HDFS-9204. DatanodeDescriptor#PendingReplicationWithoutTargets is wrongly calculated. Contributed by Mingliang Liu.

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

@@ -356,6 +356,9 @@ Trunk (Unreleased)
 
     HDFS-8800. hdfs --daemon stop namenode corrupts logs (John Smith via aw)
 
+    HDFS-9204. DatanodeDescriptor#PendingReplicationWithoutTargets is wrongly
+    calculated. (Mingliang Liu via jing9)
+
     BREAKDOWN OF HDFS-7285 SUBTASKS AND RELATED JIRAS
 
       HDFS-7347. Configurable erasure coding policy for individual files and

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

@@ -256,7 +256,7 @@ public class DatanodeDescriptor extends DatanodeInfo {
   private boolean disallowed = false;
 
   // The number of replication work pending before targets are determined
-  private int PendingReplicationWithoutTargets = 0;
+  private int pendingReplicationWithoutTargets = 0;
 
   // HB processing can use it to tell if it is the first HB since DN restarted
   private boolean heartbeatedSinceRegistration = false;
@@ -594,11 +594,11 @@ public class DatanodeDescriptor extends DatanodeInfo {
   }
 
   void incrementPendingReplicationWithoutTargets() {
-    PendingReplicationWithoutTargets++;
+    pendingReplicationWithoutTargets++;
   }
 
   void decrementPendingReplicationWithoutTargets() {
-    PendingReplicationWithoutTargets--;
+    pendingReplicationWithoutTargets--;
   }
 
   /**
@@ -651,7 +651,7 @@ public class DatanodeDescriptor extends DatanodeInfo {
    * The number of work items that are pending to be replicated
    */
   int getNumberOfBlocksToBeReplicated() {
-    return PendingReplicationWithoutTargets + replicateBlocks.size();
+    return pendingReplicationWithoutTargets + replicateBlocks.size();
   }
 
   /**

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

@@ -30,6 +30,9 @@ class ReplicationWork extends BlockRecoveryWork {
       int priority) {
     super(block, bc, srcNodes, containingNodes,
         liveReplicaStorages, additionalReplRequired, priority);
+    assert getSrcNodes().length == 1 :
+        "There should be exactly 1 source node that have been selected";
+    getSrcNodes()[0].incrementPendingReplicationWithoutTargets();
     BlockManager.LOG.debug("Creating a ReplicationWork to recover " + block);
   }
 

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

@@ -31,6 +31,7 @@ import org.apache.hadoop.hdfs.MiniDFSCluster;
 import org.apache.hadoop.hdfs.protocol.ExtendedBlock;
 import org.apache.hadoop.hdfs.server.datanode.DataNodeTestUtils;
 import org.junit.Test;
+import org.mockito.internal.util.reflection.Whitebox;
 
 import java.util.Iterator;
 
@@ -135,6 +136,10 @@ public class TestUnderReplicatedBlocks {
       assertEquals(NUM_OF_BLOCKS, bm.getUnderReplicatedNotMissingBlocks());
       bm.computeDatanodeWork();
 
+      assertTrue("The number of replication work pending before targets are " +
+              "determined should be non-negative.",
+          (Integer)Whitebox.getInternalState(secondDn,
+              "pendingReplicationWithoutTargets") >= 0);
 
       assertTrue("The number of blocks to be replicated should be less than "
           + "or equal to " + bm.replicationStreamsHardLimit,