HADOOP-8246. Auto-HA: automatically scope znode by nameservice ID. Contributed by Todd Lipcon.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/branches/HDFS-3042@1310919 13f79535-47bb-0310-9956-ffa450edef68

hadoop-common-project/hadoop-common/CHANGES.HDFS-3042.txt

@@ -15,3 +15,5 @@ HADOOP-8245. Fix flakiness in TestZKFailoverController (todd)
 HADOOP-8257. TestZKFailoverControllerStress occasionally fails with Mockito error (todd)
 
 HADOOP-8260. Replace ClientBaseWithFixes with our own modified copy of the class (todd)
+
+HADOOP-8246. Auto-HA: automatically scope znode by nameservice ID (todd)

hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/ha/ZKFailoverController.java

@@ -80,8 +80,6 @@ public abstract class ZKFailoverController implements Tool {
 
   private HAServiceTarget localTarget;
 
-  private String parentZnode;
-
   private State lastHealthState = State.INITIALIZING;
 
   /** Set if a fatal error occurs */
@@ -95,10 +93,18 @@ public abstract class ZKFailoverController implements Tool {
   
 
   protected abstract byte[] targetToData(HAServiceTarget target);
-  protected abstract HAServiceTarget getLocalTarget();  
+  protected abstract HAServiceTarget getLocalTarget();
   protected abstract HAServiceTarget dataToTarget(byte[] data);
   protected abstract void loginAsFCUser() throws IOException;
 
+  /**
+   * Return the name of a znode inside the configured parent znode in which
+   * the ZKFC will do all of its work. This is so that multiple federated
+   * nameservices can run on the same ZK quorum without having to manually
+   * configure them to separate subdirectories.
+   */
+  protected abstract String getScopeInsideParentNode();
+
   @Override
   public Configuration getConf() {
     return conf;
@@ -204,6 +210,7 @@ public abstract class ZKFailoverController implements Tool {
   }
 
   private boolean confirmFormat() {
+    String parentZnode = getParentZnode();
     System.err.println(
         "===============================================\n" +
         "The configured parent znode " + parentZnode + " already exists.\n" +
@@ -234,9 +241,6 @@ public abstract class ZKFailoverController implements Tool {
     String zkQuorum = conf.get(ZK_QUORUM_KEY);
     int zkTimeout = conf.getInt(ZK_SESSION_TIMEOUT_KEY,
         ZK_SESSION_TIMEOUT_DEFAULT);
-    parentZnode = conf.get(ZK_PARENT_ZNODE_KEY,
-        ZK_PARENT_ZNODE_DEFAULT);
-    
     // Parse ACLs from configuration.
     String zkAclConf = conf.get(ZK_ACL_KEY, ZK_ACL_DEFAULT);
     zkAclConf = HAZKUtil.resolveConfIndirection(zkAclConf);
@@ -264,10 +268,19 @@ public abstract class ZKFailoverController implements Tool {
     
 
     elector = new ActiveStandbyElector(zkQuorum,
-        zkTimeout, parentZnode, zkAcls, zkAuths,
+        zkTimeout, getParentZnode(), zkAcls, zkAuths,
         new ElectorCallbacks());
   }
   
+  private String getParentZnode() {
+    String znode = conf.get(ZK_PARENT_ZNODE_KEY,
+        ZK_PARENT_ZNODE_DEFAULT);
+    if (!znode.endsWith("/")) {
+      znode += "/";
+    }
+    return znode + getScopeInsideParentNode();
+  }
+
   private synchronized void mainLoop() throws InterruptedException {
     while (fatalError == null) {
       wait();

hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/ha/MiniZKFCCluster.java

@@ -187,8 +187,7 @@ public class MiniZKFCCluster {
       throws NoNodeException {
     Stat stat = new Stat();
     byte[] data = zks.getZKDatabase().getData(
-        ZKFailoverController.ZK_PARENT_ZNODE_DEFAULT + "/" +
-        ActiveStandbyElector.LOCK_FILENAME, stat, null);
+        DummyZKFC.LOCK_ZNODE, stat, null);
     
