HDFS-4288. NN accepts incremental BR as IBR in safemode. contributed by Daryn Sharp.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1440192 13f79535-47bb-0310-9956-ffa450edef68

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

@@ -2233,6 +2233,8 @@ Release 0.23.7 - UNRELEASED
 
   BUG FIXES
 
+    HDFS-4288. NN accepts incremental BR as IBR in safemode (daryn via kihwal)
+
 Release 0.23.6 - UNRELEASED
 
   INCOMPATIBLE CHANGES

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

@@ -88,11 +88,7 @@ public class BlockInfo extends Block implements LightWeightGSet.LinkedElement {
   DatanodeDescriptor getDatanode(int index) {
     assert this.triplets != null : "BlockInfo is not initialized";
     assert index >= 0 && index*3 < triplets.length : "Index is out of bound";
-    DatanodeDescriptor node = (DatanodeDescriptor)triplets[index*3];
-    assert node == null || 
-        DatanodeDescriptor.class.getName().equals(node.getClass().getName()) : 
-              "DatanodeDescriptor is expected at " + index*3;
-    return node;
+    return (DatanodeDescriptor)triplets[index*3];
   }
 
   private BlockInfo getPrevious(int index) {

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

@@ -62,6 +62,7 @@ import org.apache.hadoop.hdfs.server.common.HdfsServerConstants.ReplicaState;
 import org.apache.hadoop.hdfs.server.namenode.FSClusterStats;
 import org.apache.hadoop.hdfs.server.namenode.NameNode;
 import org.apache.hadoop.hdfs.server.namenode.Namesystem;
+import org.apache.hadoop.hdfs.server.namenode.metrics.NameNodeMetrics;
 import org.apache.hadoop.hdfs.server.protocol.BlockCommand;
 import org.apache.hadoop.hdfs.server.protocol.BlocksWithLocations;
 import org.apache.hadoop.hdfs.server.protocol.BlocksWithLocations.BlockWithLocations;
@@ -1576,7 +1577,10 @@ public class BlockManager {
     }
 
     // Log the block report processing stats from Namenode perspective
-    NameNode.getNameNodeMetrics().addBlockReport((int) (endTime - startTime));
+    final NameNodeMetrics metrics = NameNode.getNameNodeMetrics();
+    if (metrics != null) {
+      metrics.addBlockReport((int) (endTime - startTime));
+    }
     blockLog.info("BLOCK* processReport: from "
         + nodeID + ", blocks: " + newReport.getNumberOfBlocks()
         + ", processing time: " + (endTime - startTime) + " msecs");

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

@@ -547,6 +547,7 @@ public class DatanodeDescriptor extends DatanodeInfo {
   @Override
   public void updateRegInfo(DatanodeID nodeReg) {
     super.updateRegInfo(nodeReg);
+    firstBlockReport = true; // must re-process IBR after re-registration
   }
 
   /**

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

@@ -419,7 +419,7 @@ public class DatanodeManager {
   }
 
   /** Add a datanode. */
-  private void addDatanode(final DatanodeDescriptor node) {
+  void addDatanode(final DatanodeDescriptor node) {
     // To keep host2DatanodeMap consistent with datanodeMap,
     // remove  from host2DatanodeMap the datanodeDescriptor removed
     // from datanodeMap before adding node to host2DatanodeMap.

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

@@ -34,13 +34,16 @@ import org.apache.hadoop.hdfs.DFSConfigKeys;
 import org.apache.hadoop.hdfs.DFSTestUtil;
 import org.apache.hadoop.hdfs.HdfsConfiguration;
 import org.apache.hadoop.hdfs.protocol.Block;
+import org.apache.hadoop.hdfs.protocol.BlockListAsLongs;
 import org.apache.hadoop.hdfs.protocol.HdfsConstants;
 import org.apache.hadoop.hdfs.server.blockmanagement.DatanodeDescriptor.BlockTargetPair;
 import org.apache.hadoop.hdfs.server.namenode.FSNamesystem;
+import org.apache.hadoop.hdfs.server.protocol.DatanodeRegistration;
 import org.apache.hadoop.net.NetworkTopology;
 import org.junit.Before;
 import org.junit.Test;
 import org.mockito.Mockito;
+import static org.mockito.Mockito.*;
 
 import com.google.common.base.Joiner;
 import com.google.common.collect.ImmutableList;
@@ -485,4 +488,70 @@ public class TestBlockManager {
             new NumberReplicas(),
             UnderReplicatedBlocks.QUEUE_HIGHEST_PRIORITY));
   }
+
+  @Test
+  public void testSafeModeIBR() throws Exception {
+    DatanodeDescriptor node = spy(nodes.get(0));
+    node.setStorageID("dummy-storage");
+    node.isAlive = true;
+
+    DatanodeRegistration nodeReg =
+        new DatanodeRegistration(node, null, null, "");
+
+    // pretend to be in safemode
+    doReturn(true).when(fsn).isInStartupSafeMode();
+    
+    // register new node
+    bm.getDatanodeManager().registerDatanode(nodeReg);
+    bm.getDatanodeManager().addDatanode(node); // swap in spy    
+    assertEquals(node, bm.getDatanodeManager().getDatanode(node));
+    assertTrue(node.isFirstBlockReport());
+    // send block report, should be processed
+    reset(node);
+    bm.processReport(node, "pool", new BlockListAsLongs(null, null));
+    verify(node).receivedBlockReport();
+    assertFalse(node.isFirstBlockReport());
+    // send block report again, should NOT be processed
+    reset(node);
+    bm.processReport(node, "pool", new BlockListAsLongs(null, null));
+    verify(node, never()).receivedBlockReport();
+    assertFalse(node.isFirstBlockReport());
+
+    // re-register as if node restarted, should update existing node
+    bm.getDatanodeManager().removeDatanode(node);
+    reset(node);
+    bm.getDatanodeManager().registerDatanode(nodeReg);
+    verify(node).updateRegInfo(nodeReg);
+    assertTrue(node.isFirstBlockReport()); // ready for report again
+    // send block report, should be processed after restart
+    reset(node);
+    bm.processReport(node, "pool", new BlockListAsLongs(null, null));
+    verify(node).receivedBlockReport();
+    assertFalse(node.isFirstBlockReport());
+  }
+  
+  @Test
+  public void testSafeModeIBRAfterIncremental() throws Exception {
+    DatanodeDescriptor node = spy(nodes.get(0));
+    node.setStorageID("dummy-storage");
+    node.isAlive = true;
+
+    DatanodeRegistration nodeReg =
+        new DatanodeRegistration(node, null, null, "");
+
+    // pretend to be in safemode
+    doReturn(true).when(fsn).isInStartupSafeMode();
+
+    // register new node
+    bm.getDatanodeManager().registerDatanode(nodeReg);
+    bm.getDatanodeManager().addDatanode(node); // swap in spy    
+    assertEquals(node, bm.getDatanodeManager().getDatanode(node));
+    assertTrue(node.isFirstBlockReport());
+    // send block report while pretending to already have blocks
+    reset(node);
+    doReturn(1).when(node).numBlocks();
+    bm.processReport(node, "pool", new BlockListAsLongs(null, null));
+    verify(node).receivedBlockReport();
+    assertFalse(node.isFirstBlockReport());
+  }
 }