HDFS-7185. The active NameNode will not accept an fsimage sent from the standby during rolling upgrade. Contributed by Jing Zhao.

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

@@ -604,6 +604,9 @@ Release 2.6.0 - UNRELEASED
     HDFS-7237. The command "hdfs namenode -rollingUpgrade" throws
     ArrayIndexOutOfBoundsException.  (szetszwo)
 
+    HDFS-7185. The active NameNode will not accept an fsimage sent from the
+    standby during rolling upgrade. (jing9)
+
     BREAKDOWN OF HDFS-6134 AND HADOOP-10150 SUBTASKS AND RELATED JIRAS
   
       HDFS-6387. HDFS CLI admin tool for creating & deleting an

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

@@ -322,7 +322,7 @@ public class FSImage implements Closeable {
         if (curState != StorageState.NOT_FORMATTED 
             && startOpt != StartupOption.ROLLBACK) {
           // read and verify consistency with other directories
-          storage.readProperties(sd);
+          storage.readProperties(sd, startOpt);
           isFormatted = true;
         }
         if (startOpt == StartupOption.IMPORT && isFormatted)
@@ -563,7 +563,7 @@ public class FSImage implements Closeable {
     assert editLog != null : "editLog must be initialized";
     editLog.openForWrite();
     storage.writeTransactionIdFileToStorage(editLog.getCurSegmentTxId());
-  };
+  }
   
   /**
    * Toss the current image and namesystem, reloading from the specified
@@ -572,7 +572,7 @@ public class FSImage implements Closeable {
   void reloadFromImageFile(File file, FSNamesystem target) throws IOException {
     target.clear();
     LOG.debug("Reloading namespace from " + file);
-    loadFSImage(file, target, null);
+    loadFSImage(file, target, null, false);
   }
 
   /**
@@ -603,7 +603,8 @@ public class FSImage implements Closeable {
       // otherwise we can load from both IMAGE and IMAGE_ROLLBACK
       nnfs = EnumSet.of(NameNodeFile.IMAGE, NameNodeFile.IMAGE_ROLLBACK);
     }
-    final FSImageStorageInspector inspector = storage.readAndInspectDirs(nnfs);
+    final FSImageStorageInspector inspector = storage
+        .readAndInspectDirs(nnfs, startOpt);
 
     isUpgradeFinalized = inspector.isUpgradeFinalized();
     List<FSImageFile> imageFiles = inspector.getLatestImages();
@@ -659,7 +660,7 @@ public class FSImage implements Closeable {
     for (int i = 0; i < imageFiles.size(); i++) {
       try {
         imageFile = imageFiles.get(i);
-        loadFSImageFile(target, recovery, imageFile);
+        loadFSImageFile(target, recovery, imageFile, startOpt);
         break;
       } catch (IOException ioe) {
         LOG.error("Failed to load image from " + imageFile, ioe);
@@ -712,16 +713,18 @@ public class FSImage implements Closeable {
   }
 
   void loadFSImageFile(FSNamesystem target, MetaRecoveryContext recovery,
-      FSImageFile imageFile) throws IOException {
+      FSImageFile imageFile, StartupOption startupOption) throws IOException {
     LOG.debug("Planning to load image :\n" + imageFile);
     StorageDirectory sdForProperties = imageFile.sd;
-    storage.readProperties(sdForProperties);
+    storage.readProperties(sdForProperties, startupOption);
 
