HDFS-6295. Add decommissioning state and node state filtering to dfsadmin. Contributed by Andrew Wang.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1592438 13f79535-47bb-0310-9956-ffa450edef68

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

@@ -340,6 +340,9 @@ Release 2.5.0 - UNRELEASED
     HDFS-6304. Consolidate the logic of path resolution in FSDirectory.
     (wheat9)
 
+    HDFS-6295. Add "decommissioning" state and node state filtering to
+    dfsadmin. (wang)
+
   OPTIMIZATIONS
 
     HDFS-6214. Webhdfs has poor throughput for files >2GB (daryn)

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/protocol/HdfsConstants.java

@@ -104,7 +104,7 @@ public class HdfsConstants {
 
   // type of the datanode report
   public static enum DatanodeReportType {
-    ALL, LIVE, DEAD
+    ALL, LIVE, DEAD, DECOMMISSIONING
   }
 
   // An invalid transaction ID that will never be seen in a real namesystem.

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/protocolPB/PBHelper.java

@@ -1414,6 +1414,7 @@ public class PBHelper {
     case ALL: return DatanodeReportTypeProto.ALL;
     case LIVE: return DatanodeReportTypeProto.LIVE;
     case DEAD: return DatanodeReportTypeProto.DEAD;
+    case DECOMMISSIONING: return DatanodeReportTypeProto.DECOMMISSIONING;
     default: 
       throw new IllegalArgumentException("Unexpected data type report:" + t);
     }
@@ -1425,6 +1426,7 @@ public class PBHelper {
     case ALL: return DatanodeReportType.ALL;
     case LIVE: return DatanodeReportType.LIVE;
     case DEAD: return DatanodeReportType.DEAD;
+    case DECOMMISSIONING: return DatanodeReportType.DECOMMISSIONING;
     default: 
       throw new IllegalArgumentException("Unexpected data type report:" + t);
     }

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

@@ -1239,10 +1239,15 @@ public class DatanodeManager {
   /** For generating datanode reports */
   public List<DatanodeDescriptor> getDatanodeListForReport(
       final DatanodeReportType type) {
-    boolean listLiveNodes = type == DatanodeReportType.ALL ||
-                            type == DatanodeReportType.LIVE;
-    boolean listDeadNodes = type == DatanodeReportType.ALL ||
-                            type == DatanodeReportType.DEAD;
+    final boolean listLiveNodes =
+        type == DatanodeReportType.ALL ||
+        type == DatanodeReportType.LIVE;
+    final boolean listDeadNodes =
+        type == DatanodeReportType.ALL ||
+        type == DatanodeReportType.DEAD;
+    final boolean listDecommissioningNodes =
+        type == DatanodeReportType.ALL ||
+        type == DatanodeReportType.DECOMMISSIONING;
 
     ArrayList<DatanodeDescriptor> nodes;
     final HostFileManager.HostSet foundNodes = new HostFileManager.HostSet();
@@ -1253,7 +1258,10 @@ public class DatanodeManager {
       nodes = new ArrayList<DatanodeDescriptor>(datanodeMap.size());
       for (DatanodeDescriptor dn : datanodeMap.values()) {
         final boolean isDead = isDatanodeDead(dn);
-        if ((listLiveNodes && !isDead) || (listDeadNodes && isDead)) {
+        final boolean isDecommissioning = dn.isDecommissionInProgress();
+        if ((listLiveNodes && !isDead) ||
+            (listDeadNodes && isDead) ||
+            (listDecommissioningNodes && isDecommissioning)) {
             nodes.add(dn);
         }
         foundNodes.add(HostFileManager.resolvedAddressFromDatanodeID(dn));

