1. Install the JournalNode daemons on each of the machines where they will run.

   To install JournalNode on Red Hat-compatible systems:

   $ sudo yum install hadoop-hdfs-journalnode

   To install JournalNode on Ubuntu and Debian systems:

   $ sudo apt-get install hadoop-hdfs-journalnode 

   To install JournalNode on SLES systems:

   $ sudo zypper install hadoop-hdfs-journalnode

2. Start the JournalNode daemons on each of the machines where they will run:

   sudo service hadoop-hdfs-journalnode start 

Wait for the daemons to start before formatting the primary NameNode (in a new cluster) and before starting the NameNodes (in all cases).

1. Start the primary (formatted) NameNode:

   $ sudo service hadoop-hdfs-namenode start

2. Start the standby NameNode:

   $ sudo -u hdfs hdfs namenode -bootstrapStandby
   $ sudo service hadoop-hdfs-namenode start 

     Note:

   If Kerberos is enabled, do not use commands in the form sudo -u <user> <command>; they will fail with a security error. Instead, use the following commands: $ kinit <user> (if you are using a password) or $ kinit -kt <keytab> <principal> (if you are using a keytab) and then, for each command executed by this user, $ <command>

Starting the standby NameNode with the -bootstrapStandby option copies over the contents of the primary NameNode's metadata directories (including the namespace information and most recent checkpoint) to the standby NameNode. (The location of the directories containing the NameNode metadata is configured via the configuration options dfs.namenode.name.dir and/or dfs.namenode.edits.dir.)

You can visit each NameNode's web page by browsing to its configured HTTP address. Notice that next to the configured address is the HA state of the NameNode (either "Standby" or "Active".) Whenever an HA NameNode starts and automatic failover is not enabled, it is initially in the Standby state. If automatic failover is enabled the first NameNode that is started will become active.

If you are converting from a non-HA to an HA configuration, you need to restart the JobTracker and TaskTracker (for MRv1, if used), or ResourceManager, NodeManager, and JobHistory Server (for YARN), and the DataNodes:

On each DataNode:

$ sudo service hadoop-hdfs-datanode start

On each TaskTracker system (MRv1):

$ sudo service hadoop-0.20-mapreduce-tasktracker start

On the JobTracker system (MRv1):

$ sudo service hadoop-0.20-mapreduce-jobtracker start

Verify that the JobTracker and TaskTracker started properly:

sudo jps | grep Tracker

On the ResourceManager system (YARN):

$ sudo service hadoop-yarn-resourcemanager start

On each NodeManager system (YARN; typically the same ones where DataNode service runs):

$ sudo service hadoop-yarn-nodemanager start

On the MapReduce JobHistory Server system (YARN):

$ sudo service hadoop-mapreduce-historyserver start

If you have configured automatic failover using the ZooKeeper FailoverController (ZKFC), you must install and start the zkfc daemon on each of the machines that runs a NameNode. Proceed as follows.

To install ZKFC on Red Hat-compatible systems:

$ sudo yum install hadoop-hdfs-zkfc

To install ZKFC on Ubuntu and Debian systems:

$ sudo apt-get install hadoop-hdfs-zkfc

To install ZKFC on SLES systems:

$ sudo zypper install hadoop-hdfs-zkfc

To start the zkfc daemon:

$ sudo service hadoop-hdfs-zkfc start

It is not important that you start the ZKFC and NameNode daemons in a particular order. On any given node you can start the ZKFC before or after its corresponding NameNode.

You should add monitoring on each host that runs a NameNode to ensure that the ZKFC remains running. In some types of ZooKeeper failures, for example, the ZKFC may unexpectedly exit, and should be restarted to ensure that the system is ready for automatic failover.

Additionally, you should monitor each of the servers in the ZooKeeper quorum. If ZooKeeper crashes, then automatic failover will not function. If the ZooKeeper cluster crashes, no automatic failovers will be triggered. However, HDFS will continue to run without any impact. When ZooKeeper is restarted, HDFS will reconnect with no issues.

After the initial deployment of a cluster with automatic failover enabled, you should test its operation. To do so, first locate the active NameNode. As mentioned above, you can tell which node is active by visiting the NameNode web interfaces.

Once you have located your active NameNode, you can cause a failure on that node. For example, you can use kill -9 <pid of NN> to simulate a JVM crash. Or you can power-cycle the machine or its network interface to simulate different kinds of outages. After you trigger the outage you want to test, the other NameNode should automatically become active within several seconds. The amount of time required to detect a failure and trigger a failover depends on the configuration of ha.zookeeper.session-timeout.ms, but defaults to 5 seconds.

If the test does not succeed, you may have a misconfiguration. Check the logs for the zkfc daemons as well as the NameNode daemons in order to further diagnose the issue.