HDFS-7281. Missing block is marked as corrupted block (Ming Ma via Yongjun Zhang)

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

@@ -20,6 +20,9 @@ Trunk (Unreleased)
 
     HDFS-7985. WebHDFS should be always enabled. (Li Lu via wheat9)
 
+    HDFS-7281. Missing block is marked as corrupted block (Ming Ma via
+    Yongjun Zhang)
+ 
   NEW FEATURES
 
     HDFS-3125. Add JournalService to enable Journal Daemon. (suresh)

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockManager.java

@@ -849,7 +849,8 @@ public class BlockManager {
     }
 
     final int numNodes = blocksMap.numNodes(blk);
-    final boolean isCorrupt = numCorruptNodes == numNodes;
+    final boolean isCorrupt = numCorruptNodes != 0 &&
+        numCorruptNodes == numNodes;
     final int numMachines = isCorrupt ? numNodes: numNodes - numCorruptNodes;
     final DatanodeStorageInfo[] machines = new DatanodeStorageInfo[numMachines];
     int j = 0;

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

@@ -531,6 +531,7 @@ public class NamenodeFsck implements DataEncryptionKeyFactory {
     int missing = 0;
     int corrupt = 0;
     long missize = 0;
+    long corruptSize = 0;
     int underReplicatedPerFile = 0;
     int misReplicatedPerFile = 0;
     StringBuilder report = new StringBuilder();
@@ -570,10 +571,11 @@ public class NamenodeFsck implements DataEncryptionKeyFactory {
       // count corrupt blocks
       boolean isCorrupt = lBlk.isCorrupt();
       if (isCorrupt) {
+        res.addCorrupt(block.getNumBytes());
         corrupt++;
-        res.corruptBlocks++;
-        out.print("\n" + path + ": CORRUPT blockpool " + block.getBlockPoolId() + 
-            " block " + block.getBlockName()+"\n");
+        corruptSize += block.getNumBytes();
+        out.print("\n" + path + ": CORRUPT blockpool " +
+            block.getBlockPoolId() + " block " + block.getBlockName() + "\n");
       }
 
       // count minimally replicated blocks
@@ -619,7 +621,11 @@ public class NamenodeFsck implements DataEncryptionKeyFactory {
       // report
       String blkName = block.toString();
       report.append(blockNumber + ". " + blkName + " len=" + block.getNumBytes());
-      if (totalReplicasPerBlock == 0) {
+      if (totalReplicasPerBlock == 0 && !isCorrupt) {
+        // If the block is corrupted, it means all its available replicas are
+        // corrupted. We don't mark it as missing given these available replicas
+        // might still be accessible as the block might be incorrectly marked as
+        // corrupted by client machines.
         report.append(" MISSING!");
         res.addMissing(block.toString(), block.getNumBytes());
         missing++;
@@ -674,9 +680,15 @@ public class NamenodeFsck implements DataEncryptionKeyFactory {
 
     // count corrupt file & move or delete if necessary
     if ((missing > 0) || (corrupt > 0)) {
-      if (!showFiles && (missing > 0)) {
-        out.print("\n" + path + ": MISSING " + missing
-            + " blocks of total size " + missize + " B.");
+      if (!showFiles) {
+        if (missing > 0) {
+          out.print("\n" + path + ": MISSING " + missing
+              + " blocks of total size " + missize + " B.");
+        }
+        if (corrupt > 0) {
+          out.print("\n" + path + ": CORRUPT " + corrupt
+              + " blocks of total size " + corruptSize + " B.");
+        }
       }
       res.corruptFiles++;
       if (isOpen) {
@@ -688,9 +700,16 @@ public class NamenodeFsck implements DataEncryptionKeyFactory {
     }
 