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/tools/DFSAdmin.java

@@ -25,6 +25,7 @@ import java.net.URI;
 import java.net.URL;
 import java.security.PrivilegedExceptionAction;
 import java.util.ArrayList;
+import java.util.Arrays;
 import java.util.Collections;
 import java.util.HashMap;
 import java.util.List;
@@ -371,65 +372,96 @@ public class DFSAdmin extends FsShell {
    * Gives a report on how the FileSystem is doing.
    * @exception IOException if the filesystem does not exist.
    */
-  public void report() throws IOException {
-      DistributedFileSystem dfs = getDFS();
-      FsStatus ds = dfs.getStatus();
-      long capacity = ds.getCapacity();
-      long used = ds.getUsed();
-      long remaining = ds.getRemaining();
-      long presentCapacity = used + remaining;
-      boolean mode = dfs.setSafeMode(HdfsConstants.SafeModeAction.SAFEMODE_GET);
-      if (mode) {
-        System.out.println("Safe mode is ON");
-      }
-      System.out.println("Configured Capacity: " + capacity
-                         + " (" + StringUtils.byteDesc(capacity) + ")");
-      System.out.println("Present Capacity: " + presentCapacity
-          + " (" + StringUtils.byteDesc(presentCapacity) + ")");
-      System.out.println("DFS Remaining: " + remaining
-          + " (" + StringUtils.byteDesc(remaining) + ")");
-      System.out.println("DFS Used: " + used
-                         + " (" + StringUtils.byteDesc(used) + ")");
-      System.out.println("DFS Used%: "
-          + StringUtils.formatPercent(used/(double)presentCapacity, 2));
-      
-      /* These counts are not always upto date. They are updated after  
-       * iteration of an internal list. Should be updated in a few seconds to 
-       * minutes. Use "-metaSave" to list of all such blocks and accurate 
-       * counts.
-       */
-      System.out.println("Under replicated blocks: " + 
-                         dfs.getUnderReplicatedBlocksCount());
-      System.out.println("Blocks with corrupt replicas: " + 
-                         dfs.getCorruptBlocksCount());
-      System.out.println("Missing blocks: " + 
-                         dfs.getMissingBlocksCount());
-                           
-      System.out.println();
+  public void report(String[] argv, int i) throws IOException {
+    DistributedFileSystem dfs = getDFS();
+    FsStatus ds = dfs.getStatus();
+    long capacity = ds.getCapacity();
+    long used = ds.getUsed();
+    long remaining = ds.getRemaining();
+    long presentCapacity = used + remaining;
+    boolean mode = dfs.setSafeMode(HdfsConstants.SafeModeAction.SAFEMODE_GET);
+    if (mode) {
+      System.out.println("Safe mode is ON");
+    }
+    System.out.println("Configured Capacity: " + capacity
+                       + " (" + StringUtils.byteDesc(capacity) + ")");
+    System.out.println("Present Capacity: " + presentCapacity
+        + " (" + StringUtils.byteDesc(presentCapacity) + ")");
+    System.out.println("DFS Remaining: " + remaining
+        + " (" + StringUtils.byteDesc(remaining) + ")");
+    System.out.println("DFS Used: " + used
+                       + " (" + StringUtils.byteDesc(used) + ")");
+    System.out.println("DFS Used%: "
+        + StringUtils.formatPercent(used/(double)presentCapacity, 2));
+    
+    /* These counts are not always upto date. They are updated after  
+     * iteration of an internal list. Should be updated in a few seconds to 
+     * minutes. Use "-metaSave" to list of all such blocks and accurate 
+     * counts.
+     */
+    System.out.println("Under replicated blocks: " + 
+                       dfs.getUnderReplicatedBlocksCount());
+    System.out.println("Blocks with corrupt replicas: " + 
+                       dfs.getCorruptBlocksCount());
+    System.out.println("Missing blocks: " + 
+                       dfs.getMissingBlocksCount());
 
