HDFS-2822. processMisReplicatedBlock incorrectly identifies under-construction blocks as under-replicated. Contributed by Todd Lipcon.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/branches/branch-0.23@1234926 13f79535-47bb-0310-9956-ffa450edef68

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

@@ -171,6 +171,9 @@ Release 0.23.1 - UNRELEASED
     HDFS-2816. Fix missing license header in httpfs findbugsExcludeFile.xml.
     (hitesh via tucu)
 
+    HDFS-2822. processMisReplicatedBlock incorrectly identifies
+    under-construction blocks as under-replicated. (todd)
+
 Release 0.23.0 - 2011-11-01 
 
   INCOMPATIBLE CHANGES

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

@@ -1791,7 +1791,8 @@ public class BlockManager {
   public void processMisReplicatedBlocks() {
     assert namesystem.hasWriteLock();
 
-    long nrInvalid = 0, nrOverReplicated = 0, nrUnderReplicated = 0;
+    long nrInvalid = 0, nrOverReplicated = 0, nrUnderReplicated = 0,
+         nrUnderConstruction = 0;
     neededReplications.clear();
     for (BlockInfo block : blocksMap.getBlocks()) {
       INodeFile fileINode = block.getINode();
@@ -1801,6 +1802,12 @@ public class BlockManager {
         addToInvalidates(block);
         continue;
       }
+      if (!block.isComplete()) {
+        // Incomplete blocks are never considered mis-replicated --
+        // they'll be reached when they are completed or recovered.
+        nrUnderConstruction++;
+        continue;
+      }
       // calculate current replication
       short expectedReplication = fileINode.getReplication();
       NumberReplicas num = countNodes(block);
@@ -1824,6 +1831,7 @@ public class BlockManager {
     LOG.info("Number of invalid blocks          = " + nrInvalid);
     LOG.info("Number of under-replicated blocks = " + nrUnderReplicated);
     LOG.info("Number of  over-replicated blocks = " + nrOverReplicated);
+    LOG.info("Number of blocks being written    = " + nrUnderConstruction);
   }
 
   /** Set replication for the blocks. */

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestSafeMode.java

@@ -19,22 +19,30 @@
 package org.apache.hadoop.hdfs;
 
 import java.io.IOException;
+import java.util.List;
 
 import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.fs.FSDataOutputStream;
 import org.apache.hadoop.fs.Path;
 import org.apache.hadoop.fs.FileSystem;
 import org.apache.hadoop.fs.permission.FsPermission;
 import org.apache.hadoop.hdfs.protocol.HdfsConstants.SafeModeAction;
+import org.apache.hadoop.hdfs.server.blockmanagement.BlockManagerTestUtil;
+import org.apache.hadoop.hdfs.server.namenode.FSNamesystem;
+import org.apache.hadoop.io.IOUtils;
 
 import static org.junit.Assert.*;
 import org.junit.Before;
 import org.junit.After;
 import org.junit.Test;
 
+import com.google.common.collect.Lists;
+
 /**
  * Tests to verify safe mode correctness.
  */
 public class TestSafeMode {
+  private static final int BLOCK_SIZE = 1024;
   Configuration conf; 
   MiniDFSCluster cluster;
   FileSystem fs;
@@ -43,6 +51,7 @@ public class TestSafeMode {
   @Before
   public void startUp() throws IOException {
     conf = new HdfsConfiguration();
+    conf.setInt(DFSConfigKeys.DFS_BLOCK_SIZE_KEY, BLOCK_SIZE);
     cluster = new MiniDFSCluster.Builder(conf).numDataNodes(1).build();
     cluster.waitActive();      
     fs = cluster.getFileSystem();
@@ -128,6 +137,46 @@ public class TestSafeMode {
     assertEquals("", status);
   }
 
+  /**
+   * Test that, when under-replicated blocks are processed at the end of
+   * safe-mode, blocks currently under construction are not considered
+   * under-construction or missing. Regression test for HDFS-2822.
+   */
+  @Test
+  public void testRbwBlocksNotConsideredUnderReplicated() throws IOException {
+    List<FSDataOutputStream> stms = Lists.newArrayList();
+    try {
+      // Create some junk blocks so that the NN doesn't just immediately
+      // exit safemode on restart.
+      DFSTestUtil.createFile(fs, new Path("/junk-blocks"),
+          BLOCK_SIZE*4, (short)1, 1L);
+      // Create several files which are left open. It's important to
+      // create several here, because otherwise the first iteration of the
+      // replication monitor will pull them off the replication queue and
+      // hide this bug from the test!
+      for (int i = 0; i < 10; i++) {
+        FSDataOutputStream stm = fs.create(
+            new Path("/append-" + i), true, BLOCK_SIZE, (short) 1, BLOCK_SIZE);
+        stms.add(stm);
+        stm.write(1);
+        stm.hflush();
+      }
+
+      cluster.restartNameNode();
+      FSNamesystem ns = cluster.getNameNode(0).getNamesystem();
+      BlockManagerTestUtil.updateState(ns.getBlockManager());
+      assertEquals(0, ns.getPendingReplicationBlocks());
+      assertEquals(0, ns.getCorruptReplicaBlocks());
+      assertEquals(0, ns.getMissingBlocksCount());
+
+    } finally {
+      for (FSDataOutputStream stm : stms) {
+        IOUtils.closeStream(stm);
+      }
+      cluster.shutdown();
+    }
+  }
+
   public interface FSRun {
     public abstract void run(FileSystem fs) throws IOException;
   }