HDFS-14557. JournalNode error: Can't scan a pre-transactional edit log. Contributed by Stephen O'Donnell.

Signed-off-by: Wei-Chiu Chuang <weichiu@apache.org>

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

@@ -163,6 +163,16 @@ public class EditLogFileInputStream extends EditLogInputStream {
       } catch (EOFException eofe) {
         throw new LogHeaderCorruptException("No header found in log");
       }
+      if (logVersion == -1) {
+        // The edits in progress file is pre-allocated with 1MB of "-1" bytes
+        // when it is created, then the header is written. If the header is
+        // -1, it indicates the an exception occurred pre-allocating the file
+        // and the header was never written. Therefore this is effectively a
+        // corrupt and empty log.
+        throw new LogHeaderCorruptException("No header present in log (value " +
+            "is -1), probably due to disk space issues when it was created. " +
+            "The log has no transactions and will be sidelined.");
+      }
       // We assume future layout will also support ADD_LAYOUT_FLAGS
       if (NameNodeLayoutVersion.supports(
           LayoutVersion.Feature.ADD_LAYOUT_FLAGS, logVersion) ||

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

@@ -1285,6 +1285,12 @@ public class FSEditLogLoader {
             + lastPos, t);
         in.resync();
         FSImage.LOG.warn("After resync, position is " + in.getPosition());
+        if (in.getPosition() <= lastPos) {
+          FSImage.LOG.warn("After resync, the position, {} is not greater " +
+              "than the previous position {}. Skipping remainder of this log.",
+              in.getPosition(), lastPos);
+          break;
+        }
         continue;
       }
       if (lastTxId == HdfsServerConstants.INVALID_TXID || txid > lastTxId) {

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/qjournal/server/TestJournal.java

@@ -28,6 +28,7 @@ import java.io.ByteArrayOutputStream;
 import java.io.DataOutputStream;
 import java.io.File;
 import java.io.IOException;
+import java.io.RandomAccessFile;
 
 import org.apache.hadoop.conf.Configuration;
 import org.apache.hadoop.fs.FileUtil;
@@ -125,6 +126,48 @@ public class TestJournal {
     Assert.assertEquals(1, segmentState.getStartTxId());
   }
 
+  /**
+   * Test for HDFS-14557 to ensure that a edit file that failed to fully
+   * allocate and has a header byte of -1 is moved aside to allow startup
+   * to progress.
+   */
+  @Test
+  public void testEmptyEditsInProgressMovedAside() throws Exception {
+    // First, write 5 transactions to the journal
+    journal.startLogSegment(makeRI(1), 1,
+        NameNodeLayoutVersion.CURRENT_LAYOUT_VERSION - 1);
+    final int numTxns = 5;
+    byte[] ops = QJMTestUtil.createTxnData(1, 5);
+    journal.journal(makeRI(2), 1, 1, numTxns, ops);
+    // Now close the segment
+    journal.finalizeLogSegment(makeRI(3), 1, numTxns);
+
+    // Create a new segment creating a new edits_inprogress file
+    journal.startLogSegment(makeRI(4), 6,
+        NameNodeLayoutVersion.CURRENT_LAYOUT_VERSION - 1);
+    ops = QJMTestUtil.createTxnData(6, 5);
+    journal.journal(makeRI(5), 6, 6, numTxns, ops);
+    File eip = journal.getStorage().getInProgressEditLog(6);
+
+    // Now stop the journal without finalizing the segment
+    journal.close();
+
+    // Now "zero out" the EIP file with -1 bytes, similar to how it would
+    // appear if the pre-allocation failed
+    RandomAccessFile rwf = new RandomAccessFile(eip, "rw");
+    for (int i=0; i<rwf.length(); i++) {
+      rwf.write(-1);
+    }
+    rwf.close();
+
+    // Finally start the Journal again, and ensure the "zeroed out" file
+    // is renamed with a .empty extension
+    journal = new Journal(conf, TEST_LOG_DIR, JID, StartupOption.REGULAR,
+        mockErrorReporter);
+    File movedTo = new File(eip.getAbsolutePath()+".empty");
+    assertTrue(movedTo.exists());
+  }
+
   @Test (timeout = 10000)
   public void testEpochHandling() throws Exception {
     assertEquals(0, journal.getLastPromisedEpoch());

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

@@ -28,6 +28,7 @@ import java.io.ByteArrayInputStream;
 import java.io.File;
 import java.io.IOException;
 import java.io.RandomAccessFile;
+import java.io.FileOutputStream;
 import java.net.HttpURLConnection;
 import java.net.URL;
 import java.util.EnumMap;
@@ -160,4 +161,25 @@ public class TestEditLogFileInputStream {
     }
     elis.close();
   }
+
+  /**
+   * Regression test for HDFS-14557 which verifies that an edit log filled
+   * with only "-1" bytes is moved aside and does not prevent the Journal
+   * node from starting.
+   */
+  @Test(timeout=60000)
+  public void testScanEditThatFailedDuringPreAllocate() throws Exception {
+    Configuration conf = new Configuration();
+    File editLog = new File(GenericTestUtils.getTempPath("testCorruptEditLog"));
+    FileOutputStream os = new FileOutputStream(editLog);
+    for (int i=0; i<1024; i++) {
+      os.write(-1);
+    }
+    os.close();
+    FSEditLogLoader.EditLogValidation val =
+        EditLogFileInputStream.scanEditLog(editLog, 1234, false);
+    assertEquals(true, val.hasCorruptHeader());
+    assertEquals(HdfsServerConstants.INVALID_TXID, val.getEndTxId());
+  }
+
 }