HADOOP-8077. HA: fencing method should be able to be configured on a per-NN or per-NS basis. Contributed by Todd Lipcon.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1310173 13f79535-47bb-0310-9956-ffa450edef68

hadoop-common-project/hadoop-common/CHANGES.txt

@@ -252,6 +252,9 @@ Release 2.0.0 - UNRELEASED
 
     HADOOP-8007. Use substitution tokens for fencing argument (todd)
 
+    HADOOP-8077. HA: fencing method should be able to be configured on
+    a per-NN or per-NS basis (todd)
+
   OPTIMIZATIONS
 
   BUG FIXES

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

@@ -53,9 +53,6 @@ import com.google.common.collect.Lists;
 @InterfaceAudience.Private
 @InterfaceStability.Evolving
 public class NodeFencer {
-  public static final String CONF_METHODS_KEY =
-    "dfs.ha.fencing.methods";
-  
   private static final String CLASS_RE = "([a-zA-Z0-9\\.\\$]+)";
   private static final Pattern CLASS_WITH_ARGUMENT =
     Pattern.compile(CLASS_RE + "\\((.+?)\\)");
@@ -76,18 +73,18 @@ public class NodeFencer {
   
   private final List<FenceMethodWithArg> methods;
   
-  public NodeFencer(Configuration conf)
+  NodeFencer(Configuration conf, String spec)
       throws BadFencingConfigurationException {
-    this.methods = parseMethods(conf);
+    this.methods = parseMethods(conf, spec);
   }
   
-  public static NodeFencer create(Configuration conf)
+  public static NodeFencer create(Configuration conf, String confKey)
       throws BadFencingConfigurationException {
-    String confStr = conf.get(CONF_METHODS_KEY);
+    String confStr = conf.get(confKey);
     if (confStr == null) {
       return null;
     }
-    return new NodeFencer(conf);
+    return new NodeFencer(conf, confStr);
   }
 
   public boolean fence(HAServiceTarget fromSvc) {
@@ -115,10 +112,10 @@ public class NodeFencer {
     return false;
   }
 
-  private static List<FenceMethodWithArg> parseMethods(Configuration conf)
+  private static List<FenceMethodWithArg> parseMethods(Configuration conf,
+      String spec)
       throws BadFencingConfigurationException {
-    String confStr = conf.get(CONF_METHODS_KEY);
+    String[] lines = spec.split("\\s*\n\\s*");
-    String[] lines = confStr.split("\\s*\n\\s*");
     
     List<FenceMethodWithArg> methods = Lists.newArrayList();
     for (String line : lines) {

hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/ha/TestNodeFencer.java

@@ -132,8 +132,7 @@ public class TestNodeFencer {
       throws BadFencingConfigurationException {
     System.err.println("Testing configuration:\n" + confStr);
     Configuration conf = new Configuration();
-    conf.set(NodeFencer.CONF_METHODS_KEY, confStr);
+    return new NodeFencer(conf, confStr);
-    return new NodeFencer(conf);
   }
   
   /**

hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/ha/TestShellCommandFencer.java

@@ -71,8 +71,7 @@ public class TestShellCommandFencer {
   public void testCheckNoArgs() {
     try {
       Configuration conf = new Configuration();
-      conf.set(NodeFencer.CONF_METHODS_KEY, "shell");
+      new NodeFencer(conf, "shell");
-      new NodeFencer(conf);
       fail("Didn't throw when passing no args to shell");
     } catch (BadFencingConfigurationException confe) {
       assertTrue(
@@ -85,8 +84,7 @@ public class TestShellCommandFencer {
   public void testCheckParensNoArgs() {
     try {
       Configuration conf = new Configuration();
-      conf.set(NodeFencer.CONF_METHODS_KEY, "shell()");
+      new NodeFencer(conf, "shell()");
-      new NodeFencer(conf);
       fail("Didn't throw when passing no args to shell");
     } catch (BadFencingConfigurationException confe) {
       assertTrue(