     if (NameNodeLayoutVersion.supports(
         LayoutVersion.Feature.TXID_BASED_LAYOUT, getLayoutVersion())) {
       // For txid-based layout, we should have a .md5 file
       // next to the image file
-      loadFSImage(imageFile.getFile(), target, recovery);
+      boolean isRollingRollback = RollingUpgradeStartupOption.ROLLBACK
+          .matches(startupOption);
+      loadFSImage(imageFile.getFile(), target, recovery, isRollingRollback);
     } else if (NameNodeLayoutVersion.supports(
         LayoutVersion.Feature.FSIMAGE_CHECKSUM, getLayoutVersion())) {
       // In 0.22, we have the checksum stored in the VERSION file.
@@ -733,10 +736,11 @@ public class FSImage implements Closeable {
             NNStorage.DEPRECATED_MESSAGE_DIGEST_PROPERTY +
             " not set for storage directory " + sdForProperties.getRoot());
       }
-      loadFSImage(imageFile.getFile(), new MD5Hash(md5), target, recovery);
+      loadFSImage(imageFile.getFile(), new MD5Hash(md5), target, recovery,
+          false);
     } else {
       // We don't have any record of the md5sum
-      loadFSImage(imageFile.getFile(), null, target, recovery);
+      loadFSImage(imageFile.getFile(), null, target, recovery, false);
     }
   }
 
@@ -894,13 +898,15 @@ public class FSImage implements Closeable {
    * it against the MD5 sum stored in its associated .md5 file.
    */
   private void loadFSImage(File imageFile, FSNamesystem target,
-      MetaRecoveryContext recovery) throws IOException {
+      MetaRecoveryContext recovery, boolean requireSameLayoutVersion)
+      throws IOException {
     MD5Hash expectedMD5 = MD5FileUtils.readStoredMd5ForFile(imageFile);
     if (expectedMD5 == null) {
       throw new IOException("No MD5 file found corresponding to image file "
           + imageFile);
     }
-    loadFSImage(imageFile, expectedMD5, target, recovery);
+    loadFSImage(imageFile, expectedMD5, target, recovery,
+        requireSameLayoutVersion);
   }
   
   /**
@@ -908,13 +914,14 @@ public class FSImage implements Closeable {
    * filenames and blocks.
    */
   private void loadFSImage(File curFile, MD5Hash expectedMd5,
-      FSNamesystem target, MetaRecoveryContext recovery) throws IOException {
+      FSNamesystem target, MetaRecoveryContext recovery,
+      boolean requireSameLayoutVersion) throws IOException {
     // BlockPoolId is required when the FsImageLoader loads the rolling upgrade
     // information. Make sure the ID is properly set.
     target.setBlockPoolId(this.getBlockPoolID());
 
     FSImageFormat.LoaderDelegator loader = FSImageFormat.newLoader(conf, target);
-    loader.load(curFile);
+    loader.load(curFile, requireSameLayoutVersion);
 
     // Check that the image digest we loaded matches up with what
     // we expected
@@ -1033,7 +1040,7 @@ public class FSImage implements Closeable {
   }
 
   /**
-   * @see #saveNamespace(FSNamesystem, Canceler)
+   * @see #saveNamespace(FSNamesystem, NameNodeFile, Canceler)
    */
   public synchronized void saveNamespace(FSNamesystem source)
       throws IOException {
@@ -1072,7 +1079,7 @@ public class FSImage implements Closeable {
   }
 
   /**
-   * @see #saveFSImageInAllDirs(FSNamesystem, long, Canceler)
+   * @see #saveFSImageInAllDirs(FSNamesystem, NameNodeFile, long, Canceler)
    */
   protected synchronized void saveFSImageInAllDirs(FSNamesystem source, long txid)
       throws IOException {

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

@@ -210,7 +210,8 @@ public class FSImageFormat {
       return impl.getLoadedImageTxId();
     }
 
-    public void load(File file) throws IOException {
+    public void load(File file, boolean requireSameLayoutVersion)
+        throws IOException {
       Preconditions.checkState(impl == null, "Image already loaded!");
 
       FileInputStream is = null;
@@ -220,7 +221,7 @@ public class FSImageFormat {
         IOUtils.readFully(is, magic, 0, magic.length);
         if (Arrays.equals(magic, FSImageUtil.MAGIC_HEADER)) {
           FSImageFormatProtobuf.Loader loader = new FSImageFormatProtobuf.Loader(
-              conf, fsn);
+              conf, fsn, requireSameLayoutVersion);
           impl = loader;
           loader.load(file);
         } else {
@@ -228,7 +229,6 @@ public class FSImageFormat {
           impl = loader;
           loader.load(file);
         }
-
       } finally {
         IOUtils.cleanup(LOG, is);
       }