     assertArrayEquals(Ints.toByteArray(svcs[idx].index), data);
     long session = stat.getEphemeralOwner();
@@ -206,7 +205,7 @@ public class MiniZKFCCluster {
       throws Exception {
     DummyHAService svc = idx == null ? null : svcs[idx];
     ActiveStandbyElectorTestUtil.waitForActiveLockData(ctx, zks,
-        ZKFailoverController.ZK_PARENT_ZNODE_DEFAULT,
+        DummyZKFC.SCOPED_PARENT_ZNODE,
         (idx == null) ? null : Ints.toByteArray(svc.index));
   }
   
@@ -255,6 +254,12 @@ public class MiniZKFCCluster {
   }
   
   static class DummyZKFC extends ZKFailoverController {
+    private static final String DUMMY_CLUSTER = "dummy-cluster";
+    public static final String SCOPED_PARENT_ZNODE =
+      ZKFailoverController.ZK_PARENT_ZNODE_DEFAULT + "/" +
+      DUMMY_CLUSTER;
+    private static final String LOCK_ZNODE = 
+      SCOPED_PARENT_ZNODE + "/" + ActiveStandbyElector.LOCK_FILENAME;
     private final DummyHAService localTarget;
     
     public DummyZKFC(DummyHAService localTarget) {
@@ -280,5 +285,10 @@ public class MiniZKFCCluster {
     @Override
     protected void loginAsFCUser() throws IOException {
     }
+
+    @Override
+    protected String getScopeInsideParentNode() {
+      return DUMMY_CLUSTER;
+    }
   }
 }

hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/ha/TestZKFailoverController.java

@@ -94,6 +94,40 @@ public class TestZKFailoverController extends ClientBaseWithFixes {
     assertEquals(0, runFC(svc, "-formatZK", "-force"));
   }
   
+  @Test
+  public void testFormatOneClusterLeavesOtherClustersAlone() throws Exception {
+    DummyHAService svc = cluster.getService(1);
+
+    DummyZKFC zkfcInOtherCluster = new DummyZKFC(cluster.getService(1)) {
+      @Override
+      protected String getScopeInsideParentNode() {
+        return "other-scope";
+      }
+    };
+    zkfcInOtherCluster.setConf(conf);
+    
+    // Run without formatting the base dir,
+    // should barf
+    assertEquals(ZKFailoverController.ERR_CODE_NO_PARENT_ZNODE,
+        runFC(svc));
+
+    // Format the base dir, should succeed
+    assertEquals(0, runFC(svc, "-formatZK"));
+    
+    // Run the other cluster without formatting, should barf because
+    // it uses a different parent znode
+    assertEquals(ZKFailoverController.ERR_CODE_NO_PARENT_ZNODE,
+        zkfcInOtherCluster.run(new String[]{}));
+    
+    // Should succeed in formatting the second cluster
+    assertEquals(0, zkfcInOtherCluster.run(new String[]{"-formatZK"}));
+
+    // But should not have deleted the original base node from the first
+    // cluster
+    assertEquals(ZKFailoverController.ERR_CODE_FORMAT_DENIED,
+        runFC(svc, "-formatZK", "-nonInteractive"));
+  }
+  
   /**
    * Test that, if ACLs are specified in the configuration, that
    * it sets the ACLs when formatting the parent node.

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

@@ -117,6 +117,11 @@ public class DFSZKFailoverController extends ZKFailoverController {
         DFS_NAMENODE_USER_NAME_KEY, socAddr.getHostName());
   }
   
+  @Override
+  protected String getScopeInsideParentNode() {
+    return localTarget.getNameServiceId();
+  }
+
   public static void main(String args[])
       throws Exception {
     System.exit(ToolRunner.run(