bootstraping ZKFC hangs forever

[wuzesheng]

Forks may find that when bootstraping ZKFC after cleaning up, it will hang forever, to fix this, you need to patch the following diff to your ZKFC:

diff --git a/hadoop/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/ha/ZKFailoverController.java b/hadoop/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/ha/ZKFailoverController.java
--- a/hadoop/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/ha/ZKFailoverController.java
+++ b/hadoop/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/ha/ZKFailoverController.java
@@ -195,6 +195,8 @@
           }   
         }   
         return formatZK(force, interactive);
+      } else if ("-clearZK".equals(args[0])) {
+        return clearZK();
       } else {
         badArg(args[0]);
       }   
@@ -272,6 +274,19 @@
     return 0;
   }   

+  private int clearZK()
+      throws IOException, InterruptedException {
+    if (elector.parentZNodeExists()) {
+      try {
+        elector.clearParentZNode();
+      } catch (IOException e) {
+        LOG.error("Unable to clear zk parent znode", e); 
+        return 1;
+      }   
+    }   
+    return 0;
+  }
+   
   private boolean confirmFormat() {
     String parentZnode = getParentZnode();
     System.err.println(

[hpttlook]

There is another choice which you don't need to re-build your hadoop project.
Change the follwing line in bootstrap_hdfs.sh.tmpl
exec $java -cp $class_path $params $@
to
echo Y | exec $java -cp $class_path $params $@