YARN-8409. Fixed NPE in ActiveStandbyElectorBasedElectorService.

Contributed by Chandni Singh

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

@@ -329,9 +329,12 @@ public class ActiveStandbyElector implements StatCallback, StringCallback {
    * This recursively creates the znode as well as all of its parents.
    */
   public synchronized void ensureParentZNode()
-      throws IOException, InterruptedException {
+      throws IOException, InterruptedException, KeeperException {
     Preconditions.checkState(!wantToBeInElection,
         "ensureParentZNode() may not be called while in the election");
+    if (zkClient == null) {
+      createConnection();
+    }
 
     String pathParts[] = znodeWorkingDir.split("/");
     Preconditions.checkArgument(pathParts.length >= 1 &&

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

@@ -269,7 +269,7 @@ public abstract class ZKFailoverController {
   }
 
   private int formatZK(boolean force, boolean interactive)
-      throws IOException, InterruptedException {
+      throws IOException, InterruptedException, KeeperException {
     if (elector.parentZNodeExists()) {
       if (!force && (!interactive || !confirmFormat())) {
         return ERR_CODE_FORMAT_DENIED;

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

@@ -22,7 +22,10 @@ import org.apache.commons.logging.LogFactory;
 import org.apache.hadoop.conf.Configuration;
 import org.apache.hadoop.ha.ClientBaseWithFixes;
 import org.apache.hadoop.ha.ServiceFailedException;
+import org.apache.hadoop.service.ServiceStateException;
+import org.apache.hadoop.test.GenericTestUtils;
 import org.apache.hadoop.yarn.conf.YarnConfiguration;
+import org.junit.Assert;
 import org.junit.Before;
 import org.junit.Test;
 
@@ -278,6 +281,26 @@ public class TestRMEmbeddedElector extends ClientBaseWithFixes {
     verify(as, atMost(1)).transitionToStandby(any());
   }
 
+  /**
+   * Test that active elector service triggers a fatal RM Event when connection
+   * to ZK fails. YARN-8409
+   */
+  @Test
+  public void testFailureToConnectToZookeeper() throws Exception {
+    stopServer();
+    Configuration myConf = new Configuration(conf);
+    ResourceManager rm = new MockRM(conf);
+
+    ActiveStandbyElectorBasedElectorService ees =
+        new ActiveStandbyElectorBasedElectorService(rm);
+    try {
+      ees.init(myConf);
+      Assert.fail("expect failure to connect to Zookeeper");
+    } catch (ServiceStateException sse) {
+      Assert.assertTrue(sse.getMessage().contains("ConnectionLoss"));
+    }
+  }
+
   private class MockRMWithElector extends MockRM {
     private long delayMs = 0;