;
 import org.apache.hadoop.security.UserGroupInformation;
 import org.apache.hadoop.security.UserGroupInformation.AuthenticationMethod;
@@ -3122,7 +3123,8 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
     final long blockSize;
     final int replication;
     final byte storagePolicyID;
-    DatanodeDescriptor clientNode = null;
+    Node clientNode = null;
+    String clientMachine = null;
 
     if(NameNode.stateChangeLog.isDebugEnabled()) {
       NameNode.stateChangeLog.debug("BLOCK* NameSystem.getAdditionalBlock: "
@@ -3153,14 +3155,20 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
             + maxBlocksPerFile);
       }
       blockSize = pendingFile.getPreferredBlockSize();
+      clientMachine = pendingFile.getFileUnderConstructionFeature()
+          .getClientMachine();
       clientNode = blockManager.getDatanodeManager().getDatanodeByHost(
-              pendingFile.getFileUnderConstructionFeature().getClientMachine());
+          clientMachine);
       replication = pendingFile.getFileReplication();
       storagePolicyID = pendingFile.getStoragePolicyID();
     } finally {
       readUnlock();
     }
 
+    if (clientNode == null) {
+      clientNode = getClientNode(clientMachine);
+    }
+
     // choose targets for the new block to be allocated.
     final DatanodeStorageInfo targets[] = getBlockManager().chooseTarget4NewBlock( 
         src, replication, clientNode, excludedNodes, blockSize, favoredNodes,
@@ -3217,6 +3225,24 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
     return makeLocatedBlock(newBlock, targets, offset);
   }
 
+  /*
+   * Resolve clientmachine address to get a network location path
+   */
+  private Node getClientNode(String clientMachine) {
+    List<String> hosts = new ArrayList<String>(1);
+    hosts.add(clientMachine);
+    List<String> rName = getBlockManager().getDatanodeManager()
+        .resolveNetworkLocation(hosts);
+    Node clientNode = null;
+    if (rName != null) {
+      // Able to resolve clientMachine mapping.
+      // Create a temp node to findout the rack local nodes
+      clientNode = new NodeBase(rName.get(0) + NodeBase.PATH_SEPARATOR_STR
+          + clientMachine);
+    }
+    return clientNode;
+  }
+
   static class FileState {
     public final INodeFile inode;
     public final String path;
@@ -3350,7 +3376,8 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
     //check if the feature is enabled
     dtpReplaceDatanodeOnFailure.checkEnabled();
 
-    final DatanodeDescriptor clientnode;
+    Node clientnode = null;
+    String clientMachine;
     final long preferredblocksize;
     final byte storagePolicyID;
     final List<DatanodeStorageInfo> chosen;
@@ -3375,8 +3402,7 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
         if (inode != null) src = inode.getFullPathName();
       }
       final INodeFile file = checkLease(src, clientName, inode, fileId);
-      String clientMachine = file.getFileUnderConstructionFeature()
-              .getClientMachine();
+      clientMachine = file.getFileUnderConstructionFeature().getClientMachine();
       clientnode = blockManager.getDatanodeManager().getDatanodeByHost(clientMachine);
       preferredblocksize = file.getPreferredBlockSize();
       storagePolicyID = file.getStoragePolicyID();
@@ -3388,6 +3414,10 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
       readUnlock();
     }
 
+    if (clientnode == null) {
+      clientnode = getClientNode(clientMachine);
+    }
+
     // choose new datanodes.
     final DatanodeStorageInfo[] targets = blockManager.chooseTarget4AdditionalDatanode(
         src, numAdditionalNodes, clientnode, chosen, 

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/blockmanagement/TestReplicationPolicyConsiderLoad.java

@@ -34,7 +34,6 @@ import org.apache.hadoop.hdfs.DFSConfigKeys;
 import org.apache.hadoop.hdfs.DFSTestUtil;
 import org.apache.hadoop.hdfs.HdfsConfiguration;
 import org.apache.hadoop.hdfs.TestBlockStoragePolicy;
-import org.apache.hadoop.hdfs.protocol.DatanodeInfo;
 import org.apache.hadoop.hdfs.protocol.HdfsConstants;
 import org.apache.hadoop.hdfs.security.token.block.ExportedBlockKeys;
 import org.apache.hadoop.hdfs.server.common.HdfsServerConstants.NodeType;
@@ -142,10 +141,18 @@ public class TestReplicationPolicyConsiderLoad {
       }
       assertEquals((double)load/3, fsn.getInServiceXceiverAverage(), EPSILON);
 
+      // update references of writer DN to update the de-commissioned state
+      List<DatanodeDescriptor> liveNodes = new ArrayList<DatanodeDescriptor>();
+      dnManager.fetchDatanodes(liveNodes, null, false);
+      DatanodeDescriptor writerDn = null;
+      if (liveNodes.contains(dataNodes[0])) {
+        writerDn = liveNodes.get(liveNodes.indexOf(dataNodes[0]));
+      }
+
       // Call chooseTarget()
       DatanodeStorageInfo[] targets = namenode.getNamesystem().getBlockManager()
           .getBlockPlacementPolicy().chooseTarget("testFile.txt", 3,
-              dataNodes[0], new ArrayList<DatanodeStorageInfo>(), false, null,
+              writerDn, new ArrayList<DatanodeStorageInfo>(), false, null,
               1024, TestBlockStoragePolicy.DEFAULT_STORAGE_POLICY);
 
       assertEquals(3, targets.length);