Mailing-List: contact hdfs-commits-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Reply-To: hdfs-dev@hadoop.apache.org
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Subject: svn commit: r1245690 - in
 /hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs: ./
 src/main/java/ src/main/java/org/apache/hadoop/hdfs/ src/main/native/
 src/main/webapps/datanode/ src/main/webapps/hdfs/
 src/main/webapps/secondary/ src/test...
Date: Fri, 17 Feb 2012 18:19:34 -0000
To: hdfs-commits@hadoop.apache.org
From: atm@apache.org
Message-Id: <20120217181935.1A2052388ABA@eris.apache.org>

Author: atm
Date: Fri Feb 17 18:19:31 2012
New Revision: 1245690

URL: http://svn.apache.org/viewvc?rev=1245690&view=rev
Log:
Merge trunk into HA branch.

Modified:
    hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/   (props changed)
    hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
    hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/main/java/   (props changed)
    hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/BlockReaderLocal.java
    hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/main/native/   (props changed)
    hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/datanode/   (props changed)
    hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/hdfs/   (props changed)
    hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/secondary/   (props changed)
    hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/test/hdfs/   (props changed)
    hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestShortCircuitLocalRead.java

Propchange: hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/
------------------------------------------------------------------------------
--- svn:mergeinfo (original)
+++ svn:mergeinfo Fri Feb 17 18:19:31 2012
@@ -1,4 +1,4 @@
-/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs:1152502-1245115
+/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs:1152502-1245677
 /hadoop/core/branches/branch-0.19/hdfs:713112
 /hadoop/hdfs/branches/HDFS-1052:987665-1095512
 /hadoop/hdfs/branches/HDFS-265:796829-820463

Modified: hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt?rev=1245690&r1=1245689&r2=1245690&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt (original)
+++ hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt Fri Feb 17 18:19:31 2012
@@ -132,6 +132,9 @@ Trunk (unreleased changes)
     HDFS-2878. Fix TestBlockRecovery and move it back into main test directory.
     (todd)
 
+    HDFS-2655. BlockReaderLocal#skip performs unnecessary IO. (Brandon Li
+    via jitendra) 
+
   OPTIMIZATIONS
     HDFS-2477. Optimize computing the diff between a block report and the
     namenode state. (Tomasz Nykiel via hairong)
@@ -210,9 +213,6 @@ Trunk (unreleased changes)
     dfs.client.block.write.replace-datanode-on-failure.enable to be mistakenly
     disabled. (atm)
 
-    HDFS-2525. Race between BlockPoolSliceScanner and append. (Brandon Li
-    via jitendra)
-
 Release 0.23.2 - UNRELEASED 
 
   INCOMPATIBLE CHANGES
@@ -251,6 +251,9 @@ Release 0.23.2 - UNRELEASED 
     HDFS-2950. Secondary NN HTTPS address should be listed as a
     NAMESERVICE_SPECIFIC_KEY. (todd)
 
+    HDFS-2525. Race between BlockPoolSliceScanner and append. (Brandon Li
+    via jitendra)
+
     HDFS-2938. Recursive delete of a large directory make namenode
     unresponsive. (Hari Mankude via suresh)
 

Propchange: hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/main/java/
------------------------------------------------------------------------------
--- svn:mergeinfo (original)
+++ svn:mergeinfo Fri Feb 17 18:19:31 2012
@@ -1,4 +1,4 @@
-/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/main/java:1159757-1245115
+/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/main/java:1159757-1245677
 /hadoop/core/branches/branch-0.19/hdfs/src/java:713112
 /hadoop/core/branches/branch-0.19/hdfs/src/main/java:713112
 /hadoop/core/trunk/src/hdfs:776175-785643,785929-786278

