HDFS-6348. SecondaryNameNode not terminating properly on runtime exceptions (Contributed by Rakesh R)

hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt

@@ -1,4 +1,4 @@
-Hadoop HDFS Change Log
+ Hadoop HDFS Change Log
 
 Trunk (Unreleased)
 
@@ -788,6 +788,9 @@ Release 2.8.0 - UNRELEASED
     HDFS-8403. Eliminate retries in TestFileCreation
     #testOverwriteOpenForWrite. (Arpit Agarwal via wheat9)
 
+    HDFS-6348. SecondaryNameNode not terminating properly on runtime exceptions
+    (Rakesh R via vinayakumarb)
+
 Release 2.7.1 - UNRELEASED
 
   INCOMPATIBLE CHANGES

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/SecondaryNameNode.java

@@ -667,29 +667,28 @@ public class SecondaryNameNode implements Runnable,
       opts.usage();
       System.exit(0);
     }
-    
-    StringUtils.startupShutdownMessage(SecondaryNameNode.class, argv, LOG);
-    Configuration tconf = new HdfsConfiguration();
-    SecondaryNameNode secondary = null;
+
     try {
+      StringUtils.startupShutdownMessage(SecondaryNameNode.class, argv, LOG);
+      Configuration tconf = new HdfsConfiguration();
+      SecondaryNameNode secondary = null;
       secondary = new SecondaryNameNode(tconf, opts);
-    } catch (IOException ioe) {
-      LOG.fatal("Failed to start secondary namenode", ioe);
-      terminate(1);
-    }
 
-    if (opts != null && opts.getCommand() != null) {
-      int ret = secondary.processStartupCommand(opts);
-      terminate(ret);
-    }
+      if (opts != null && opts.getCommand() != null) {
+        int ret = secondary.processStartupCommand(opts);
+        terminate(ret);
+      }
 
-    if (secondary != null) {
-      secondary.startCheckpointThread();
-      secondary.join();
+      if (secondary != null) {
+        secondary.startCheckpointThread();
+        secondary.join();
+      }
+    } catch (Throwable e) {
+      LOG.fatal("Failed to start secondary namenode", e);
+      terminate(1);
     }
   }
-  
-  
+
   public void startCheckpointThread() {
     Preconditions.checkState(checkpointThread == null,
         "Should not already have a thread");

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestStartup.java

@@ -60,6 +60,8 @@ import org.apache.hadoop.hdfs.util.MD5FileUtils;
 import org.apache.hadoop.io.MD5Hash;
 import org.apache.hadoop.test.GenericTestUtils;
 import org.apache.hadoop.test.PathUtils;
+import org.apache.hadoop.util.ExitUtil.ExitException;
+import org.apache.hadoop.util.ExitUtil;
 import org.apache.hadoop.util.StringUtils;
 import org.apache.log4j.Logger;
 import org.junit.After;
@@ -87,6 +89,8 @@ public class TestStartup {
 
   @Before
   public void setUp() throws Exception {
+    ExitUtil.disableSystemExit();
+    ExitUtil.resetFirstExitException();
     config = new HdfsConfiguration();
     hdfsDir = new File(MiniDFSCluster.getBaseDirectory());
 
@@ -403,7 +407,19 @@ public class TestStartup {
         cluster.shutdown();
     }
   }
-  
+
+  @Test(timeout = 30000)
+  public void testSNNStartupWithRuntimeException() throws Exception {
+    String[] argv = new String[] { "-checkpoint" };
+    try {
+      SecondaryNameNode.main(argv);
+      fail("Failed to handle runtime exceptions during SNN startup!");
+    } catch (ExitException ee) {
+      GenericTestUtils.assertExceptionContains("ExitException", ee);
+      assertTrue("Didn't termiated properly ", ExitUtil.terminateCalled());
+    }
+  }
+
   @Test
   public void testCompression() throws IOException {
     LOG.info("Test compressing image.");