YARN-2131. Add a way to format the RMStateStore. (Robert Kanter via kasha)

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1609278 13f79535-47bb-0310-9956-ffa450edef68

hadoop-yarn-project/CHANGES.txt

@@ -24,6 +24,8 @@ Release 2.6.0 - UNRELEASED
 
   NEW FEATURES
 
+    YARN-2131. Add a way to format the RMStateStore. (Robert Kanter via kasha)
+
   IMPROVEMENTS
 
     YARN-2242. Improve exception information on AM launch crashes. (Li Lu 

hadoop-yarn-project/hadoop-yarn/bin/yarn

@@ -61,20 +61,21 @@ HADOOP_LIBEXEC_DIR=${HADOOP_LIBEXEC_DIR:-$DEFAULT_LIBEXEC_DIR}
 function print_usage(){
   echo "Usage: yarn [--config confdir] COMMAND"
   echo "where COMMAND is one of:"
-  echo "  resourcemanager      run the ResourceManager" 
-  echo "  nodemanager          run a nodemanager on each slave" 
-  echo "  timelineserver       run the timeline server"
-  echo "  rmadmin              admin tools" 
-  echo "  version              print the version"
-  echo "  jar <jar>            run a jar file"
-  echo "  application          prints application(s) report/kill application"
-  echo "  applicationattempt   prints applicationattempt(s) report"
-  echo "  container            prints container(s) report"
-  echo "  node                 prints node report(s)"
-  echo "  logs                 dump container logs"
-  echo "  classpath            prints the class path needed to get the"
-  echo "                       Hadoop jar and the required libraries"
-  echo "  daemonlog            get/set the log level for each daemon"
+  echo "  resourcemanager -format deletes the RMStateStore"
+  echo "  resourcemanager         run the ResourceManager"
+  echo "  nodemanager             run a nodemanager on each slave"
+  echo "  timelineserver          run the timeline server"
+  echo "  rmadmin                 admin tools"
+  echo "  version                 print the version"
+  echo "  jar <jar>               run a jar file"
+  echo "  application             prints application(s) report/kill application"
+  echo "  applicationattempt      prints applicationattempt(s) report"
+  echo "  container               prints container(s) report"
+  echo "  node                    prints node report(s)"
+  echo "  logs                    dump container logs"
+  echo "  classpath               prints the class path needed to get the"
+  echo "                          Hadoop jar and the required libraries"
+  echo "  daemonlog               get/set the log level for each daemon"
   echo " or"
   echo "  CLASSNAME            run the class named CLASSNAME"
   echo "Most commands print help when invoked w/o parameters."

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ResourceManager.java

@@ -1035,12 +1035,17 @@ public class ResourceManager extends CompositeService implements Recoverable {
     StringUtils.startupShutdownMessage(ResourceManager.class, argv, LOG);
     try {
       Configuration conf = new YarnConfiguration();
-      ResourceManager resourceManager = new ResourceManager();
-      ShutdownHookManager.get().addShutdownHook(
-        new CompositeServiceShutdownHook(resourceManager),
-        SHUTDOWN_HOOK_PRIORITY);
-      resourceManager.init(conf);
-      resourceManager.start();
+      // If -format, then delete RMStateStore; else startup normally
+      if (argv.length == 1 && argv[0].equals("-format")) {
+        deleteRMStateStore(conf);
+      } else {
+        ResourceManager resourceManager = new ResourceManager();
+        ShutdownHookManager.get().addShutdownHook(
+          new CompositeServiceShutdownHook(resourceManager),
+          SHUTDOWN_HOOK_PRIORITY);
+        resourceManager.init(conf);
+        resourceManager.start();
+      }
     } catch (Throwable t) {
       LOG.fatal("Error starting ResourceManager", t);
       System.exit(-1);
@@ -1077,4 +1082,23 @@ public class ResourceManager extends CompositeService implements Recoverable {
     return conf.getSocketAddr(YarnConfiguration.RM_ADDRESS,
       YarnConfiguration.DEFAULT_RM_ADDRESS, YarnConfiguration.DEFAULT_RM_PORT);
   }
+
+  /**
+   * Deletes the RMStateStore
+   *
+   * @param conf
+   * @throws Exception
+   */
+  private static void deleteRMStateStore(Configuration conf) throws Exception {
+    RMStateStore rmStore = RMStateStoreFactory.getStore(conf);
+    rmStore.init(conf);
+    rmStore.start();
+    try {
+      LOG.info("Deleting ResourceManager state store...");
+      rmStore.deleteStore();
+      LOG.info("State store deleted");
+    } finally {
+      rmStore.stop();
+    }
+  }
 }

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/FileSystemRMStateStore.java

@@ -515,6 +515,13 @@ public class FileSystemRMStateStore extends RMStateStore {
     deleteFile(nodeCreatePath);
   }
 
+  @Override
+  public synchronized void deleteStore() throws IOException {
+    if (fs.exists(rootDirPath)) {
+      fs.delete(rootDirPath, true);
+    }
+  }
+
   private Path getAppDir(Path root, String appId) {
     return getNodePath(root, appId);
   }

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/MemoryRMStateStore.java

@@ -267,4 +267,8 @@ public class MemoryRMStateStore extends RMStateStore {
     return null;
   }
 
+  @Override
+  public void deleteStore() throws Exception {
+  }
+
 }

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/NullRMStateStore.java

@@ -138,4 +138,9 @@ public class NullRMStateStore extends RMStateStore {
     return null;
   }
 