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/DFSConfigKeys.java

@@ -345,4 +345,5 @@ public class DFSConfigKeys extends CommonConfigurationKeys {
   public static final int DFS_HA_LOGROLL_PERIOD_DEFAULT = 2 * 60; // 2m
   public static final String DFS_HA_TAILEDITS_PERIOD_KEY = "dfs.ha.tail-edits.period";
   public static final int DFS_HA_TAILEDITS_PERIOD_DEFAULT = 60; // 1m
+  public static final String DFS_HA_FENCE_METHODS_KEY = "dfs.ha.fencing.methods";
 }

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

@@ -40,6 +40,8 @@ import org.apache.hadoop.fs.FileSystem;
 import org.apache.hadoop.fs.FileUtil;
 import org.apache.hadoop.fs.Trash;
 import static org.apache.hadoop.hdfs.DFSConfigKeys.*;
+
+import org.apache.hadoop.hdfs.DFSConfigKeys;
 import org.apache.hadoop.hdfs.DFSUtil;
 import org.apache.hadoop.hdfs.HAUtil;
 import org.apache.hadoop.hdfs.HdfsConfiguration;
@@ -160,7 +162,8 @@ public class NameNode {
     DFS_SECONDARY_NAMENODE_KEYTAB_FILE_KEY,
     DFS_NAMENODE_BACKUP_ADDRESS_KEY,
     DFS_NAMENODE_BACKUP_HTTP_ADDRESS_KEY,
-    DFS_NAMENODE_BACKUP_SERVICE_RPC_ADDRESS_KEY
+    DFS_NAMENODE_BACKUP_SERVICE_RPC_ADDRESS_KEY,
+    DFS_HA_FENCE_METHODS_KEY
   };
   
   public long getProtocolVersion(String protocol, 

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/tools/NNHAServiceTarget.java

@@ -24,6 +24,7 @@ import org.apache.hadoop.classification.InterfaceAudience;
 import org.apache.hadoop.ha.BadFencingConfigurationException;
 import org.apache.hadoop.ha.HAServiceTarget;
 import org.apache.hadoop.ha.NodeFencer;
+import org.apache.hadoop.hdfs.DFSConfigKeys;
 import org.apache.hadoop.hdfs.DFSUtil;
 import org.apache.hadoop.hdfs.HdfsConfiguration;
 import org.apache.hadoop.hdfs.server.namenode.NameNode;
@@ -75,7 +76,8 @@ public class NNHAServiceTarget extends HAServiceTarget {
     this.addr = NetUtils.createSocketAddr(serviceAddr,
         NameNode.DEFAULT_PORT);
     try {
-      this.fencer = NodeFencer.create(targetConf);
+      this.fencer = NodeFencer.create(targetConf,
+          DFSConfigKeys.DFS_HA_FENCE_METHODS_KEY);
     } catch (BadFencingConfigurationException e) {
       this.fenceConfigError = e;
     }