-      System.out.println("-------------------------------------------------");
-      
+    System.out.println();
+
+    System.out.println("-------------------------------------------------");
+    
+    // Parse arguments for filtering the node list
+    List<String> args = Arrays.asList(argv);
+    // Truncate already handled arguments before parsing report()-specific ones
+    args = new ArrayList<String>(args.subList(i, args.size()));
+    final boolean listLive = StringUtils.popOption("-live", args);
+    final boolean listDead = StringUtils.popOption("-dead", args);
+    final boolean listDecommissioning =
+        StringUtils.popOption("-decommissioning", args);
+
+    // If no filter flags are found, then list all DN types
+    boolean listAll = (!listLive && !listDead && !listDecommissioning);
+
+    if (listAll || listLive) {
       DatanodeInfo[] live = dfs.getDataNodeStats(DatanodeReportType.LIVE);
-      DatanodeInfo[] dead = dfs.getDataNodeStats(DatanodeReportType.DEAD);
-      System.out.println("Datanodes available: " + live.length +
-                         " (" + (live.length + dead.length) + " total, " + 
-                         dead.length + " dead)\n");
-      
-      if(live.length > 0) {
-        System.out.println("Live datanodes:");
+      if (live.length > 0 || listLive) {
+        System.out.println("Live datanodes (" + live.length + "):\n");
+      }
+      if (live.length > 0) {
         for (DatanodeInfo dn : live) {
           System.out.println(dn.getDatanodeReport());
           System.out.println();
         }
       }
-      
-      if(dead.length > 0) {
-        System.out.println("Dead datanodes:");
+    }
+
+    if (listAll || listDead) {
+      DatanodeInfo[] dead = dfs.getDataNodeStats(DatanodeReportType.DEAD);
+      if (dead.length > 0 || listDead) {
+        System.out.println("Dead datanodes (" + dead.length + "):\n");
+      }
+      if (dead.length > 0) {
         for (DatanodeInfo dn : dead) {
           System.out.println(dn.getDatanodeReport());
           System.out.println();
-        }     
+        }
+      }
+    }
+
+    if (listAll || listDecommissioning) {
+      DatanodeInfo[] decom =
+          dfs.getDataNodeStats(DatanodeReportType.DECOMMISSIONING);
+      if (decom.length > 0 || listDecommissioning) {
+        System.out.println("Decommissioning datanodes (" + decom.length
+            + "):\n");
+      }
+      if (decom.length > 0) {
+        for (DatanodeInfo dn : decom) {
+          System.out.println(dn.getDatanodeReport());
+          System.out.println();
+        }
       }
+    }
   }
 
   /**
@@ -639,7 +671,9 @@ public class DFSAdmin extends FsShell {
   private void printHelp(String cmd) {
     String summary = "hadoop dfsadmin performs DFS administrative commands.\n" +
       "The full syntax is: \n\n" +
-      "hadoop dfsadmin [-report] [-safemode <enter | leave | get | wait>]\n" +
+      "hadoop dfsadmin\n" +
+      "\t[-report [-live] [-dead] [-decommissioning]]\n" +
+      "\t[-safemode <enter | leave | get | wait>]\n" +
       "\t[-saveNamespace]\n" +
       "\t[-rollEdits]\n" +
       "\t[-restoreFailedStorage true|false|check]\n" +
@@ -665,8 +699,11 @@ public class DFSAdmin extends FsShell {
       "\t[-getDatanodeInfo <datanode_host:ipc_port>\n" +
       "\t[-help [cmd]]\n";
 
-    String report ="-report: \tReports basic filesystem information and statistics.\n";
-        
+    String report ="-report [-live] [-dead] [-decommissioning]:\n" +
+      "\tReports basic filesystem information and statistics.\n" +
+      "\tOptional flags may be used to filter the list of displayed DNs.\n";
+    
+
     String safemode = "-safemode <enter|leave|get|wait>:  Safe mode maintenance command.\n" + 
       "\t\tSafe mode is a Namenode state in which it\n" +
       "\t\t\t1.  does not accept changes to the name space (read-only)\n" +
@@ -1069,7 +1106,7 @@ public class DFSAdmin extends FsShell {
   private static void printUsage(String cmd) {
     if ("-report".equals(cmd)) {
       System.err.println("Usage: java DFSAdmin"
-                         + " [-report]");
+          + " [-report] [-live] [-dead] [-decommissioning]");
     } else if ("-safemode".equals(cmd)) {
       System.err.println("Usage: java DFSAdmin"
           + " [-safemode enter | leave | get | wait]");
@@ -1210,7 +1247,7 @@ public class DFSAdmin extends FsShell {
         return exitCode;
       }
     } else if ("-report".equals(cmd)) {
-      if (argv.length != 1) {
+      if (argv.length < 1) {
         printUsage(cmd);
         return exitCode;
       }
@@ -1312,7 +1349,7 @@ public class DFSAdmin extends FsShell {
     exitCode = 0;
     try {
       if ("-report".equals(cmd)) {
-        report();
+        report(argv, i);
       } else if ("-safemode".equals(cmd)) {
         setSafeMode(argv, i);
       } else if ("-allowSnapshot".equalsIgnoreCase(cmd)) {

