HDFS-2988. Improve error message when storage directory lock fails. Contributed by Miomir Boljanovic. (harsh)

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1358683 13f79535-47bb-0310-9956-ffa450edef68

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

@@ -276,6 +276,9 @@ Branch-2 ( Unreleased changes )
 
     HDFS-3604. Add dfs.webhdfs.enabled to hdfs-default.xml. (eli)
 
+    HDFS-2988. Improve error message when storage directory lock fails
+    (Miomir Boljanovic via harsh)
+
   OPTIMIZATIONS
 
     HDFS-2982. Startup performance suffers when there are many edit log

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/common/Storage.java

@@ -22,6 +22,7 @@
 import java.io.FileOutputStream;
 import java.io.IOException;
 import java.io.RandomAccessFile;
+import java.lang.management.ManagementFactory;
 import java.nio.channels.FileLock;
 import java.nio.channels.OverlappingFileLockException;
 import java.util.ArrayList;
@@ -615,14 +616,20 @@ FileLock tryLock() throws IOException {
         deletionHookAdded = true;
       }
       RandomAccessFile file = new RandomAccessFile(lockF, "rws");
+      String jvmName = ManagementFactory.getRuntimeMXBean().getName();
       FileLock res = null;
       try {
         res = file.getChannel().tryLock();
+        file.write(jvmName.getBytes());
+        LOG.info("Lock on " + lockF + " acquired by nodename " + jvmName);
       } catch(OverlappingFileLockException oe) {
+        LOG.error("It appears that another namenode " + file.readLine() 
+            + " has already locked the storage directory");
         file.close();
         return null;
       } catch(IOException e) {
-        LOG.error("Cannot create lock on " + lockF, e);
+        LOG.error("Failed to acquire lock on " + lockF + ". If this storage directory is mounted via NFS, " 
+            + "ensure that the appropriate nfs lock services are running.", e);
         file.close();
         throw e;
       }

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

@@ -19,6 +19,8 @@
 
 import static org.apache.hadoop.hdfs.server.common.Util.fileAsURI;
 import junit.framework.TestCase;
+
+import java.lang.management.ManagementFactory;
 import java.net.InetSocketAddress;
 import java.io.File;
 import java.io.IOException;
@@ -48,6 +50,7 @@
 import org.apache.hadoop.hdfs.MiniDFSNNTopology;
 import org.apache.hadoop.hdfs.protocol.HdfsConstants.SafeModeAction;
 import org.apache.hadoop.hdfs.server.common.HdfsServerConstants.StartupOption;
+import org.apache.hadoop.hdfs.server.common.Storage;
 import org.apache.hadoop.hdfs.server.common.Storage.StorageDirectory;
 import org.apache.hadoop.hdfs.server.common.StorageInfo;
 import org.apache.hadoop.hdfs.server.namenode.NNStorage.NameNodeDirType;
@@ -60,6 +63,7 @@
 import org.apache.hadoop.net.NetUtils;
 import org.apache.hadoop.test.GenericTestUtils;
 import org.apache.hadoop.test.GenericTestUtils.DelayAnswer;
+import org.apache.hadoop.test.GenericTestUtils.LogCapturer;
 import org.apache.hadoop.util.StringUtils;
 import org.apache.log4j.Level;
 import org.mockito.ArgumentMatcher;
@@ -679,6 +683,38 @@ public void testSecondaryNameNodeLocking() throws Exception {
     }
   }
   
+  /**
+   * Test that, an attempt to lock a storage that is already locked by a nodename,
+   * logs error message that includes JVM name of the namenode that locked it.
+   */
+  public void testStorageAlreadyLockedErrorMessage() throws Exception {
+    Configuration conf = new HdfsConfiguration();
+    MiniDFSCluster cluster = new MiniDFSCluster.Builder(conf)
+      .numDataNodes(0)
+      .build();
+    
+    StorageDirectory savedSd = null;
+    try {
+      NNStorage storage = cluster.getNameNode().getFSImage().getStorage();
+      for (StorageDirectory sd : storage.dirIterable(null)) {
+        assertLockFails(sd);
+        savedSd = sd;
+      }
+      
+      LogCapturer logs = GenericTestUtils.LogCapturer.captureLogs(LogFactory.getLog(Storage.class));
+      try {
+        // try to lock the storage that's already locked
+        savedSd.lock();
+        fail("Namenode should not be able to lock a storage that is already locked");
+      } catch (IOException ioe) {
+        String jvmName = ManagementFactory.getRuntimeMXBean().getName();
+        assertTrue("Error message does not include JVM name '" + jvmName 
+            + "'", logs.getOutput().contains(jvmName));
+      }
+    } finally {
+      cluster.shutdown();
+    }
+  }
 
   /**
    * Assert that the given storage directory can't be locked, because