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

@@ -42,9 +42,12 @@ import org.apache.commons.logging.Log;
 import org.apache.commons.logging.LogFactory;
 import org.apache.hadoop.classification.InterfaceAudience;
 import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.hdfs.protocol.HdfsConstants;
 import org.apache.hadoop.hdfs.protocol.proto.ClientNamenodeProtocolProtos.CacheDirectiveInfoProto;
 import org.apache.hadoop.hdfs.protocol.proto.ClientNamenodeProtocolProtos.CachePoolInfoProto;
 import org.apache.hadoop.hdfs.security.token.delegation.DelegationTokenSecretManager;
+import org.apache.hadoop.hdfs.server.common.InconsistentFSStateException;
+import org.apache.hadoop.hdfs.server.common.IncorrectVersionException;
 import org.apache.hadoop.hdfs.server.namenode.FsImageProto.CacheManagerSection;
 import org.apache.hadoop.hdfs.server.namenode.FsImageProto.FileSummary;
 import org.apache.hadoop.hdfs.server.namenode.FsImageProto.NameSystemSection;
@@ -139,11 +142,19 @@ public final class FSImageFormatProtobuf {
     private MD5Hash imgDigest;
     /** The transaction ID of the last edit represented by the loaded file */
     private long imgTxId;
+    /**
+     * Whether the image's layout version must be the same with
+     * {@link HdfsConstants#NAMENODE_LAYOUT_VERSION}. This is only set to true
+     * when we're doing (rollingUpgrade rollback).
+     */
+    private final boolean requireSameLayoutVersion;
 
-    Loader(Configuration conf, FSNamesystem fsn) {
+    Loader(Configuration conf, FSNamesystem fsn,
+        boolean requireSameLayoutVersion) {
       this.conf = conf;
       this.fsn = fsn;
       this.ctx = new LoaderContext();
+      this.requireSameLayoutVersion = requireSameLayoutVersion;
     }
 
     @Override
@@ -181,6 +192,12 @@ public final class FSImageFormatProtobuf {
         throw new IOException("Unrecognized file format");
       }
       FileSummary summary = FSImageUtil.loadSummary(raFile);
+      if (requireSameLayoutVersion && summary.getLayoutVersion() !=
+          HdfsConstants.NAMENODE_LAYOUT_VERSION) {
+        throw new IOException("Image version " + summary.getLayoutVersion() +
+            " is not equal to the software version " +
+            HdfsConstants.NAMENODE_LAYOUT_VERSION);
+      }
 
       FileChannel channel = fin.getChannel();
 

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

@@ -998,7 +998,8 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
       MetaRecoveryContext recovery = startOpt.createRecoveryContext();
       final boolean staleImage
           = fsImage.recoverTransitionRead(startOpt, this, recovery);
-      if (RollingUpgradeStartupOption.ROLLBACK.matches(startOpt)) {
+      if (RollingUpgradeStartupOption.ROLLBACK.matches(startOpt) ||
+          RollingUpgradeStartupOption.DOWNGRADE.matches(startOpt)) {
         rollingUpgradeInfo = null;
       }
       final boolean needToSave = staleImage && !haEnabled && !isRollingUpgrade(); 
@@ -1008,6 +1009,8 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
       if (needToSave) {
         fsImage.saveNamespace(this);
       } else {
+        updateStorageVersionForRollingUpgrade(fsImage.getLayoutVersion(),
+            startOpt);
         // No need to save, so mark the phase done.
         StartupProgress prog = NameNode.getStartupProgress();
         prog.beginPhase(Phase.SAVING_CHECKPOINT);
@@ -1029,6 +1032,18 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
     imageLoadComplete();
   }
 
