Regarding Hadoop secondarynamenode concept

Question

As per the documnetation (http://hadoop.apache.org/common/docs/r0.20.203.0/hdfs_user_guide.html) secondarynamenode is deprecated in hadoop0.20.203.0 release onwards and replaced by checkpointnode and backupnode. But in cluster set up doc (http://hadoop.apache.org/common/docs/r0.20.203.0/cluster_setup.html) no where mentioned about that change. More over bin/start-dfs.sh starting secondaryname node in the address mentioned in conf/masters file.

Can some one provide what is difference? Does it mean configuration not changed. only internal architecture of secondarynamenode is chnaged..

Also on hadoop0.23.0 release there is no conf/masters file where we used to specify the host address where secondary name node needs to start.

Thanks MRK

Answer 1

Also on hadoop0.23.0 release there is no conf/masters file where we used to specify the host address where secondary name node needs to start.

The documentation for 0.23 is a bit sparse and I had to dig through the code. In the DFSConfigKeys.java the below variables have been defined. Set the dfs.namenode.secondary.http-address key to ip:port in hdfs-site.xml and start the secondary namenode using the sbin/hadoop-daemon.sh start secondarynamenode command. Run jps command to check if the secondary namenode is running and also make sure to check the log file also for any errors.

public static final String DFS_NAMENODE_SECONDARY_HTTP_ADDRESS_KEY = "dfs.namenode.secondary.http-address";

Can some one provide what is difference?

Secondary NN is being called CheckPoint NN. But, the code is still using Secondary NN and people still refer it as Secondary NN.

From the description of the HDFS-2141 Jira which is related to 0.23 release

Active and Standby are the state of the NameNode. While Backup and CheckPoint are the name/role of the daemons that are started.