YARN-3410. YARN admin should be able to remove individual application records from RMStateStore. (Rohith Sharmaks via wangda)

hadoop-yarn-project/CHANGES.txt

@@ -42,6 +42,9 @@ Release 2.8.0 - UNRELEASED
     YARN-3463. Integrate OrderingPolicy Framework with CapacityScheduler.
     (Craig Welch via wangda)
 
+    YARN-3410. YARN admin should be able to remove individual application 
+    records from RMStateStore. (Rohith Sharmaks via wangda)
+
   IMPROVEMENTS
 
     YARN-1880. Cleanup TestApplicationClientProtocolOnHA

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

@@ -69,8 +69,10 @@ function print_usage(){
   echo "  CLASSNAME                             run the class named CLASSNAME"
   echo " or"
   echo "  where COMMAND is one of:"
-  echo "  resourcemanager -format-state-store   deletes the RMStateStore"
   echo "  resourcemanager                       run the ResourceManager"
+  echo "                                        Use -format-state-store for deleting the RMStateStore."
+  echo "                                        Use -remove-application-from-state-store <appId> for "
+  echo "                                            removing application from RMStateStore."
   echo "  nodemanager                           run a nodemanager on each slave"
   echo "  timelineserver                        run the timeline server"
   echo "  rmadmin                               admin tools"

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

@@ -93,6 +93,7 @@ import org.apache.hadoop.yarn.server.webproxy.AppReportFetcher;
 import org.apache.hadoop.yarn.server.webproxy.ProxyUriUtils;
 import org.apache.hadoop.yarn.server.webproxy.WebAppProxy;
 import org.apache.hadoop.yarn.server.webproxy.WebAppProxyServlet;
+import org.apache.hadoop.yarn.util.ConverterUtils;
 import org.apache.hadoop.yarn.webapp.WebApp;
 import org.apache.hadoop.yarn.webapp.WebApps;
 import org.apache.hadoop.yarn.webapp.WebApps.Builder;
@@ -100,6 +101,7 @@ import org.apache.hadoop.yarn.webapp.util.WebAppUtils;
 
 import java.io.IOException;
 import java.io.InputStream;
+import java.io.PrintStream;
 import java.net.InetSocketAddress;
 import java.security.PrivilegedExceptionAction;
 import java.util.ArrayList;
@@ -1197,8 +1199,15 @@ public class ResourceManager extends CompositeService implements Recoverable {
       GenericOptionsParser hParser = new GenericOptionsParser(conf, argv);
       argv = hParser.getRemainingArgs();
       // If -format-state-store, then delete RMStateStore; else startup normally
-      if (argv.length == 1 && argv[0].equals("-format-state-store")) {
-        deleteRMStateStore(conf);
+      if (argv.length >= 1) {
+        if (argv[0].equals("-format-state-store")) {
+          deleteRMStateStore(conf);
+        } else if (argv[0].equals("-remove-application-from-state-store")
+            && argv.length == 2) {
+          removeApplication(conf, argv[1]);
+        } else {
+          printUsage(System.err);
+        }
       } else {
         ResourceManager resourceManager = new ResourceManager();
         ShutdownHookManager.get().addShutdownHook(
@@ -1275,4 +1284,25 @@ public class ResourceManager extends CompositeService implements Recoverable {
       rmStore.stop();
     }
   }
+
+  private static void removeApplication(Configuration conf, String applicationId)
+      throws Exception {
+    RMStateStore rmStore = RMStateStoreFactory.getStore(conf);
+    rmStore.init(conf);
+    rmStore.start();
+    try {
+      ApplicationId removeAppId = ConverterUtils.toApplicationId(applicationId);
+      LOG.info("Deleting application " + removeAppId + " from state store");
+      rmStore.removeApplication(removeAppId);
+      LOG.info("Application is deleted from state store");
+    } finally {
+      rmStore.stop();
+    }
+  }
+
+  private static void printUsage(PrintStream out) {
+    out.println("Usage: java ResourceManager [-format-state-store]");
+    out.println("                            "
+        + "[-remove-application-from-state-store <appId>]" + "\n");
+  }
 }

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

@@ -559,6 +559,15 @@ public class FileSystemRMStateStore extends RMStateStore {
     }
   }
 
+  @Override
+  public synchronized void removeApplication(ApplicationId removeAppId)
+      throws Exception {
+    Path nodeRemovePath = getAppDir(rmAppRoot, removeAppId);
+    if (existsWithRetries(nodeRemovePath)) {
+      deleteFileWithRetries(nodeRemovePath);
+    }
+  }
+
   private Path getAppDir(Path root, ApplicationId appId) {
     return getNodePath(root, appId.toString());
   }

