HDFS-3885. QJM: optimize log sync when JN is lagging behind. Contributed by Todd Lipcon.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/branches/HDFS-3077@1383039 13f79535-47bb-0310-9956-ffa450edef68

hadoop-hdfs-project/hadoop-hdfs/CHANGES.HDFS-3077.txt

@@ -54,3 +54,5 @@ HDFS-3893. QJM: Make QJM work with security enabled. (atm)
 HDFS-3897. QJM: TestBlockToken fails after HDFS-3893. (atm)
 
 HDFS-3898. QJM: enable TCP_NODELAY for IPC (todd)
+
+HDFS-3885. QJM: optimize log sync when JN is lagging behind (todd)

hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/main/java/org/apache/hadoop/contrib/bkjournal/BookKeeperEditLogOutputStream.java

@@ -84,7 +84,7 @@ class BookKeeperEditLogOutputStream
   @Override
   public void close() throws IOException {
     setReadyToFlush();
-    flushAndSync();
+    flushAndSync(true);
     try {
       lh.close();
     } catch (InterruptedException ie) {
@@ -130,7 +130,7 @@ class BookKeeperEditLogOutputStream
   }
 
   @Override
-  public void flushAndSync() throws IOException {
+  public void flushAndSync(boolean durable) throws IOException {
     assert(syncLatch != null);
     try {
       syncLatch.await();

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/qjournal/client/QuorumOutputStream.java

@@ -77,7 +77,7 @@ class QuorumOutputStream extends EditLogOutputStream {
   }
 
   @Override
-  protected void flushAndSync() throws IOException {
+  protected void flushAndSync(boolean durable) throws IOException {
     int numReadyBytes = buf.countReadyBytes();
     if (numReadyBytes > 0) {
       int numReadyTxns = buf.countReadyTxns();

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/qjournal/server/Journal.java

@@ -301,17 +301,23 @@ class Journal implements Closeable {
     if (LOG.isTraceEnabled()) {
       LOG.trace("Writing txid " + firstTxnId + "-" + lastTxnId);
     }
+
+    // If the edit has already been marked as committed, we know
+    // it has been fsynced on a quorum of other nodes, and we are
+    // "catching up" with the rest. Hence we do not need to fsync.
+    boolean isLagging = lastTxnId <= committedTxnId.get();
+    boolean shouldFsync = !isLagging;
     
     curSegment.writeRaw(records, 0, records.length);
     curSegment.setReadyToFlush();
     Stopwatch sw = new Stopwatch();
     sw.start();
-    curSegment.flush();
+    curSegment.flush(shouldFsync);
     sw.stop();
     
     metrics.addSync(sw.elapsedTime(TimeUnit.MICROSECONDS));
-    
-    if (committedTxnId.get() > lastTxnId) {
+
+    if (isLagging) {
       // This batch of edits has already been committed on a quorum of other
       // nodes. So, we are in "catch up" mode. This gets its own metric.
       metrics.batchesWrittenWhileLagging.incr(1);

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/EditLogBackupOutputStream.java

@@ -114,7 +114,7 @@ class EditLogBackupOutputStream extends EditLogOutputStream {
   }
 
   @Override // EditLogOutputStream
-  protected void flushAndSync() throws IOException {
+  protected void flushAndSync(boolean durable) throws IOException {
     assert out.getLength() == 0 : "Output buffer is not empty";
     
     int numReadyTxns = doubleBuf.countReadyTxns();

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/EditLogFileOutputStream.java

@@ -176,7 +176,7 @@ public class EditLogFileOutputStream extends EditLogOutputStream {
    * accumulates new log records while readyBuffer will be flushed and synced.
    */
   @Override
-  public void flushAndSync() throws IOException {
+  public void flushAndSync(boolean durable) throws IOException {
     if (fp == null) {
       throw new IOException("Trying to use aborted output stream");
     }
@@ -186,7 +186,7 @@ public class EditLogFileOutputStream extends EditLogOutputStream {
     }
     preallocate(); // preallocate file if necessay
     doubleBuf.flushTo(fp);
-    if (!shouldSkipFsyncForTests) {
+    if (durable && !shouldSkipFsyncForTests) {
       fc.force(false); // metadata updates not needed
     }
   }

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/EditLogOutputStream.java

@@ -93,18 +93,24 @@ public abstract class EditLogOutputStream implements Closeable {
   /**
    * Flush and sync all data that is ready to be flush 
    * {@link #setReadyToFlush()} into underlying persistent store.
+   * @param durable if true, the edits should be made truly durable before
+   * returning
    * @throws IOException
    */
-  abstract protected void flushAndSync() throws IOException;
+  abstract protected void flushAndSync(boolean durable) throws IOException;
 
   /**
    * Flush data to persistent store.
    * Collect sync metrics.
    */
   public void flush() throws IOException {
+    flush(true);
+  }
+  
+  public void flush(boolean durable) throws IOException {
     numSync++;
     long start = now();
-    flushAndSync();
+    flushAndSync(durable);
     long end = now();
     totalTimeSync += (end - start);
   }

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/JournalSet.java

@@ -471,12 +471,12 @@ public class JournalSet implements JournalManager {
     }
 
     @Override
-    protected void flushAndSync() throws IOException {
+    protected void flushAndSync(final boolean durable) throws IOException {
       mapJournalsAndReportErrors(new JournalClosure() {
         @Override
         public void apply(JournalAndStream jas) throws IOException {
           if (jas.isActive()) {
-            jas.getCurrentStream().flushAndSync();
+            jas.getCurrentStream().flushAndSync(durable);
           }
         }
       }, "flushAndSync");

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/tools/offlineEditsViewer/BinaryEditsVisitor.java

@@ -56,7 +56,7 @@ public class BinaryEditsVisitor implements OfflineEditsVisitor {
   @Override
   public void close(Throwable error) throws IOException {
     elfos.setReadyToFlush();
-    elfos.flushAndSync();
+    elfos.flushAndSync(true);
     elfos.close();
   }
 

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

@@ -1222,7 +1222,7 @@ public class TestEditLog {
       elfos.create();
       elfos.writeRaw(garbage, 0, garbage.length);
       elfos.setReadyToFlush();
-      elfos.flushAndSync();
+      elfos.flushAndSync(true);
       elfos.close();
       elfos = null;
       file = new File(TEST_LOG_NAME);

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

@@ -55,7 +55,7 @@ public class TestEditLogFileOutputStream {
   static void flushAndCheckLength(EditLogFileOutputStream elos,
       long expectedLength) throws IOException {
     elos.setReadyToFlush();
-    elos.flushAndSync();
+    elos.flushAndSync(true);
     assertEquals(expectedLength, elos.getFile().length());
   }
   

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

@@ -74,7 +74,7 @@ public class TestNameNodeRecovery {
 
       elts.addTransactionsToLog(elfos, cache);
       elfos.setReadyToFlush();
-      elfos.flushAndSync();
+      elfos.flushAndSync(true);
       elfos.close();
       elfos = null;
       file = new File(TEST_LOG_NAME);