+  private void updateStorageVersionForRollingUpgrade(final long layoutVersion,
+      StartupOption startOpt) throws IOException {
+    boolean rollingStarted = RollingUpgradeStartupOption.STARTED
+        .matches(startOpt) && layoutVersion > HdfsConstants
+        .NAMENODE_LAYOUT_VERSION;
+    boolean rollingRollback = RollingUpgradeStartupOption.ROLLBACK
+        .matches(startOpt);
+    if (rollingRollback || rollingStarted) {
+      fsImage.updateStorageVersion();
+    }
+  }
+
   private void startSecretManager() {
     if (dtSecretManager != null) {
       try {

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

@@ -39,9 +39,11 @@ import org.apache.hadoop.fs.FileUtil;
 import org.apache.hadoop.hdfs.DFSUtil;
 import org.apache.hadoop.hdfs.protocol.HdfsConstants;
 import org.apache.hadoop.hdfs.protocol.LayoutVersion;
+import org.apache.hadoop.hdfs.server.common.HdfsServerConstants;
 import org.apache.hadoop.hdfs.server.common.HdfsServerConstants.NodeType;
 import org.apache.hadoop.hdfs.server.common.HdfsServerConstants.StartupOption;
 import org.apache.hadoop.hdfs.server.common.InconsistentFSStateException;
+import org.apache.hadoop.hdfs.server.common.IncorrectVersionException;
 import org.apache.hadoop.hdfs.server.common.Storage;
 import org.apache.hadoop.hdfs.server.common.StorageErrorReporter;
 import org.apache.hadoop.hdfs.server.common.Util;
@@ -620,6 +622,23 @@ public class NNStorage extends Storage implements Closeable,
     setDeprecatedPropertiesForUpgrade(props);
   }
 
+  void readProperties(StorageDirectory sd, StartupOption startupOption)
+      throws IOException {
+    Properties props = readPropertiesFile(sd.getVersionFile());
+    if (HdfsServerConstants.RollingUpgradeStartupOption.ROLLBACK.matches
+        (startupOption)) {
+      int lv = Integer.parseInt(getProperty(props, sd, "layoutVersion"));
+      if (lv > getServiceLayoutVersion()) {
+        // we should not use a newer version for rollingUpgrade rollback
+        throw new IncorrectVersionException(getServiceLayoutVersion(), lv,
+            "storage directory " + sd.getRoot().getAbsolutePath());
+      }
+      props.setProperty("layoutVersion",
+          Integer.toString(HdfsConstants.NAMENODE_LAYOUT_VERSION));
+    }
+    setFieldsFromProperties(props, sd);
+  }
+
   /**
    * Pull any properties out of the VERSION file that are from older
    * versions of HDFS and only necessary during upgrade.
@@ -1002,8 +1021,8 @@ public class NNStorage extends Storage implements Closeable,
    * <b>Note:</b> this can mutate the storage info fields (ctime, version, etc).
    * @throws IOException if no valid storage dirs are found or no valid layout version
    */
-  FSImageStorageInspector readAndInspectDirs(EnumSet<NameNodeFile> fileTypes)
+  FSImageStorageInspector readAndInspectDirs(EnumSet<NameNodeFile> fileTypes,
-      throws IOException {
+      StartupOption startupOption) throws IOException {
     Integer layoutVersion = null;
     boolean multipleLV = false;
     StringBuilder layoutVersions = new StringBuilder();
@@ -1016,7 +1035,7 @@ public class NNStorage extends Storage implements Closeable,
         FSImage.LOG.warn("Storage directory " + sd + " contains no VERSION file. Skipping...");
         continue;
       }
-      readProperties(sd); // sets layoutVersion
+      readProperties(sd, startupOption); // sets layoutVersion
       int lv = getLayoutVersion();
       if (layoutVersion == null) {
         layoutVersion = Integer.valueOf(lv);

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

@@ -157,7 +157,7 @@ public class TestFSImageWithSnapshot {
     fsn.writeLock();
     fsn.getFSDirectory().writeLock();
     try {
-      loader.load(imageFile);
+      loader.load(imageFile, false);
       FSImage.updateCountForQuota(
           INodeDirectory.valueOf(fsn.getFSDirectory().getINode("/"), "/"));
     } finally {