hadoop-hdfs-project/hadoop-hdfs/src/main/proto/ClientNamenodeProtocol.proto

@@ -267,6 +267,7 @@ enum DatanodeReportTypeProto {  // type of the datanode report
   ALL = 1;
   LIVE = 2;
   DEAD = 3;
+  DECOMMISSIONING = 4;
 }
 
 message GetDatanodeReportRequestProto {

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

@@ -21,12 +21,15 @@ import static org.junit.Assert.assertEquals;
 import static org.junit.Assert.assertTrue;
 
 import java.io.IOException;
+import java.io.PrintStream;
 import java.net.InetSocketAddress;
 import java.util.ArrayList;
+import java.util.Arrays;
 import java.util.Iterator;
 import java.util.List;
 import java.util.Random;
 
+import org.apache.commons.io.output.ByteArrayOutputStream;
 import org.apache.hadoop.conf.Configuration;
 import org.apache.hadoop.fs.CommonConfigurationKeys;
 import org.apache.hadoop.fs.FSDataOutputStream;
@@ -35,12 +38,15 @@ import org.apache.hadoop.fs.Path;
 import org.apache.hadoop.hdfs.DFSClient;
 import org.apache.hadoop.hdfs.DFSConfigKeys;
 import org.apache.hadoop.hdfs.DFSTestUtil;
+import org.apache.hadoop.hdfs.DistributedFileSystem;
 import org.apache.hadoop.hdfs.HdfsConfiguration;
 import org.apache.hadoop.hdfs.MiniDFSCluster;
+import org.apache.hadoop.hdfs.protocol.DatanodeID;
 import org.apache.hadoop.hdfs.protocol.DatanodeInfo;
 import org.apache.hadoop.hdfs.protocol.HdfsConstants.DatanodeReportType;
 import org.apache.hadoop.hdfs.server.blockmanagement.DatanodeDescriptor;
 import org.apache.hadoop.hdfs.server.blockmanagement.DatanodeManager;
+import org.apache.hadoop.hdfs.tools.DFSAdmin;
 import org.junit.AfterClass;
 import org.junit.BeforeClass;
 import org.junit.Test;
@@ -167,7 +173,51 @@ public class TestDecommissioningStatus {
     assertEquals(decommNode.decommissioningStatus
         .getUnderReplicatedInOpenFiles(), expectedUnderRepInOpenFiles);
   }
-  
+
+  private void checkDFSAdminDecommissionStatus(
+      List<DatanodeDescriptor> expectedDecomm, DistributedFileSystem dfs,
+      DFSAdmin admin) throws IOException {
+    ByteArrayOutputStream baos = new ByteArrayOutputStream();
+    PrintStream ps = new PrintStream(baos);
+    PrintStream oldOut = System.out;
+    System.setOut(ps);
+    try {
+      // Parse DFSAdmin just to check the count
+      admin.report(new String[] {"-decommissioning"}, 0);
+      String[] lines = baos.toString().split("\n");
+      Integer num = null;
+      int count = 0;
+      for (String line: lines) {
+        if (line.startsWith("Decommissioning datanodes")) {
+          // Pull out the "(num)" and parse it into an int
+          String temp = line.split(" ")[2];
+          num =
+              Integer.parseInt((String) temp.subSequence(1, temp.length() - 2));
+        }
+        if (line.contains("Decommission in progress")) {
+          count++;
+        }
+      }
+      assertTrue("No decommissioning output", num != null);
+      assertEquals("Unexpected number of decomming DNs", expectedDecomm.size(),
+          num.intValue());
+      assertEquals("Unexpected number of decomming DNs", expectedDecomm.size(),
+          count);
+
+      // Check Java API for correct contents
+      List<DatanodeInfo> decomming =
+          new ArrayList<DatanodeInfo>(Arrays.asList(dfs
+              .getDataNodeStats(DatanodeReportType.DECOMMISSIONING)));
+      assertEquals("Unexpected number of decomming DNs", expectedDecomm.size(),
+          decomming.size());
+      for (DatanodeID id : expectedDecomm) {
+        assertTrue("Did not find expected decomming DN " + id,
+            decomming.contains(id));
+      }
+    } finally {
+      System.setOut(oldOut);
+    }
+  }
   /**
    * Tests Decommissioning Status in DFS.
    */
@@ -179,7 +229,8 @@ public class TestDecommissioningStatus {
     DFSClient client = new DFSClient(addr, conf);
     DatanodeInfo[] info = client.datanodeReport(DatanodeReportType.LIVE);
     assertEquals("Number of Datanodes ", 2, info.length);
-    FileSystem fileSys = cluster.getFileSystem();
+    DistributedFileSystem fileSys = cluster.getFileSystem();
+    DFSAdmin admin = new DFSAdmin(cluster.getConfiguration(0));
 
