YARN-4127. RM fail with noAuth error if switched from failover mode to non-failover mode. Contributed by Varun Saxena

hadoop-yarn-project/CHANGES.txt

@@ -118,6 +118,9 @@ Release 2.7.2 - UNRELEASED
     YARN-3136. getTransferredContainers can be a bottleneck during AM registration.
     (Sunil G via jianhe)
 
+    YARN-4127. RM fail with noAuth error if switched from failover mode to non-failover
+    mode. (Varun Saxena via jianhe)
+
 Release 2.7.1 - 2015-07-06
 
   INCOMPATIBLE CHANGES

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

@@ -305,8 +305,9 @@ public class ZKRMStateStore extends RMStateStore {
     // ensure root dirs exist
     createRootDirRecursively(znodeWorkingPath);
     createRootDir(zkRootNodePath);
+    setRootNodeAcls();
+    deleteFencingNodePath();
     if (HAUtil.isHAEnabled(getConfig())){
-      fence();
       verifyActiveStatusThread = new VerifyActiveStatusThread();
       verifyActiveStatusThread.start();
     }
@@ -352,31 +353,41 @@ public class ZKRMStateStore extends RMStateStore {
     LOG.debug(builder.toString());
   }
 
-  private synchronized void fence() throws Exception {
-    if (LOG.isTraceEnabled()) {
-      logRootNodeAcls("Before fencing\n");
-    }
-
-    new ZKAction<Void>() {
-      @Override
-      public Void run() throws KeeperException, InterruptedException {
-        zkClient.setACL(zkRootNodePath, zkRootNodeAcl, -1);
-        return null;
-      }
-    }.runWithRetries();
-
-    // delete fencingnodepath
+  private void deleteFencingNodePath() throws Exception {
     new ZKAction<Void>() {
       @Override
       public Void run() throws KeeperException, InterruptedException {
         try {
           zkClient.multi(Collections.singletonList(deleteFencingNodePathOp));
         } catch (KeeperException.NoNodeException nne) {
-          LOG.info("Fencing node " + fencingNodePath + " doesn't exist to delete");
+          LOG.info("Fencing node " + fencingNodePath +
+              " doesn't exist to delete");
         }
         return null;
       }
     }.runWithRetries();
+  }
+
+  private void setAcl(final String zkPath, final List<ACL> acl)
+      throws Exception {
+    new ZKAction<Void>() {
+      @Override
+      public Void run() throws KeeperException, InterruptedException {
+        zkClient.setACL(zkPath, acl, -1);
+        return null;
+      }
+    }.runWithRetries();
+  }
+
+  private void setRootNodeAcls() throws Exception {
+    if (LOG.isTraceEnabled()) {
+      logRootNodeAcls("Before fencing\n");
+    }
+    if (HAUtil.isHAEnabled(getConfig())) {
+      setAcl(zkRootNodePath, zkRootNodeAcl);
+    } else {
+      setAcl(zkRootNodePath, zkAcl);
+    }
 
     if (LOG.isTraceEnabled()) {
       logRootNodeAcls("After fencing\n");
@@ -1027,7 +1038,8 @@ public class ZKRMStateStore extends RMStateStore {
     }.runWithRetries();
   }
 
-  private List<ACL> getACLWithRetries(
+  @VisibleForTesting
+  List<ACL> getACLWithRetries(
       final String path, final Stat stat) throws Exception {
     return new ZKAction<List<ACL>>() {
       @Override