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/tools/TestDFSHAAdmin.java

@@ -158,7 +158,7 @@ public class TestDFSHAAdmin {
   public void testFailoverWithFencerConfigured() throws Exception {
     Mockito.doReturn(STANDBY_READY_RESULT).when(mockProtocol).getServiceStatus();
     HdfsConfiguration conf = getHAConf();
-    conf.set(NodeFencer.CONF_METHODS_KEY, "shell(true)");
+    conf.set(DFSConfigKeys.DFS_HA_FENCE_METHODS_KEY, "shell(true)");
     tool.setConf(conf);
     assertEquals(0, runTool("-failover", "nn1", "nn2"));
   }
@@ -167,7 +167,7 @@ public class TestDFSHAAdmin {
   public void testFailoverWithFencerAndNameservice() throws Exception {
     Mockito.doReturn(STANDBY_READY_RESULT).when(mockProtocol).getServiceStatus();
     HdfsConfiguration conf = getHAConf();
-    conf.set(NodeFencer.CONF_METHODS_KEY, "shell(true)");
+    conf.set(DFSConfigKeys.DFS_HA_FENCE_METHODS_KEY, "shell(true)");
     tool.setConf(conf);
     assertEquals(0, runTool("-ns", "ns1", "-failover", "nn1", "nn2"));
   }
@@ -176,7 +176,7 @@ public class TestDFSHAAdmin {
   public void testFailoverWithFencerConfiguredAndForce() throws Exception {
     Mockito.doReturn(STANDBY_READY_RESULT).when(mockProtocol).getServiceStatus();
     HdfsConfiguration conf = getHAConf();
-    conf.set(NodeFencer.CONF_METHODS_KEY, "shell(true)");
+    conf.set(DFSConfigKeys.DFS_HA_FENCE_METHODS_KEY, "shell(true)");
     tool.setConf(conf);
     assertEquals(0, runTool("-failover", "nn1", "nn2", "--forcefence"));
   }
@@ -185,7 +185,7 @@ public class TestDFSHAAdmin {
   public void testFailoverWithForceActive() throws Exception {
     Mockito.doReturn(STANDBY_READY_RESULT).when(mockProtocol).getServiceStatus();
     HdfsConfiguration conf = getHAConf();
-    conf.set(NodeFencer.CONF_METHODS_KEY, "shell(true)");
+    conf.set(DFSConfigKeys.DFS_HA_FENCE_METHODS_KEY, "shell(true)");
     tool.setConf(conf);
     assertEquals(0, runTool("-failover", "nn1", "nn2", "--forceactive"));
   }
@@ -194,7 +194,7 @@ public class TestDFSHAAdmin {
   public void testFailoverWithInvalidFenceArg() throws Exception {
     Mockito.doReturn(STANDBY_READY_RESULT).when(mockProtocol).getServiceStatus();
     HdfsConfiguration conf = getHAConf();
-    conf.set(NodeFencer.CONF_METHODS_KEY, "shell(true)");
+    conf.set(DFSConfigKeys.DFS_HA_FENCE_METHODS_KEY, "shell(true)");
     tool.setConf(conf);
     assertEquals(-1, runTool("-failover", "nn1", "nn2", "notforcefence"));
   }
@@ -209,7 +209,7 @@ public class TestDFSHAAdmin {
   public void testFailoverWithFenceAndBadFencer() throws Exception {
     Mockito.doReturn(STANDBY_READY_RESULT).when(mockProtocol).getServiceStatus();
     HdfsConfiguration conf = getHAConf();
-    conf.set(NodeFencer.CONF_METHODS_KEY, "foobar!");
+    conf.set(DFSConfigKeys.DFS_HA_FENCE_METHODS_KEY, "foobar!");
     tool.setConf(conf);
     assertEquals(-1, runTool("-failover", "nn1", "nn2", "--forcefence"));
   }
@@ -218,7 +218,7 @@ public class TestDFSHAAdmin {
   public void testForceFenceOptionListedBeforeArgs() throws Exception {
     Mockito.doReturn(STANDBY_READY_RESULT).when(mockProtocol).getServiceStatus();
     HdfsConfiguration conf = getHAConf();
-    conf.set(NodeFencer.CONF_METHODS_KEY, "shell(true)");
+    conf.set(DFSConfigKeys.DFS_HA_FENCE_METHODS_KEY, "shell(true)");
     tool.setConf(conf);
     assertEquals(0, runTool("-failover", "--forcefence", "nn1", "nn2"));
   }
@@ -241,6 +241,40 @@ public class TestDFSHAAdmin {
     assertOutputContains("Health check failed: fake health check failure");
   }
   
