HDFS-5723. Append failed FINALIZED replica should not be accepted as valid when that block is underconstruction. Contributed by Vinayakumar B.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1615491 13f79535-47bb-0310-9956-ffa450edef68

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

@@ -429,6 +429,10 @@ Release 2.6.0 - UNRELEASED
 
     HDFS-6810. StorageReport array is initialized with wrong size in
     DatanodeDescriptor#getStorageReports. (szetszwo via Arpit Agarwal)
+
+    HDFS-5723. Append failed FINALIZED replica should not be accepted as valid
+    when that block is underconstruction (vinayakumarb)
+
 Release 2.5.0 - UNRELEASED
 
   INCOMPATIBLE CHANGES

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

@@ -2167,6 +2167,16 @@ public class BlockManager {
         } else {
           return null; // not corrupt
         }
+      case UNDER_CONSTRUCTION:
+        if (storedBlock.getGenerationStamp() > reported.getGenerationStamp()) {
+          final long reportedGS = reported.getGenerationStamp();
+          return new BlockToMarkCorrupt(storedBlock, reportedGS, "block is "
+              + ucState + " and reported state " + reportedState
+              + ", But reported genstamp " + reportedGS
+              + " does not match genstamp in block map "
+              + storedBlock.getGenerationStamp(), Reason.GENSTAMP_MISMATCH);
+        }
+        return null;
       default:
         return null;
       }

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

@@ -19,6 +19,7 @@ package org.apache.hadoop.hdfs;
 
 import static org.junit.Assert.assertEquals;
 import static org.junit.Assert.assertTrue;
+import static org.junit.Assert.fail;
 
 import java.io.File;
 import java.io.FileNotFoundException;
@@ -32,6 +33,7 @@ import org.apache.hadoop.fs.FSDataOutputStream;
 import org.apache.hadoop.fs.FileSystem;
 import org.apache.hadoop.fs.HardLink;
 import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.hdfs.MiniDFSCluster.DataNodeProperties;
 import org.apache.hadoop.hdfs.protocol.AlreadyBeingCreatedException;
 import org.apache.hadoop.hdfs.protocol.ExtendedBlock;
 import org.apache.hadoop.hdfs.protocol.LocatedBlock;
@@ -39,6 +41,7 @@ import org.apache.hadoop.hdfs.protocol.LocatedBlocks;
 import org.apache.hadoop.hdfs.server.datanode.DataNode;
 import org.apache.hadoop.hdfs.server.datanode.DataNodeTestUtils;
 import org.apache.hadoop.hdfs.server.datanode.SimulatedFSDataset;
+import org.apache.hadoop.io.IOUtils;
 import org.apache.hadoop.ipc.RemoteException;
 import org.junit.Assert;
 import org.junit.Test;
@@ -169,6 +172,7 @@ public class TestFileAppend{
       }
 
     } finally {
+      client.close();
       fs.close();
       cluster.shutdown();
     }
@@ -380,4 +384,57 @@ public class TestFileAppend{
     }
   }
 
+  /**
+   * Old replica of the block should not be accepted as valid for append/read
+   */
+  @Test
+  public void testFailedAppendBlockRejection() throws Exception {
+    Configuration conf = new HdfsConfiguration();
+    conf.set("dfs.client.block.write.replace-datanode-on-failure.enable",
+        "false");
+    MiniDFSCluster cluster = new MiniDFSCluster.Builder(conf).numDataNodes(3)
+        .build();
+    DistributedFileSystem fs = null;
+    try {
+      fs = cluster.getFileSystem();
+      Path path = new Path("/test");
+      FSDataOutputStream out = fs.create(path);
+      out.writeBytes("hello\n");
+      out.close();
+
+      // stop one datanode
+      DataNodeProperties dnProp = cluster.stopDataNode(0);
+      String dnAddress = dnProp.datanode.getXferAddress().toString();
+      if (dnAddress.startsWith("/")) {
+        dnAddress = dnAddress.substring(1);
+      }
+
+      // append again to bump genstamps
+      for (int i = 0; i < 2; i++) {
+        out = fs.append(path);
+        out.writeBytes("helloagain\n");
+        out.close();
+      }
+
+      // re-open and make the block state as underconstruction
+      out = fs.append(path);
+      cluster.restartDataNode(dnProp, true);
+      // wait till the block report comes
+      Thread.sleep(2000);
+      // check the block locations, this should not contain restarted datanode
+      BlockLocation[] locations = fs.getFileBlockLocations(path, 0,
+          Long.MAX_VALUE);
+      String[] names = locations[0].getNames();
+      for (String node : names) {
+        if (node.equals(dnAddress)) {
+          fail("Failed append should not be present in latest block locations.");
+        }
+      }
+      out.close();
+    } finally {
+      IOUtils.closeStream(fs);
+      cluster.shutdown();
+    }
+  }
+
 }

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/ha/TestRetryCacheWithHA.java

@@ -50,6 +50,7 @@ import org.apache.hadoop.fs.XAttrSetFlag;
 import org.apache.hadoop.fs.permission.FsPermission;
 import org.apache.hadoop.hdfs.DFSClient;
 import org.apache.hadoop.hdfs.DFSConfigKeys;
+import org.apache.hadoop.hdfs.DFSOutputStream;
 import org.apache.hadoop.hdfs.DFSTestUtil;
 import org.apache.hadoop.hdfs.DistributedFileSystem;
 import org.apache.hadoop.hdfs.HdfsConfiguration;
@@ -725,7 +726,12 @@ public class TestRetryCacheWithHA {
       
       client.getNamenode().updatePipeline(client.getClientName(), oldBlock,
           newBlock, newNodes, storageIDs);
-      out.close();
+      // close can fail if the out.close() commit the block after block received
+      // notifications from Datanode.
+      // Since datanodes and output stream have still old genstamps, these
+      // blocks will be marked as corrupt after HDFS-5723 if RECEIVED
+      // notifications reaches namenode first and close() will fail.
+      DFSTestUtil.abortStream((DFSOutputStream) out.getWrappedStream());
     }
 
     @Override