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

@@ -53,6 +53,7 @@ import org.apache.hadoop.yarn.conf.YarnConfiguration;
 import org.apache.hadoop.yarn.security.client.RMDelegationTokenIdentifier;
 import org.apache.hadoop.yarn.server.records.Version;
 import org.apache.hadoop.yarn.server.records.impl.pb.VersionPBImpl;
+import org.apache.hadoop.yarn.server.resourcemanager.MockRM;
 import org.apache.hadoop.yarn.server.resourcemanager.RMZKUtils;
 import org.apache.hadoop.yarn.server.resourcemanager.ResourceManager;
 import org.apache.hadoop.yarn.server.resourcemanager.recovery.records.ApplicationAttemptStateData;
@@ -65,8 +66,10 @@ import org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptS
 import org.apache.hadoop.yarn.server.resourcemanager.security.ClientToAMTokenSecretManagerInRM;
 import org.apache.hadoop.yarn.util.ConverterUtils;
 import org.apache.zookeeper.KeeperException;
+import org.apache.zookeeper.ZooDefs.Perms;
 import org.apache.zookeeper.ZooKeeper;
 import org.apache.zookeeper.CreateMode;
+import org.apache.zookeeper.data.ACL;
 import org.apache.zookeeper.data.Stat;
 import org.junit.Assert;
 import org.junit.Test;
@@ -236,6 +239,73 @@ public class TestZKRMStateStore extends RMStateStoreTestBase {
     return conf;
   }
 
+  private static boolean verifyZKACL(String id, String scheme, int perm,
+      List<ACL> acls) {
+    for (ACL acl : acls) {
+      if (acl.getId().getScheme().equals(scheme) &&
+          acl.getId().getId().startsWith(id) &&
+          acl.getPerms() == perm) {
+        return true;
+      }
+    }
+    return false;
+  }
+
+  /**
+   * Test if RM can successfully start in HA disabled mode if it was previously
+   * running in HA enabled mode. And then start it in HA mode after running it
+   * with HA disabled. NoAuth Exception should not be sent by zookeeper and RM
+   * should start successfully.
+   */
+  @Test
+  public void testZKRootPathAcls() throws Exception {
+    StateChangeRequestInfo req = new StateChangeRequestInfo(
+        HAServiceProtocol.RequestSource.REQUEST_BY_USER);
+    String rootPath =
+        YarnConfiguration.DEFAULT_ZK_RM_STATE_STORE_PARENT_PATH + "/" +
+            ZKRMStateStore.ROOT_ZNODE_NAME;
+
+    // Start RM with HA enabled
+    Configuration conf = createHARMConf("rm1,rm2", "rm1", 1234);
+    conf.setBoolean(YarnConfiguration.AUTO_FAILOVER_ENABLED, false);
+    ResourceManager rm = new MockRM(conf);
+    rm.start();
+    rm.getRMContext().getRMAdminService().transitionToActive(req);
+    Stat stat = new Stat();
+    List<ACL> acls = ((ZKRMStateStore)rm.getRMContext().getStateStore()).
+        getACLWithRetries(rootPath, stat);
+    assertEquals(acls.size(), 2);
+    // CREATE and DELETE permissions for root node based on RM ID
+    verifyZKACL("digest", "localhost", Perms.CREATE | Perms.DELETE, acls);
+    verifyZKACL(
+        "world", "anyone", Perms.ALL ^ (Perms.CREATE | Perms.DELETE), acls);
+    rm.close();
+
+    // Now start RM with HA disabled. NoAuth Exception should not be thrown.
+    conf.setBoolean(YarnConfiguration.RM_HA_ENABLED, false);
+    rm = new MockRM(conf);
+    rm.start();
+    rm.getRMContext().getRMAdminService().transitionToActive(req);
+    acls = ((ZKRMStateStore)rm.getRMContext().getStateStore()).
+        getACLWithRetries(rootPath, stat);
+    assertEquals(acls.size(), 1);
+    verifyZKACL("world", "anyone", Perms.ALL, acls);
+    rm.close();
+
+    // Start RM with HA enabled.
+    conf.setBoolean(YarnConfiguration.RM_HA_ENABLED, true);
+    rm = new MockRM(conf);
+    rm.start();
+    rm.getRMContext().getRMAdminService().transitionToActive(req);
+    acls = ((ZKRMStateStore)rm.getRMContext().getStateStore()).
+        getACLWithRetries(rootPath, stat);
+    assertEquals(acls.size(), 2);
+    verifyZKACL("digest", "localhost", Perms.CREATE | Perms.DELETE, acls);
+    verifyZKACL(
+        "world", "anyone", Perms.ALL ^ (Perms.CREATE | Perms.DELETE), acls);
+    rm.close();
+  }
+
   @SuppressWarnings("unchecked")
   @Test
   public void testFencing() throws Exception {