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

@@ -659,6 +659,18 @@ public class LeveldbRMStateStore extends RMStateStore {
     fs.delete(root, true);
   }
 
+  @Override
+  public synchronized void removeApplication(ApplicationId removeAppId)
+      throws IOException {
+    String appKey = getApplicationNodeKey(removeAppId);
+    LOG.info("Removing state for app " + removeAppId);
+    try {
+      db.delete(bytes(appKey));
+    } catch (DBException e) {
+      throw new IOException(e);
+    }
+  }
+
   @VisibleForTesting
   int getNumEntriesInDatabase() throws IOException {
     int numEntries = 0;

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

@@ -251,4 +251,8 @@ public class MemoryRMStateStore extends RMStateStore {
   public void deleteStore() throws Exception {
   }
 
+  @Override
+  public void removeApplication(ApplicationId removeAppId) 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

@@ -150,4 +150,9 @@ public class NullRMStateStore extends RMStateStore {
     // Do nothing
   }
 
+  @Override
+  public void removeApplication(ApplicationId removeAppId) 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

@@ -907,6 +907,15 @@ public abstract class RMStateStore extends AbstractService {
    */
   public abstract void deleteStore() throws Exception;
 
+  /**
+   * Derived classes must implement this method to remove application from the
+   * state store
+   * 
+   * @throws Exception
+   */
+  public abstract void removeApplication(ApplicationId removeAppId)
+      throws Exception;
+
   public void setResourceManager(ResourceManager rm) {
     this.resourceManager = rm;
   }

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

@@ -823,6 +823,15 @@ public class ZKRMStateStore extends RMStateStore {
     }
   }
 
+  @Override
+  public synchronized void removeApplication(ApplicationId removeAppId)
+      throws Exception {
+    String appIdRemovePath = getNodePath(rmAppRoot, removeAppId.toString());
+    if (existsWithRetries(appIdRemovePath, true) != null) {
+      deleteWithRetries(appIdRemovePath, true);
+    }
+  }
+
   // ZK related code
   /**
    * Watcher implementation which forward events to the ZKRMStateStore This

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

@@ -591,6 +591,21 @@ public class RMStateStoreTestBase extends ClientBaseWithFixes{
     }
   }
 
+  public void testRemoveApplication(RMStateStoreHelper stateStoreHelper)
+      throws Exception {
+    RMStateStore store = stateStoreHelper.getRMStateStore();
+    int noOfApps = 2;
+    ArrayList<RMApp> appList =
+        createAndStoreApps(stateStoreHelper, store, noOfApps);
+
+    RMApp rmApp1 = appList.get(0);
+    store.removeApplication(rmApp1.getApplicationId());
+    Assert.assertFalse(stateStoreHelper.appExists(rmApp1));
+
+    RMApp rmApp2 = appList.get(1);
+    Assert.assertTrue(stateStoreHelper.appExists(rmApp2));
+  }
+
   protected void modifyAppState() throws Exception {
 
   }

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

@@ -171,6 +171,7 @@ public class TestFSRMStateStore extends RMStateStoreTestBase {
       testEpoch(fsTester);
       testAppDeletion(fsTester);
       testDeleteStore(fsTester);
+      testRemoveApplication(fsTester);
       testAMRMTokenSecretManagerStateStore(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/TestLeveldbRMStateStore.java

@@ -90,6 +90,12 @@ public class TestLeveldbRMStateStore extends RMStateStoreTestBase {
     testDeleteStore(tester);
   }
 
+  @Test(timeout = 60000)
+  public void testRemoveApplication() throws Exception {
+    LeveldbStateStoreTester tester = new LeveldbStateStoreTester();
+    testRemoveApplication(tester);
+  }
+
   @Test(timeout = 60000)
   public void testAMTokens() throws Exception {
     LeveldbStateStoreTester tester = new LeveldbStateStoreTester();

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

@@ -146,6 +146,7 @@ public class TestZKRMStateStore extends RMStateStoreTestBase {
     testEpoch(zkTester);
     testAppDeletion(zkTester);
     testDeleteStore(zkTester);
+    testRemoveApplication(zkTester);
     testAMRMTokenSecretManagerStateStore(zkTester);
   }
 

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-site/src/site/markdown/YarnCommands.md

@@ -193,6 +193,7 @@ Usage: `yarn resourcemanager [-format-state-store]`
 | COMMAND\_OPTIONS | Description |
 |:---- |:---- |
 | -format-state-store | 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. |
+| -remove-application-from-state-store <appId> | Remove the application from RMStateStore. This should be run only when the ResourceManager is not running. |
 
 Start the ResourceManager