     short replicas = 2;
     //
@@ -205,12 +256,16 @@ public class TestDecommissioningStatus {
         assertEquals(decommissioningNodes.size(), 1);
         DatanodeDescriptor decommNode = decommissioningNodes.get(0);
         checkDecommissionStatus(decommNode, 4, 0, 2);
+        checkDFSAdminDecommissionStatus(decommissioningNodes.subList(0, 1),
+            fileSys, admin);
       } else {
         assertEquals(decommissioningNodes.size(), 2);
         DatanodeDescriptor decommNode1 = decommissioningNodes.get(0);
         DatanodeDescriptor decommNode2 = decommissioningNodes.get(1);
         checkDecommissionStatus(decommNode1, 4, 4, 2);
         checkDecommissionStatus(decommNode2, 4, 4, 2);
+        checkDFSAdminDecommissionStatus(decommissioningNodes.subList(0, 2),
+            fileSys, admin);
       }
     }
     // Call refreshNodes on FSNamesystem with empty exclude file.

hadoop-hdfs-project/hadoop-hdfs/src/test/resources/testHDFSConf.xml

@@ -15220,7 +15220,11 @@
       <comparators>
         <comparator>
           <type>RegexpComparator</type>
-          <expected-output>^-report:( |\t)*Reports basic filesystem information and statistics.( )*</expected-output>
+          <expected-output>^-report \[-live\] \[-dead\] \[-decommissioning\]:(.)*</expected-output>
+        </comparator>
+        <comparator>
+          <type>RegexpComparator</type>
+          <expected-output>^[ \t]*Reports basic filesystem information and statistics.( )*</expected-output>
         </comparator>
       </comparators>
     </test>
@@ -15900,9 +15904,9 @@
           <type>RegexpComparator</type>
           <expected-output>DFS Used\%: [0-9\.]+%</expected-output>
         </comparator>
-	<comparator>
+        <comparator>
           <type>RegexpComparator</type>
-          <expected-output>Datanodes available: [0-9]+ \([0-9]+ total, [0-9]+ dead\)</expected-output>
+          <expected-output>Live datanodes \([0-9]+\):</expected-output>
         </comparator>
         <comparator>
           <type>RegexpComparator</type>
@@ -15930,7 +15934,7 @@
         </comparator>
         <comparator>
           <type>TokenComparator</type>
-          <expected-output>Live datanodes:</expected-output>
+          <expected-output>Live datanodes</expected-output>
         </comparator>
       </comparators>
     </test>
@@ -16018,10 +16022,6 @@
           <type>RegexpComparator</type>
           <expected-output>DFS Used\%: [0-9\.]+%</expected-output>
         </comparator>
-        <comparator>
-          <type>RegexpComparator</type>
-          <expected-output>Datanodes available: [0-9]+ \([0-9]+ total, [0-9]+ dead\)</expected-output>
-        </comparator>
         <comparator>
           <type>RegexpComparator</type>
           <expected-output>Name: [0-9\.:]+ \([-.a-zA-z0-9\.]+\)</expected-output>
@@ -16048,7 +16048,7 @@
         </comparator>
 	<comparator>
           <type>TokenComparator</type>
-          <expected-output>Live datanodes:</expected-output>
+          <expected-output>Live datanodes</expected-output>
         </comparator>
       </comparators>
     </test>