Modified: hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/BlockReaderLocal.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/BlockReaderLocal.java?rev=1245690&r1=1245689&r2=1245690&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/BlockReaderLocal.java (original)
+++ hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/BlockReaderLocal.java Fri Feb 17 18:19:31 2012
@@ -369,26 +369,68 @@ class BlockReaderLocal implements BlockR
     if (LOG.isDebugEnabled()) {
       LOG.debug("skip " + n);
     }
+    if (n <= 0) {
+      return 0;
+    }
     if (!verifyChecksum) {
       return dataIn.skip(n);
     }
-    // Skip by reading the data so we stay in sync with checksums.
-    // This could be implemented more efficiently in the future to
-    // skip to the beginning of the appropriate checksum chunk
-    // and then only read to the middle of that chunk.
+  
+    // caller made sure newPosition is not beyond EOF.
+    int remaining = dataBuff.remaining();
+    int position = dataBuff.position();
+    int newPosition = position + (int)n;
+  
+    // if the new offset is already read into dataBuff, just reposition
+    if (n <= remaining) {
+      assert offsetFromChunkBoundary == 0;
+      dataBuff.position(newPosition);
+      return n;
+    }
+  
+    // for small gap, read through to keep the data/checksum in sync
+    if (n - remaining <= bytesPerChecksum) {
+      dataBuff.position(position + remaining);
+      if (skipBuf == null) {
+        skipBuf = new byte[bytesPerChecksum];
+      }
+      int ret = read(skipBuf, 0, (int)(n - remaining));
+      return ret;
+    }
+  
+    // optimize for big gap: discard the current buffer, skip to
+    // the beginning of the appropriate checksum chunk and then
+    // read to the middle of that chunk to be in sync with checksums.
+    this.offsetFromChunkBoundary = newPosition % bytesPerChecksum;
+    long toskip = n - remaining - this.offsetFromChunkBoundary;
+  
+    dataBuff.clear();
+    checksumBuff.clear();
+  
+    long dataSkipped = dataIn.skip(toskip);
+    if (dataSkipped != toskip) {
+      throw new IOException("skip error in data input stream");
+    }
+    long checkSumOffset = (dataSkipped / bytesPerChecksum) * checksumSize;
+    if (checkSumOffset > 0) {
+      long skipped = checksumIn.skip(checkSumOffset);
+      if (skipped != checkSumOffset) {
+        throw new IOException("skip error in checksum input stream");
+      }
+    }
+
+    // read into the middle of the chunk
     if (skipBuf == null) {
       skipBuf = new byte[bytesPerChecksum];
     }
-    long nSkipped = 0;
-    while ( nSkipped < n ) {
-      int toSkip = (int)Math.min(n-nSkipped, skipBuf.length);
-      int ret = read(skipBuf, 0, toSkip);
-      if ( ret <= 0 ) {
-        return nSkipped;
-      }
-      nSkipped += ret;
+    assert skipBuf.length == bytesPerChecksum;
+    assert this.offsetFromChunkBoundary < bytesPerChecksum;
+    int ret = read(skipBuf, 0, this.offsetFromChunkBoundary);
+    if (ret == -1) {  // EOS
+      return toskip;
+    } else {
+      return (toskip + ret);
     }
-    return nSkipped;
   }
 
   @Override

Propchange: hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/main/native/
------------------------------------------------------------------------------
--- svn:mergeinfo (original)
+++ svn:mergeinfo Fri Feb 17 18:19:31 2012
@@ -1,4 +1,4 @@
-/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/main/native:1152502-1245115
+/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/main/native:1152502-1245677
 /hadoop/core/branches/branch-0.19/hdfs/src/main/native:713112
 /hadoop/core/branches/branch-0.19/mapred/src/c++/libhdfs:713112
 /hadoop/core/trunk/src/c++/libhdfs:776175-784663

Propchange: hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/datanode/
------------------------------------------------------------------------------
--- svn:mergeinfo (original)
+++ svn:mergeinfo Fri Feb 17 18:19:31 2012
@@ -1,4 +1,4 @@
-/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/datanode:1159757-1245115
+/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/datanode:1159757-1245677
 /hadoop/core/branches/branch-0.19/hdfs/src/main/webapps/datanode:713112
 /hadoop/core/branches/branch-0.19/hdfs/src/webapps/datanode:713112
 /hadoop/core/trunk/src/webapps/datanode:776175-784663