+  @Override
+  public void deleteStore() throws Exception {
+    // Do nothing
+  }
+
 }

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStore.java

@@ -845,4 +845,10 @@ public abstract class RMStateStore extends AbstractService {
       handleStoreEvent(event);
     }
   }
+
+  /**
+   * Derived classes must implement this method to delete the state store
+   * @throws Exception
+   */
+  public abstract void deleteStore() throws Exception;
 }

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/ZKRMStateStore.java

@@ -805,6 +805,13 @@ public class ZKRMStateStore extends RMStateStore {
     }
   }
 
+  @Override
+  public synchronized void deleteStore() throws Exception {
+    if (existsWithRetries(zkRootNodePath, true) != null) {
+      deleteWithRetries(zkRootNodePath, true);
+    }
+  }
+
   // ZK related code
   /**
    * Watcher implementation which forward events to the ZKRMStateStore This
@@ -959,6 +966,29 @@ public class ZKRMStateStore extends RMStateStore {
     }.runWithRetries();
   }
 
+  private void deleteWithRetries(
+      final String path, final boolean watch) throws Exception {
+    new ZKAction<Void>() {
+      @Override
+      Void run() throws KeeperException, InterruptedException {
+        recursiveDeleteWithRetriesHelper(path, watch);
+        return null;
+      }
+    }.runWithRetries();
+  }
+
+  /**
+   * Helper method that deletes znodes recursively
+   */
+  private void recursiveDeleteWithRetriesHelper(String path, boolean watch)
+          throws KeeperException, InterruptedException {
+    List<String> children = zkClient.getChildren(path, watch);
+    for (String child : children) {
+      recursiveDeleteWithRetriesHelper(path + "/" + child, false);
+    }
+    zkClient.delete(path, -1);
+  }
+
   /**
    * Helper class that periodically attempts creating a znode to ensure that
    * this RM continues to be the Active.

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStoreTestBase.java

@@ -515,16 +515,33 @@ public class RMStateStoreTestBase extends ClientBaseWithFixes{
       throws Exception {
     RMStateStore store = stateStoreHelper.getRMStateStore();
     store.setRMDispatcher(new TestDispatcher());
-    // create and store apps
+    ArrayList<RMApp> appList = createAndStoreApps(stateStoreHelper, store, 5);
+
+    for (RMApp app : appList) {
+      // remove the app
+      store.removeApplication(app);
+      // wait for app to be removed.
+      while (true) {
+        if (!stateStoreHelper.appExists(app)) {
+          break;
+        } else {
+          Thread.sleep(100);
+        }
+      }
+    }
+  }
+
+  private ArrayList<RMApp> createAndStoreApps(
+      RMStateStoreHelper stateStoreHelper, RMStateStore store, int numApps)
+      throws Exception {
     ArrayList<RMApp> appList = new ArrayList<RMApp>();
-    int NUM_APPS = 5;
-    for (int i = 0; i < NUM_APPS; i++) {
+    for (int i = 0; i < numApps; i++) {
       ApplicationId appId = ApplicationId.newInstance(1383183338, i);
       RMApp app = storeApp(store, appId, 123456789, 987654321);
       appList.add(app);
     }
 
-    Assert.assertEquals(NUM_APPS, appList.size());
+    Assert.assertEquals(numApps, appList.size());
     for (RMApp app : appList) {
       // wait for app to be stored.
       while (true) {
@@ -535,18 +552,17 @@ public class RMStateStoreTestBase extends ClientBaseWithFixes{
         }
       }
     }
+    return appList;
+  }
 
+  public void testDeleteStore(RMStateStoreHelper stateStoreHelper)
+      throws Exception {
+    RMStateStore store = stateStoreHelper.getRMStateStore();
+    ArrayList<RMApp> appList = createAndStoreApps(stateStoreHelper, store, 5);
+    store.deleteStore();
+    // verify apps deleted
     for (RMApp app : appList) {
-      // remove the app
-      store.removeApplication(app);
-      // wait for app to be removed.
-      while (true) {
-        if (!stateStoreHelper.appExists(app)) {
-          break;
-        } else {
-          Thread.sleep(100);
-        }
-      }
+      Assert.assertFalse(stateStoreHelper.appExists(app));
     }
   }
 

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/TestFSRMStateStore.java

@@ -160,6 +160,7 @@ public class TestFSRMStateStore extends RMStateStoreTestBase {
       testCheckVersion(fsTester);
       testEpoch(fsTester);
       testAppDeletion(fsTester);
+      testDeleteStore(fsTester);
     } finally {
       cluster.shutdown();
     }

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/TestZKRMStateStore.java

@@ -122,6 +122,7 @@ public class TestZKRMStateStore extends RMStateStoreTestBase {
     testCheckVersion(zkTester);
     testEpoch(zkTester);
     testAppDeletion(zkTester);
+    testDeleteStore(zkTester);
   }
 
   private Configuration createHARMConf(

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-site/src/site/apt/YarnCommands.apt.vm

@@ -157,9 +157,17 @@ Usage: yarn [--config confdir] COMMAND
   Start the ResourceManager
 
 -------
-  Usage: yarn resourcemanager
+  Usage: yarn resourcemanager [-format]
 -------
 
+*---------------+--------------+
+|| COMMAND_OPTIONS || Description                   |
+*---------------+--------------+
+| -format | Formats the RMStateStore. This will clear the RMStateStore and is
+|         | useful if past applications are no longer needed. This should be run
+|         | only when the ResourceManager is not running.
+*---------------+--------------+
+
 ** nodemanager
 
   Start the NodeManager