Hadoop: How can I start tasktracker?

Question

I want to open web/ui of tasktracker. localhost:50030/jobtracker.jsp But it doesn't work since there's no running tasktracker. My hadoop applications have been working well without tasktracker. But I found that tasktracker should be running in order to open web/ui. But I don't know how to start tasktracker. How can I start tasktracker? I tried lots of things but they don't work.

My status & environment: localhost:50070 web/ui is working well. jps command shows Datanode, NameNode, NodeManager, SecondaryNameNode, ResourceManager. Hadoop version is 2.2.0. Os is Red Hat 6.3(Santiago).

Thank you for your help in advance.

Answer 1

If you are using Hadoop 2.2.0 which has YARN framework, there is no jobtracker in it. Its functionality is split and replaced by ResourceManager and ApplicationMaster. Here is expected jps prinout while running YARN

$jps
18509 Jps
17107 NameNode
17170 DataNode
17252 ResourceManager
17309 NodeManager
17626 JobHistoryServer

Answer 2

From the master node, you can start the jobtracker and the tasktrackers of the slaves (these are the nodes in the file $HADOOP_CONF_DIR/slaves) with the command:

$HADOOP_HOME/bin/start-mapred.sh

If you want to start a tasktracker separately, from this node type:

$HADOOP_HOME/bin/hadoop-daemons.sh start tasktracker

Replace start with stop in the commands above for stopping the jobtracker and tasktrackers, respectively.

By the way, I think that you need to start the jobtracker, in order to see what you want, even if no tasktracker is running.

EDIT: These commands were used in the old API. The new API uses the following commands:

$ cd $HADOOP_MAPRED_HOME
$ sbin/yarn-daemon.sh start resourcemanager
$ sbin/yarn-daemon.sh start nodemanager

, as described here.