HDFS-5504. In HA mode, OP_DELETE_SNAPSHOT is not decrementing the safemode threshold, leads to NN safemode. Contributed by Vinay.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1541773 13f79535-47bb-0310-9956-ffa450edef68

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

@@ -550,6 +550,9 @@ Release 2.3.0 - UNRELEASED
     HDFS-5075. httpfs-config.sh calls out incorrect env script name
     (Timothy St. Clair via stevel)
 
+    HDFS-5504. In HA mode, OP_DELETE_SNAPSHOT is not decrementing the safemode threshold, 
+    leads to NN safemode. (Vinay via jing9)
+
 Release 2.2.1 - UNRELEASED
 
   INCOMPATIBLE CHANGES

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

@@ -593,7 +593,7 @@ public class FSEditLogLoader {
       fsNamesys.getSnapshotManager().deleteSnapshot(
           deleteSnapshotOp.snapshotRoot, deleteSnapshotOp.snapshotName,
           collectedBlocks, removedINodes);
-      fsNamesys.removeBlocks(collectedBlocks);
+      fsNamesys.removeBlocksAndUpdateSafemodeTotal(collectedBlocks);
       collectedBlocks.clear();
       fsNamesys.dir.removeFromInodeMap(removedINodes);
       removedINodes.clear();

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

@@ -3316,6 +3316,18 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
       return;
     }
     
+    removeBlocksAndUpdateSafemodeTotal(blocks);
+  }
+
+  /**
+   * Removes the blocks from blocksmap and updates the safemode blocks total
+   * 
+   * @param blocks
+   *          An instance of {@link BlocksMapUpdateInfo} which contains a list
+   *          of blocks that need to be removed from blocksMap
+   */
+  void removeBlocksAndUpdateSafemodeTotal(BlocksMapUpdateInfo blocks) {
+    assert hasWriteLock();
     // In the case that we are a Standby tailing edits from the
     // active while in safe-mode, we need to track the total number
     // of blocks and safe blocks in the system.
@@ -3336,9 +3348,9 @@ public class FSNamesystem implements Namesystem, FSClusterStats,
     }
     if (trackBlockCounts) {
       if (LOG.isDebugEnabled()) {
-        LOG.debug("Adjusting safe-mode totals for deletion of " + src + ":" +
-            "decreasing safeBlocks by " + numRemovedSafe +
-            ", totalBlocks by " + numRemovedComplete);
+        LOG.debug("Adjusting safe-mode totals for deletion."
+            + "decreasing safeBlocks by " + numRemovedSafe
+            + ", totalBlocks by " + numRemovedComplete);
       }
       adjustSafeModeBlockTotals(-numRemovedSafe, -numRemovedComplete);
     }

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

@@ -38,6 +38,7 @@ import org.apache.hadoop.hdfs.DFSTestUtil;
 import org.apache.hadoop.hdfs.DFSUtil;
 import org.apache.hadoop.hdfs.DistributedFileSystem;
 import org.apache.hadoop.hdfs.MiniDFSCluster;
+import org.apache.hadoop.hdfs.MiniDFSNNTopology;
 import org.apache.hadoop.hdfs.protocol.HdfsConstants;
 import org.apache.hadoop.hdfs.server.blockmanagement.BlockInfo;
 import org.apache.hadoop.hdfs.server.blockmanagement.BlockManager;
@@ -47,7 +48,10 @@ import org.apache.hadoop.hdfs.server.namenode.INode;
 import org.apache.hadoop.hdfs.server.namenode.INodeDirectory;
 import org.apache.hadoop.hdfs.server.namenode.INodeDirectoryWithQuota;
 import org.apache.hadoop.hdfs.server.namenode.INodeFile;
+import org.apache.hadoop.hdfs.server.namenode.NameNode;
+import org.apache.hadoop.hdfs.server.namenode.NameNodeAdapter;
 import org.apache.hadoop.hdfs.server.namenode.Quota;
+import org.apache.hadoop.hdfs.server.namenode.ha.HATestUtil;
 import org.apache.hadoop.hdfs.server.namenode.snapshot.INodeDirectoryWithSnapshot.DirectoryDiffList;
 import org.apache.hadoop.hdfs.util.ReadOnlyList;
 import org.apache.hadoop.io.IOUtils;
@@ -949,4 +953,54 @@ public class TestSnapshotDeletion {
     psOut.close();
     out.close();
   }
+
+  /*
+   * OP_DELETE_SNAPSHOT edits op was not decrementing the safemode threshold on
+   * restart in HA mode. HDFS-5504
+   */
+  @Test(timeout = 60000)
+  public void testHANNRestartAfterSnapshotDeletion() throws Exception {
+    hdfs.close();
+    cluster.shutdown();
+    conf = new Configuration();
+    cluster = new MiniDFSCluster.Builder(conf)
+        .nnTopology(MiniDFSNNTopology.simpleHATopology()).numDataNodes(1)
+        .build();
+    cluster.transitionToActive(0);
+    // stop the standby namenode
+    NameNode snn = cluster.getNameNode(1);
+    snn.stop();
+
+    hdfs = (DistributedFileSystem) HATestUtil
+        .configureFailoverFs(cluster, conf);
+    Path dir = new Path("/dir");
+    Path subDir = new Path(dir, "sub");
+    hdfs.mkdirs(dir);
+    hdfs.allowSnapshot(dir);
+    for (int i = 0; i < 5; i++) {
+      DFSTestUtil.createFile(hdfs, new Path(subDir, "" + i), 100, (short) 1,
+          1024L);
+    }
+
+    // take snapshot
+    hdfs.createSnapshot(dir, "s0");
+
+    // delete the subdir
+    hdfs.delete(subDir, true);
+
+    // roll the edit log
+    NameNode ann = cluster.getNameNode(0);
+    ann.getRpcServer().rollEditLog();
+
+    hdfs.deleteSnapshot(dir, "s0");
+    // wait for the blocks deletion at namenode
+    Thread.sleep(2000);
+
+    NameNodeAdapter.abortEditLogs(ann);
+    cluster.restartNameNode(0, false);
+    cluster.transitionToActive(0);
+
+    // wait till the cluster becomes active
+    cluster.waitClusterUp();
+  }
 }