HDFS-14317. Ensure checkpoints are created when in-progress edit log tailing is enabled with a period shorter than the log roll period. Contributed by Ekanth Sethuramalingam.

(cherry-picked from commit 1bc282e0b3)

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/DFSConfigKeys.java

@@ -327,7 +327,8 @@ public class DFSConfigKeys extends CommonConfigurationKeys {
   public static final int     DFS_NAMENODE_QUOTA_INIT_THREADS_DEFAULT = 4;
 
   public static final String  DFS_NAMENODE_EDIT_LOG_AUTOROLL_MULTIPLIER_THRESHOLD = "dfs.namenode.edit.log.autoroll.multiplier.threshold";
-  public static final float   DFS_NAMENODE_EDIT_LOG_AUTOROLL_MULTIPLIER_THRESHOLD_DEFAULT = 2.0f;
+  public static final float
+      DFS_NAMENODE_EDIT_LOG_AUTOROLL_MULTIPLIER_THRESHOLD_DEFAULT = 0.5f;
   public static final String  DFS_NAMENODE_EDIT_LOG_AUTOROLL_CHECK_INTERVAL_MS = "dfs.namenode.edit.log.autoroll.check.interval.ms";
   public static final int     DFS_NAMENODE_EDIT_LOG_AUTOROLL_CHECK_INTERVAL_MS_DEFAULT = 5*60*1000;
 

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

@@ -563,7 +563,8 @@ public class FSEditLog implements LogsPurgeable {
   /**
    * @return the first transaction ID in the current log segment
    */
-  synchronized long getCurSegmentTxId() {
+  @VisibleForTesting
+  public synchronized long getCurSegmentTxId() {
     Preconditions.checkState(isSegmentOpen(),
         "Bad state: %s", state);
     return curSegmentTxId;

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

@@ -110,6 +110,11 @@ public class EditLogTailer {
    */
   private long lastLoadTimeMs;
 
+  /**
+   * The last time we triggered a edit log roll on active namenode.
+   */
+  private long lastRollTimeMs;
+
   /**
    * How often the Standby should roll edit logs. Since the Standby only reads
    * from finalized log segments, the Standby will only be as up-to-date as how
@@ -140,7 +145,8 @@ public class EditLogTailer {
   private int nnLoopCount = 0;
 
   /**
-   * maximum number of retries we should give each of the remote namenodes before giving up
+   * Maximum number of retries we should give each of the remote namenodes
+   * before giving up.
    */
   private int maxRetries;
 
@@ -162,6 +168,7 @@ public class EditLogTailer {
     this.editLog = namesystem.getEditLog();
     
     lastLoadTimeMs = monotonicNow();
+    lastRollTimeMs = monotonicNow();
 
     logRollPeriodMs = conf.getTimeDuration(
         DFSConfigKeys.DFS_HA_LOGROLL_PERIOD_KEY,
@@ -350,7 +357,7 @@ public class EditLogTailer {
    */
   private boolean tooLongSinceLastLoad() {
     return logRollPeriodMs >= 0 && 
-      (monotonicNow() - lastLoadTimeMs) > logRollPeriodMs ;
+      (monotonicNow() - lastRollTimeMs) > logRollPeriodMs;
   }
 
   /**
@@ -378,6 +385,7 @@ public class EditLogTailer {
     try {
       future = rollEditsRpcExecutor.submit(getNameNodeProxy());
       future.get(rollEditsTimeoutMs, TimeUnit.MILLISECONDS);
+      lastRollTimeMs = monotonicNow();
       lastRollTriggerTxId = lastLoadedTxnId;
     } catch (ExecutionException e) {
       LOG.warn("Unable to trigger a roll of the active NN", e);

hadoop-hdfs-project/hadoop-hdfs/src/main/resources/hdfs-default.xml

@@ -2600,7 +2600,7 @@
 
 <property>
   <name>dfs.namenode.edit.log.autoroll.multiplier.threshold</name>
-  <value>2.0</value>
+  <value>0.5</value>
   <description>
     Determines when an active namenode will roll its own edit log.
     The actual threshold (in number of edits) is determined by multiplying