     if (showFiles) {
-      if (missing > 0) {
-        out.print(" MISSING " + missing + " blocks of total size " + missize + " B\n");
-      }  else if (underReplicatedPerFile == 0 && misReplicatedPerFile == 0) {
+      if (missing > 0 || corrupt > 0) {
+        if (missing > 0) {
+          out.print(" MISSING " + missing + " blocks of total size " +
+              missize + " B\n");
+        }
+        if (corrupt > 0) {
+          out.print(" CORRUPT " + corrupt + " blocks of total size " +
+              corruptSize + " B\n");
+        }
+      } else if (underReplicatedPerFile == 0 && misReplicatedPerFile == 0) {
         out.print(" OK\n");
       }
       if (showBlocks) {
@@ -956,6 +975,7 @@ public class NamenodeFsck implements DataEncryptionKeyFactory {
     long missingSize = 0L;
     long corruptFiles = 0L;
     long corruptBlocks = 0L;
+    long corruptSize = 0L;
     long excessiveReplicas = 0L;
     long missingReplicas = 0L;
     long decommissionedReplicas = 0L;
@@ -998,7 +1018,13 @@ public class NamenodeFsck implements DataEncryptionKeyFactory {
       missingIds.add(id);
       missingSize += size;
     }
-    
+
+    /** Add a corrupt block. */
+    void addCorrupt(long size) {
+      corruptBlocks++;
+      corruptSize += size;
+    }
+
     /** Return the actual replication factor. */
     float getReplicationFactor() {
       if (totalBlocks == 0)
@@ -1051,7 +1077,8 @@ public class NamenodeFsck implements DataEncryptionKeyFactory {
                 "\n  MISSING SIZE:\t\t").append(missingSize).append(" B");
           }
           if (corruptBlocks > 0) {
-            res.append("\n  CORRUPT BLOCKS: \t").append(corruptBlocks);
+            res.append("\n  CORRUPT BLOCKS: \t").append(corruptBlocks).append(
+                "\n  CORRUPT SIZE:\t\t").append(corruptSize).append(" B");
           }
         }
         res.append("\n  ********************************");
@@ -1086,7 +1113,8 @@ public class NamenodeFsck implements DataEncryptionKeyFactory {
       }
       res.append("\n Default replication factor:\t").append(replication)
           .append("\n Average block replication:\t").append(
-              getReplicationFactor()).append("\n Corrupt blocks:\t\t").append(
+              getReplicationFactor()).append("\n Missing blocks:\t\t").append(
+              missingIds.size()).append("\n Corrupt blocks:\t\t").append(
               corruptBlocks).append("\n Missing replicas:\t\t").append(
               missingReplicas);
       if (totalReplicas > 0) {

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

@@ -120,7 +120,10 @@ public class TestFsck {
       "ip=/\\d{1,3}\\.\\d{1,3}\\.\\d{1,3}\\.\\d{1,3}\\s" + 
       "cmd=getfileinfo\\ssrc=\\/\\sdst=null\\s" + 
       "perm=null\\s" + "proto=.*");
-  
+
+  static final Pattern numMissingBlocksPattern = Pattern.compile(
+      ".*Missing blocks:\t\t([0123456789]*).*");
+
   static final Pattern numCorruptBlocksPattern = Pattern.compile(
       ".*Corrupt blocks:\t\t([0123456789]*).*");
   
@@ -360,19 +363,27 @@ public class TestFsck {
       // Wait for fsck to discover all the missing blocks
       while (true) {
         outStr = runFsck(conf, 1, false, "/");
+        String numMissing = null;
         String numCorrupt = null;
         for (String line : outStr.split(LINE_SEPARATOR)) {
-          Matcher m = numCorruptBlocksPattern.matcher(line);
+          Matcher m = numMissingBlocksPattern.matcher(line);
+          if (m.matches()) {
+            numMissing = m.group(1);
+          }
+          m = numCorruptBlocksPattern.matcher(line);
           if (m.matches()) {
             numCorrupt = m.group(1);
+          }
+          if (numMissing != null && numCorrupt != null) {
             break;
           }
         }
-        if (numCorrupt == null) {
-          throw new IOException("failed to find number of corrupt " +
-              "blocks in fsck output.");
+        if (numMissing == null || numCorrupt == null) {
+          throw new IOException("failed to find number of missing or corrupt" +
+              " blocks in fsck output.");
         }
-        if (numCorrupt.equals(Integer.toString(totalMissingBlocks))) {
+        if (numMissing.equals(Integer.toString(totalMissingBlocks))) {
+          assertTrue(numCorrupt.equals(Integer.toString(0)));
           assertTrue(outStr.contains(NamenodeFsck.CORRUPT_STATUS));
           break;
         }