HDFS-5999. Do not create rollback fsimage when it already exists. Contributed by Jing Zhao.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/branches/HDFS-5535@1571096 13f79535-47bb-0310-9956-ffa450edef68

hadoop-hdfs-project/hadoop-hdfs/CHANGES_HDFS-5535.txt

@@ -73,3 +73,5 @@ HDFS-5535 subtasks:
     TestOfflineEditsViewer.  (szetszwo)
 
     HDFS-5994. Fix TestDataNodeRollingUpgrade.  (Arpit Agarwal via szetszwo)
+
+    HDFS-5999. Do not create rollback fsimage when it already exists. (jing9)

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

@@ -728,11 +728,9 @@ public class FSEditLogLoader {
         }
       }
 
-      // save namespace if this is not the second edit transaction
+      // save namespace if there is no rollback image existing
-      // (the first must be OP_START_LOG_SEGMENT)
-      final boolean saveNamespace = totalEdits > 1;
       final long startTime = ((RollingUpgradeOp) op).getTime();
-      fsNamesys.startRollingUpgradeInternal(startTime, saveNamespace);
+      fsNamesys.startRollingUpgradeInternal(startTime, op.txid - 2);
       break;
     }
     case OP_ROLLING_UPGRADE_FINALIZE: {

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

@@ -327,6 +327,21 @@ public class FSImage implements Closeable {
     }
   }
 
+  /**
+   * @return true if there is rollback fsimage (for rolling upgrade) for the
+   * given txid in storage.
+   */
+  boolean hasRollbackFSImage(long txid) {
+    for (StorageDirectory sd : storage.dirIterable(NameNodeDirType.IMAGE)) {
+      final File rollbackImageFile = NNStorage.getStorageFile(sd,
+          NameNodeFile.IMAGE_ROLLBACK, txid);
+      if (rollbackImageFile.exists()) {
+        return true;
+      }
+    }
+    return false;
+  }
+
   void doUpgrade(FSNamesystem target) throws IOException {
     checkUpgrade(target);
 
@@ -1071,7 +1086,7 @@ public class FSImage implements Closeable {
   }
 
   /**
-   * Renames new image
+   * Rename FSImage
    */
   private void renameCheckpoint(long txid, NameNodeFile fromNnf,
       NameNodeFile toNnf, boolean renameMD5) throws IOException {

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

@@ -7147,7 +7147,7 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
     try {
       checkOperation(OperationCategory.WRITE);
       checkNameNodeSafeMode("Failed to start rolling upgrade");
-      startRollingUpgradeInternal(now(), true);
+      startRollingUpgradeInternal(now(), -1);
       getEditLog().logStartRollingUpgrade(rollingUpgradeInfo.getStartTime());
     } finally {
       writeUnlock();
@@ -7165,15 +7165,14 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
    * Update internal state to indicate that a rolling upgrade is in progress.
    * Ootionally create a checkpoint before starting the RU.
    * @param startTime
-   * @param saveNamespace If true then a checkpoint is created before initiating
-   *                   the rolling upgrade.
    */
-  void startRollingUpgradeInternal(long startTime, boolean saveNamespace)
+  void startRollingUpgradeInternal(long startTime, long txid)
       throws IOException {
     checkRollingUpgrade("start rolling upgrade");
     getFSImage().checkUpgrade(this);
 
-    if (saveNamespace) {
+    // if we have not made a rollback image, do it
+    if (txid < 0 || !getFSImage().hasRollbackFSImage(txid)) {
       getFSImage().saveNamespace(this, NameNodeFile.IMAGE_ROLLBACK, null);
       LOG.info("Successfully saved namespace for preparing rolling upgrade.");
     }