HDFS-5433. When reloading fsimage during checkpointing, we should clear existing snapshottable directories. Contributed by Aaron T. Myers.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1537192 13f79535-47bb-0310-9956-ffa450edef68

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

@@ -565,6 +565,9 @@ Release 2.2.1 - UNRELEASED
     HDFS-5413. hdfs.cmd does not support passthrough to any arbitrary class.
     (cnauroth)
 
+    HDFS-5433. When reloading fsimage during checkpointing, we should clear
+    existing snapshottable directories. (Aaron T. Myers via wang)
+
 Release 2.2.0 - 2013-10-13
 
   INCOMPATIBLE CHANGES

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

@@ -518,6 +518,7 @@ void clear() {
     generationStampV1Limit = GenerationStamp.GRANDFATHER_GENERATION_STAMP;
     leaseManager.removeAllLeases();
     inodeId.setCurrentValue(INodeId.LAST_RESERVED_ID);
+    snapshotManager.clearSnapshottableDirs();
   }
 
   @VisibleForTesting

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

@@ -155,7 +155,7 @@ int getMergeErrorCount() {
   }
 
   @VisibleForTesting
-  FSNamesystem getFSNamesystem() {
+  public FSNamesystem getFSNamesystem() {
     return namesystem;
   }
   
@@ -489,7 +489,8 @@ private InetSocketAddress getImageListenAddress() {
    * Create a new checkpoint
    * @return if the image is fetched from primary or not
    */
-  boolean doCheckpoint() throws IOException {
+  @VisibleForTesting
+  public boolean doCheckpoint() throws IOException {
     checkpointImage.ensureCurrentDirExists();
     NNStorage dstStorage = checkpointImage.getStorage();
     

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

@@ -358,6 +358,10 @@ public SnapshotDiffInfo diff(final String path, final String from,
     return snapshotRoot.computeDiff(from, to);
   }
   
+  public void clearSnapshottableDirs() {
+    snapshottables.clear();
+  }
+
   /**
    * Returns the maximum allowable snapshot ID based on the bit width of the
    * snapshot ID.

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

@@ -0,0 +1,131 @@
+/**
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package org.apache.hadoop.hdfs.server.namenode.snapshot;
+
+import static org.junit.Assert.*;
+
+import java.io.File;
+import java.io.IOException;
+
+import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.fs.FileSystem;
+import org.apache.hadoop.fs.FileUtil;
+import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.hdfs.DFSConfigKeys;
+import org.apache.hadoop.hdfs.HdfsConfiguration;
+import org.apache.hadoop.hdfs.MiniDFSCluster;
+import org.apache.hadoop.hdfs.client.HdfsAdmin;
+import org.apache.hadoop.hdfs.server.namenode.NameNodeAdapter;
+import org.apache.hadoop.hdfs.server.namenode.SecondaryNameNode;
+import org.junit.Before;
+import org.junit.Test;
+
+public class TestCheckpointsWithSnapshots {
+  
+  private static final Path TEST_PATH = new Path("/foo");
+  private static Configuration conf = new HdfsConfiguration();
+  static {
+    conf.set(DFSConfigKeys.DFS_NAMENODE_SECONDARY_HTTP_ADDRESS_KEY, "0.0.0.0:0");
+  }
+  
+  @Before
+  public void setUp() {
+    FileUtil.fullyDeleteContents(new File(MiniDFSCluster.getBaseDirectory()));
+  }
+
+  /**
+   * Regression test for HDFS-5433 - "When reloading fsimage during
+   * checkpointing, we should clear existing snapshottable directories"
+   */
+  @Test
+  public void testCheckpoint() throws IOException {
+    MiniDFSCluster cluster = null;
+    SecondaryNameNode secondary = null;
+    try {
+      cluster = new MiniDFSCluster.Builder(conf).build();
+      cluster.waitActive();
+      secondary = new SecondaryNameNode(conf);
+      SnapshotManager nnSnapshotManager = cluster.getNamesystem().getSnapshotManager();
+      SnapshotManager secondarySnapshotManager = secondary.getFSNamesystem().getSnapshotManager();
+      
+      FileSystem fs = cluster.getFileSystem();
+      HdfsAdmin admin =  new HdfsAdmin(FileSystem.getDefaultUri(conf), conf);
+      
+      assertEquals(0, nnSnapshotManager.getNumSnapshots());
+      assertEquals(0, nnSnapshotManager.getNumSnapshottableDirs());
+      assertEquals(0, secondarySnapshotManager.getNumSnapshots());
+      assertEquals(0, secondarySnapshotManager.getNumSnapshottableDirs());
+      
+      // 1. Create a snapshottable directory foo on the NN.
+      fs.mkdirs(TEST_PATH);
+      admin.allowSnapshot(TEST_PATH);
+      assertEquals(0, nnSnapshotManager.getNumSnapshots());
+      assertEquals(1, nnSnapshotManager.getNumSnapshottableDirs());
+      
+      // 2. Create a snapshot of the dir foo. This will be referenced both in
+      // the SnapshotManager as well as in the file system tree. The snapshot
+      // count will go up to 1.
+      Path snapshotPath = fs.createSnapshot(TEST_PATH);
+      assertEquals(1, nnSnapshotManager.getNumSnapshots());
+      assertEquals(1, nnSnapshotManager.getNumSnapshottableDirs());
+      
+      // 3. Start up a 2NN and have it do a checkpoint. It will have foo and its
+      // snapshot in its list of snapshottable dirs referenced from the
+      // SnapshotManager, as well as in the file system tree.
+      secondary.doCheckpoint();
+      assertEquals(1, secondarySnapshotManager.getNumSnapshots());
+      assertEquals(1, secondarySnapshotManager.getNumSnapshottableDirs());
+      
+      // 4. Disallow snapshots on and delete foo on the NN. The snapshot count
+      // will go down to 0 and the snapshottable dir will be removed from the fs
+      // tree.
+      fs.deleteSnapshot(TEST_PATH, snapshotPath.getName());
+      admin.disallowSnapshot(TEST_PATH);
+      assertEquals(0, nnSnapshotManager.getNumSnapshots());
+      assertEquals(0, nnSnapshotManager.getNumSnapshottableDirs());
+      
+      // 5. Have the NN do a saveNamespace, writing out a new fsimage with
+      // snapshot count 0.
+      NameNodeAdapter.enterSafeMode(cluster.getNameNode(), false);
+      NameNodeAdapter.saveNamespace(cluster.getNameNode());
+      NameNodeAdapter.leaveSafeMode(cluster.getNameNode());
+      
+      // 6. Have the still-running 2NN do a checkpoint. It will notice that the
+      // fsimage has changed on the NN and redownload/reload from that image.
+      // This will replace all INodes in the file system tree as well as reset
+      // the snapshot counter to 0 in the SnapshotManager. However, it will not
+      // clear the list of snapshottable dirs referenced from the
+      // SnapshotManager. When it writes out an fsimage, the 2NN will write out
+      // 0 for the snapshot count, but still serialize the snapshottable dir
+      // referenced in the SnapshotManager even though it no longer appears in
+      // the file system tree. The NN will not be able to start up with this.
+      secondary.doCheckpoint();
+      assertEquals(0, secondarySnapshotManager.getNumSnapshots());
+      assertEquals(0, secondarySnapshotManager.getNumSnapshottableDirs());
+    } finally {
+      if (cluster != null) {
+        cluster.shutdown();
+      }
+      if (secondary != null) {
+        secondary.shutdown();
+      }
+    }
+
+  }
+
+}