Mailing-List: contact core-commits-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Reply-To: core-dev@hadoop.apache.org
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Subject: svn commit: r752593 - in /hadoop/core/branches/branch-0.19:
 CHANGES.txt
 src/hdfs/org/apache/hadoop/hdfs/DFSClient.java
 src/test/org/apache/hadoop/hdfs/TestFileCreation.java
Date: Wed, 11 Mar 2009 19:43:49 -0000
To: core-commits@hadoop.apache.org
From: dhruba@apache.org
Message-Id: <20090311194349.80FE12388995@eris.apache.org>

Author: dhruba
Date: Wed Mar 11 19:43:48 2009
New Revision: 752593

URL: http://svn.apache.org/viewvc?rev=752593&view=rev
Log:
HADOOP-3998. Fix dfsclient exception when JVM is shutdown. (dhruba)


Modified:
    hadoop/core/branches/branch-0.19/CHANGES.txt   (contents, props changed)
    hadoop/core/branches/branch-0.19/src/hdfs/org/apache/hadoop/hdfs/DFSClient.java
    hadoop/core/branches/branch-0.19/src/test/org/apache/hadoop/hdfs/TestFileCreation.java

Modified: hadoop/core/branches/branch-0.19/CHANGES.txt
URL: http://svn.apache.org/viewvc/hadoop/core/branches/branch-0.19/CHANGES.txt?rev=752593&r1=752592&r2=752593&view=diff
==============================================================================
--- hadoop/core/branches/branch-0.19/CHANGES.txt (original)
+++ hadoop/core/branches/branch-0.19/CHANGES.txt Wed Mar 11 19:43:48 2009
@@ -60,6 +60,8 @@
 
     HADOOP-5333. libhdfs supports appending to files. (dhruba)
 
+    HADOOP-3998. Fix dfsclient exception when JVM is shutdown. (dhruba)
+ 
 Release 0.19.1 - 2009-02-23
 
   INCOMPATIBLE CHANGES

Propchange: hadoop/core/branches/branch-0.19/CHANGES.txt
------------------------------------------------------------------------------
--- svn:mergeinfo (original)
+++ svn:mergeinfo Wed Mar 11 19:43:48 2009
@@ -1,2 +1,3 @@
 /hadoop/core/branches/branch-0.18/CHANGES.txt:727226
-/hadoop/core/trunk/CHANGES.txt:697306,698176,699056,699098,699415,699424,699444,699490,699517,700163,700628,700923,701273,701398,703923,704203,704261,704701,704703,704707,704712,704732,704748,704989,705391,705420,705430,705762,706350,706707,706719,706796,706802,707258,707262,708623,708641,708710,708723,709040,709303,711717,712881,713888,720602,723013,723460,723831,723918,724883,727117,727212,727217,727228,727869,732572,732777,733887,734870,735082,736426,738697,740077,741703,741762,743296,743745,743892,745180,746902-746903,752073
+/hadoop/core/branches/branch-0.20/CHANGES.txt:752591
+/hadoop/core/trunk/CHANGES.txt:697306,698176,699056,699098,699415,699424,699444,699490,699517,700163,700628,700923,701273,701398,703923,704203,704261,704701,704703,704707,704712,704732,704748,704989,705391,705420,705430,705762,706350,706707,706719,706796,706802,707258,707262,708623,708641,708710,708723,709040,709303,711717,712881,713888,720602,723013,723460,723831,723918,724883,727117,727212,727217,727228,727869,732572,732777,733887,734870,735082,736426,738697,740077,741703,741762,743296,743745,743892,745180,746902-746903,752073,752590

Modified: hadoop/core/branches/branch-0.19/src/hdfs/org/apache/hadoop/hdfs/DFSClient.java
URL: http://svn.apache.org/viewvc/hadoop/core/branches/branch-0.19/src/hdfs/org/apache/hadoop/hdfs/DFSClient.java?rev=752593&r1=752592&r2=752593&view=diff
==============================================================================
--- hadoop/core/branches/branch-0.19/src/hdfs/org/apache/hadoop/hdfs/DFSClient.java (original)
+++ hadoop/core/branches/branch-0.19/src/hdfs/org/apache/hadoop/hdfs/DFSClient.java Wed Mar 11 19:43:48 2009
@@ -205,8 +205,8 @@
    */
   public synchronized void close() throws IOException {
     if(clientRunning) {
-      clientRunning = false;
       leasechecker.close();
+      clientRunning = false;
   
