svn merge -c 1406278 FIXES: HDFS-4075. Reduce recommissioning overhead (Kihwal Lee via daryn)

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/branches/branch-0.23@1406290 13f79535-47bb-0310-9956-ffa450edef68

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

@@ -16,6 +16,8 @@ Release 0.23.5 - UNRELEASED
     HDFS-3483. Better error message when hdfs fsck is run against a ViewFS
     config. (Stephen Fritz via atm)
 
+    HDFS-4075. Reduce recommissioning overhead (Kihwal Lee via daryn)
+
   OPTIMIZATIONS
 
   BUG FIXES

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

@@ -2311,6 +2311,7 @@ public class BlockManager {
   void processOverReplicatedBlocksOnReCommission(
       final DatanodeDescriptor srcNode) {
     final Iterator<? extends Block> it = srcNode.getBlockIterator();
+    int numOverReplicated = 0;
     while(it.hasNext()) {
       final Block block = it.next();
       INodeFile fileINode = blocksMap.getINode(block);
@@ -2320,8 +2321,11 @@ public class BlockManager {
       if (numCurrentReplica > expectedReplication) {
         // over-replicated block 
         processOverReplicatedBlock(block, expectedReplication, null, null);
+        numOverReplicated++;
       }
     }
+    LOG.info("Invalidated " + numOverReplicated + " over-replicated blocks on " +
+        srcNode + " during recommissioning");
   }
 
   /**

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

@@ -522,7 +522,11 @@ public class DatanodeManager {
     if (node.isDecommissionInProgress() || node.isDecommissioned()) {
       LOG.info("Stop Decommissioning node " + node.getName());
       heartbeatManager.stopDecommission(node);
-      blockManager.processOverReplicatedBlocksOnReCommission(node);
+      // Over-replicated blocks will be detected and processed when 
+      // the dead node comes back and send in its full block report.
+      if (node.isAlive) {
+        blockManager.processOverReplicatedBlocksOnReCommission(node);
+      }
     }
   }