Propchange: hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/hdfs/
------------------------------------------------------------------------------
--- svn:mergeinfo (original)
+++ svn:mergeinfo Fri Feb 17 18:19:31 2012
@@ -1,4 +1,4 @@
-/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/hdfs:1152502-1245115
+/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/hdfs:1152502-1245677
 /hadoop/core/branches/branch-0.19/hdfs/src/main/webapps/hdfs:713112
 /hadoop/core/branches/branch-0.19/hdfs/src/webapps/hdfs:713112
 /hadoop/core/trunk/src/webapps/hdfs:776175-784663

Propchange: hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/secondary/
------------------------------------------------------------------------------
--- svn:mergeinfo (original)
+++ svn:mergeinfo Fri Feb 17 18:19:31 2012
@@ -1,4 +1,4 @@
-/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/secondary:1152502-1245115
+/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/secondary:1152502-1245677
 /hadoop/core/branches/branch-0.19/hdfs/src/main/webapps/secondary:713112
 /hadoop/core/branches/branch-0.19/hdfs/src/webapps/secondary:713112
 /hadoop/core/trunk/src/webapps/secondary:776175-784663

Propchange: hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/test/hdfs/
------------------------------------------------------------------------------
--- svn:mergeinfo (original)
+++ svn:mergeinfo Fri Feb 17 18:19:31 2012
@@ -1,4 +1,4 @@
-/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/test/hdfs:1159757-1245115
+/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/test/hdfs:1159757-1245677
 /hadoop/core/branches/branch-0.19/hdfs/src/test/hdfs:713112
 /hadoop/core/trunk/src/test/hdfs:776175-785643
 /hadoop/hdfs/branches/HDFS-1052/src/test/hdfs:987665-1095512

Modified: hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestShortCircuitLocalRead.java
URL: http://svn.apache.org/viewvc/hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestShortCircuitLocalRead.java?rev=1245690&r1=1245689&r2=1245690&view=diff
==============================================================================
--- hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestShortCircuitLocalRead.java (original)
+++ hadoop/common/branches/HDFS-1623/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestShortCircuitLocalRead.java Fri Feb 17 18:19:31 2012
@@ -238,7 +238,53 @@ public class TestShortCircuitLocalRead {
       cluster.shutdown();
     }
   }
+
+  @Test
+  public void testSkipWithVerifyChecksum() throws IOException {
+    int size = blockSize;
+    Configuration conf = new Configuration();
+    conf.setBoolean(DFSConfigKeys.DFS_CLIENT_READ_SHORTCIRCUIT_KEY, true);
+    conf.setBoolean(DFSConfigKeys.DFS_CLIENT_READ_SHORTCIRCUIT_SKIP_CHECKSUM_KEY, false);
+    conf.set(DFSConfigKeys.DFS_BLOCK_LOCAL_PATH_ACCESS_USER_KEY,
+        UserGroupInformation.getCurrentUser().getShortUserName());
+    if (simulatedStorage) {
+      conf.setBoolean(SimulatedFSDataset.CONFIG_PROPERTY_SIMULATED, true);
+    }
+    MiniDFSCluster cluster = new MiniDFSCluster.Builder(conf).numDataNodes(1)
+        .format(true).build();
+    FileSystem fs = cluster.getFileSystem();
+    try {
+      // check that / exists
+      Path path = new Path("/");
+      assertTrue("/ should be a directory", fs.getFileStatus(path)
+          .isDirectory() == true);
+      
+      byte[] fileData = AppendTestUtil.randomBytes(seed, size*3);
+      // create a new file in home directory. Do not close it.
+      Path file1 = new Path("filelocal.dat");
+      FSDataOutputStream stm = createFile(fs, file1, 1);
   
+      // write to file
+      stm.write(fileData);
+      stm.close();
+      
+      // now test the skip function
+      FSDataInputStream instm = fs.open(file1);
+      byte[] actual = new byte[fileData.length];
+      // read something from the block first, otherwise BlockReaderLocal.skip()
+      // will not be invoked
+      int nread = instm.read(actual, 0, 3);
+      long skipped = 2*size+3;
+      instm.seek(skipped);
+      nread = instm.read(actual, (int)(skipped + nread), 3);
+      instm.close();
+        
+    } finally {
+      fs.close();
+      cluster.shutdown();
+    }
+  }
+     
   /**
    * Test to run benchmarks between shortcircuit read vs regular read with
    * specified number of threads simultaneously reading.