       // close connections to the namenode
       RPC.stopProxy(rpcNamenode);
@@ -2140,7 +2140,6 @@
       private volatile boolean closed = false;
   
       public void run() {
-
         while (!closed && clientRunning) {
 
           // if the Responder encountered an error, shutdown Responder
@@ -2475,21 +2474,30 @@
               // The original bad datanode is left in the list because it is
               // conservative to remove only one datanode in one iteration.
               for (int j = 0; j < nodes.length; j++) {
-                if (nodes[j] ==  primaryNode) {
+                if (nodes[j].equals(primaryNode)) {
                   errorIndex = j; // forget original bad node.
                 }
               }
+              // remove primary node from list
+              newnodes =  new DatanodeInfo[nodes.length-1];
+              System.arraycopy(nodes, 0, newnodes, 0, errorIndex);
+              System.arraycopy(nodes, errorIndex+1, newnodes, errorIndex,
+                               newnodes.length-errorIndex);
+              nodes = newnodes;
               LOG.warn("Error Recovery for block " + block + " failed " +
                        " because recovery from primary datanode " +
                        primaryNode + " failed " + recoveryErrorCount +
-                       " times. Marking primary datanode as bad.");
+                       " times. " + " Pipeline was " + pipelineMsg +
+                       ". Marking primary datanode as bad.");
               recoveryErrorCount = 0; 
+              errorIndex = -1;
               return true;          // sleep when we return from here
             }
             String emsg = "Error Recovery for block " + block + " failed " +
                           " because recovery from primary datanode " +
                           primaryNode + " failed " + recoveryErrorCount + 
-                          " times. Aborting...";
+                          " times. "  + " Pipeline was " + pipelineMsg +
+                          ". Aborting...";
             LOG.warn(emsg);
             lastException = new IOException(emsg);
             closed = true;
@@ -2499,7 +2507,8 @@
           LOG.warn("Error Recovery for block " + block + " failed " +
                    " because recovery from primary datanode " +
                    primaryNode + " failed " + recoveryErrorCount +
-                   " times. Will retry...");
+                   " times. "  + " Pipeline was " + pipelineMsg +
+                   ". Will retry...");
           return true;          // sleep when we return from here
         } finally {
           RPC.stopProxy(primary);

Modified: hadoop/core/branches/branch-0.19/src/test/org/apache/hadoop/hdfs/TestFileCreation.java
URL: http://svn.apache.org/viewvc/hadoop/core/branches/branch-0.19/src/test/org/apache/hadoop/hdfs/TestFileCreation.java?rev=752593&r1=752592&r2=752593&view=diff
==============================================================================
--- hadoop/core/branches/branch-0.19/src/test/org/apache/hadoop/hdfs/TestFileCreation.java (original)
+++ hadoop/core/branches/branch-0.19/src/test/org/apache/hadoop/hdfs/TestFileCreation.java Wed Mar 11 19:43:48 2009
@@ -56,6 +56,7 @@
     //((Log4JLogger)DataNode.LOG).getLogger().setLevel(Level.ALL);
     ((Log4JLogger)LeaseManager.LOG).getLogger().setLevel(Level.ALL);
     ((Log4JLogger)FSNamesystem.LOG).getLogger().setLevel(Level.ALL);
+    ((Log4JLogger)DFSClient.LOG).getLogger().setLevel(Level.ALL);
   }
 
   static final long seed = 0xDEADBEEFL;
@@ -703,4 +704,31 @@
 
     System.out.println("testLeaseExpireHardLimit successful");
   }
+
+  // test closing file system before all file handles are closed.
+  public void testFsClose() throws Exception {
+    System.out.println("test file system close start");
+    final int DATANODE_NUM = 3;
+
+    Configuration conf = new Configuration();
+
+    // create cluster
+    MiniDFSCluster cluster = new MiniDFSCluster(conf, DATANODE_NUM, true, null);
+    DistributedFileSystem dfs = null;
+    try {
+      cluster.waitActive();
+      dfs = (DistributedFileSystem)cluster.getFileSystem();
+
+      // create a new file.
+      final String f = DIR + "foofs";
+      final Path fpath = new Path(f);
+      FSDataOutputStream out = TestFileCreation.createFile(dfs, fpath, DATANODE_NUM);
+      out.write("something".getBytes());
+
+      // close file system without closing file
+      dfs.close();
+    } finally {
+      System.out.println("testFsClose successful");
+    }
+  }
 }