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

@@ -338,9 +338,97 @@ public class TestEditLogTailer {
     }
   }
 
+  @Test
+  public void testStandbyTriggersLogRollsWhenTailInProgressEdits()
+      throws Exception {
+    // Time in seconds to wait for standby to catch up to edits from active
+    final int standbyCatchupWaitTime = 2;
+    // Time in seconds to wait before checking if edit logs are rolled while
+    // expecting no edit log roll
+    final int noLogRollWaitTime = 2;
+    // Time in seconds to wait before checking if edit logs are rolled while
+    // expecting edit log roll
+    final int logRollWaitTime = 3;
+
+    Configuration conf = getConf();
+    conf.setInt(DFSConfigKeys.DFS_HA_LOGROLL_PERIOD_KEY,
+        standbyCatchupWaitTime + noLogRollWaitTime + 1);
+    conf.setInt(DFSConfigKeys.DFS_HA_TAILEDITS_PERIOD_KEY, 1);
+    conf.setBoolean(DFSConfigKeys.DFS_HA_TAILEDITS_INPROGRESS_KEY, true);
+
+    MiniDFSCluster cluster = createMiniDFSCluster(conf, 2);
+    if (cluster == null) {
+      fail("failed to start mini cluster.");
+    }
+
+    try {
+      int activeIndex = new Random().nextBoolean() ? 1 : 0;
+      int standbyIndex = (activeIndex == 0) ? 1 : 0;
+      cluster.transitionToActive(activeIndex);
+      NameNode active = cluster.getNameNode(activeIndex);
+      NameNode standby = cluster.getNameNode(standbyIndex);
+
+      long origTxId = active.getNamesystem().getFSImage().getEditLog()
+          .getCurSegmentTxId();
+      for (int i = 0; i < DIRS_TO_MAKE / 2; i++) {
+        NameNodeAdapter.mkdirs(active, getDirPath(i),
+            new PermissionStatus("test", "test",
+            new FsPermission((short)00755)), true);
+      }
+
+      long activeTxId = active.getNamesystem().getFSImage().getEditLog()
+          .getLastWrittenTxId();
+      waitForStandbyToCatchUpWithInProgressEdits(standby, activeTxId,
+          standbyCatchupWaitTime);
+
+      for (int i = DIRS_TO_MAKE / 2; i < DIRS_TO_MAKE; i++) {
+        NameNodeAdapter.mkdirs(active, getDirPath(i),
+            new PermissionStatus("test", "test",
+            new FsPermission((short)00755)), true);
+      }
+
+      boolean exceptionThrown = false;
+      try {
+        checkForLogRoll(active, origTxId, noLogRollWaitTime);
+      } catch (TimeoutException e) {
+        exceptionThrown = true;
+      }
+      assertTrue(exceptionThrown);
+
+      checkForLogRoll(active, origTxId, logRollWaitTime);
+    } finally {
+      cluster.shutdown();
+    }
+  }
+
+  private static void waitForStandbyToCatchUpWithInProgressEdits(
+      final NameNode standby, final long activeTxId,
+      int maxWaitSec) throws Exception {
+    GenericTestUtils.waitFor(new Supplier<Boolean>() {
+      @Override
+      public Boolean get() {
+        long standbyTxId = standby.getNamesystem().getFSImage()
+            .getLastAppliedTxId();
+        return (standbyTxId >= activeTxId);
+      }
+    }, 100, maxWaitSec * 1000);
+  }
+
+  private static void checkForLogRoll(final NameNode active,
+      final long origTxId, int maxWaitSec) throws Exception {
+    GenericTestUtils.waitFor(new Supplier<Boolean>() {
+      @Override
+      public Boolean get() {
+        long curSegmentTxId = active.getNamesystem().getFSImage().getEditLog()
+            .getCurSegmentTxId();
+        return (origTxId != curSegmentTxId);
+      }
+    }, 100, maxWaitSec * 1000);
+  }
+
   private static MiniDFSCluster createMiniDFSCluster(Configuration conf,
       int nnCount) throws IOException {
-    int basePort = 10060 + new Random().nextInt(100) * 2;
+    int basePort = 10060 + new Random().nextInt(1000) * 2;
 
     // By passing in basePort, name node will have IPC port set,
     // which is needed for enabling roll log.

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

@@ -255,10 +255,10 @@ public class TestFailureToReadEdits {
     
     // Once the standby catches up, it should notice that it needs to
     // do a checkpoint and save one to its local directories.
-    HATestUtil.waitForCheckpoint(cluster, 1, ImmutableList.of(0, 3));
+    HATestUtil.waitForCheckpoint(cluster, 1, ImmutableList.of(0, 5));
     
     // It should also upload it back to the active.
-    HATestUtil.waitForCheckpoint(cluster, 0, ImmutableList.of(0, 3));
+    HATestUtil.waitForCheckpoint(cluster, 0, ImmutableList.of(0, 5));
     
     causeFailureOnEditLogRead();
     
@@ -273,15 +273,15 @@ public class TestFailureToReadEdits {
     }
     
     // 5 because we should get OP_START_LOG_SEGMENT and one successful OP_MKDIR
-    HATestUtil.waitForCheckpoint(cluster, 1, ImmutableList.of(0, 3, 5));
+    HATestUtil.waitForCheckpoint(cluster, 1, ImmutableList.of(0, 5, 7));
     
     // It should also upload it back to the active.
-    HATestUtil.waitForCheckpoint(cluster, 0, ImmutableList.of(0, 3, 5));
+    HATestUtil.waitForCheckpoint(cluster, 0, ImmutableList.of(0, 5, 7));
 
     // Restart the active NN
     cluster.restartNameNode(0);
     
-    HATestUtil.waitForCheckpoint(cluster, 0, ImmutableList.of(0, 3, 5));
+    HATestUtil.waitForCheckpoint(cluster, 0, ImmutableList.of(0, 5, 7));
     
     FileSystem fs0 = null;
     try {
@@ -310,7 +310,7 @@ public class TestFailureToReadEdits {
     HATestUtil.waitForStandbyToCatchUp(nn0, nn1);
     
     // It should also upload it back to the active.
-    HATestUtil.waitForCheckpoint(cluster, 0, ImmutableList.of(0, 3));
+    HATestUtil.waitForCheckpoint(cluster, 0, ImmutableList.of(0, 5));
     
     causeFailureOnEditLogRead();