+  /**
+   * Test that the fencing configuration can be overridden per-nameservice
+   * or per-namenode
+   */
+  @Test
+  public void testFencingConfigPerNameNode() throws Exception {
+    Mockito.doReturn(STANDBY_READY_RESULT).when(mockProtocol).getServiceStatus();
+
+    final String nsSpecificKey = DFSConfigKeys.DFS_HA_FENCE_METHODS_KEY + "." + NSID;
+    final String nnSpecificKey = nsSpecificKey + ".nn1";
+    
+    HdfsConfiguration conf = getHAConf();
+    // Set the default fencer to succeed
+    conf.set(DFSConfigKeys.DFS_HA_FENCE_METHODS_KEY, "shell(true)");
+    tool.setConf(conf);
+    assertEquals(0, runTool("-failover", "nn1", "nn2", "--forcefence"));
+    
+    // Set the NN-specific fencer to fail. Should fail to fence.
+    conf.set(nnSpecificKey, "shell(false)");
+    tool.setConf(conf);
+    assertEquals(-1, runTool("-failover", "nn1", "nn2", "--forcefence"));
+    conf.unset(nnSpecificKey);
+
+    // Set an NS-specific fencer to fail. Should fail.
+    conf.set(nsSpecificKey, "shell(false)");
+    tool.setConf(conf);
+    assertEquals(-1, runTool("-failover", "nn1", "nn2", "--forcefence"));
+    
+    // Set the NS-specific fencer to succeed. Should succeed
+    conf.set(nsSpecificKey, "shell(true)");
+    tool.setConf(conf);
+    assertEquals(0, runTool("-failover", "nn1", "nn2", "--forcefence"));
+  }
+  
   private Object runTool(String ... args) throws Exception {
     errOutBytes.reset();
     LOG.info("Running: DFSHAAdmin " + Joiner.on(" ").join(args));

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/tools/TestDFSHAAdminMiniCluster.java

@@ -28,6 +28,7 @@ import org.apache.commons.logging.LogFactory;
 import org.apache.commons.logging.Log;
 import org.apache.commons.logging.impl.Log4JLogger;
 import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.hdfs.DFSConfigKeys;
 import org.apache.hadoop.hdfs.MiniDFSCluster;
 import org.apache.hadoop.hdfs.MiniDFSNNTopology;
 import org.apache.hadoop.hdfs.server.namenode.NameNode;
@@ -113,7 +114,7 @@ public class TestDFSHAAdminMiniCluster {
   
   @Test
   public void testTryFailoverToSafeMode() throws Exception {
-    conf.set(NodeFencer.CONF_METHODS_KEY, "shell(true)");
+    conf.set(DFSConfigKeys.DFS_HA_FENCE_METHODS_KEY, "shell(true)");
     tool.setConf(conf);
 
     NameNodeAdapter.enterSafeMode(cluster.getNameNode(0), false);
@@ -135,7 +136,7 @@ public class TestDFSHAAdminMiniCluster {
     // tmp file, so we can verify that the args were substituted right
     File tmpFile = File.createTempFile("testFencer", ".txt");
     tmpFile.deleteOnExit();
-    conf.set(NodeFencer.CONF_METHODS_KEY,
+    conf.set(DFSConfigKeys.DFS_HA_FENCE_METHODS_KEY,
         "shell(echo -n $target_nameserviceid.$target_namenodeid " +
         "$target_port $dfs_ha_namenode_id > " +
         tmpFile.getAbsolutePath() + ")");
@@ -168,19 +169,19 @@ public class TestDFSHAAdminMiniCluster {
 
           
     // Test failover with not fencer and forcefence option
-    conf.unset(NodeFencer.CONF_METHODS_KEY);
+    conf.unset(DFSConfigKeys.DFS_HA_FENCE_METHODS_KEY);
     tool.setConf(conf);
     assertEquals(-1, runTool("-failover", "nn1", "nn2", "--forcefence"));
     assertFalse(tmpFile.exists());
 
     // Test failover with bad fencer and forcefence option
-    conf.set(NodeFencer.CONF_METHODS_KEY, "foobar!");
+    conf.set(DFSConfigKeys.DFS_HA_FENCE_METHODS_KEY, "foobar!");
     tool.setConf(conf);
     assertEquals(-1, runTool("-failover", "nn1", "nn2", "--forcefence"));
     assertFalse(tmpFile.exists());
 
     // Test failover with force fence listed before the other arguments
-    conf.set(NodeFencer.CONF_METHODS_KEY, "shell(true)");
+    conf.set(DFSConfigKeys.DFS_HA_FENCE_METHODS_KEY, "shell(true)");
     tool.setConf(conf);
     assertEquals(0, runTool("-failover", "--forcefence", "nn1", "nn2"));
   }