Plugin Information

This plugin makes Hudson cluster act as a Hadoop cluster without any configuration.

When you install this plugin, Hudson will start a Hadoop "master" (namenode and job tracker, in their jargons) on the Hudson master. On each slave that Hudson starts, this plugin also starts a Hadoop "slave" (datanode and task tracker, in their jargons.) Hudson then instructs Hadoop to form a cluster among those nodes. All the jar files needed to run Hadoop is packaged inside this plugin, and will be automatically deployed to eveyrwhere they are needed.

In this way, this plugin takes your machines in your Hudson cluster and runs a Hadoop cluster very easily.

Hudson UI addition

Once installed, you'll see the "Hadoop" link in the Hudson top page.

This link will take you to the information page that shows links to Hadoop web UI, as well as the configuration users need on their hadoop-site.xml.

Slave log page (http://yourserver/hudson/computer/SLAVENAME/log) will also include text showing how the hadoop slave JVM is launched. Any error encountered will be recorded there, too.

Configuring Hadoop

Once you run this, you'll get $HUDSON_HOME/hadoop/dist directory that contains the stripped down version of Hadoop. You can change conf/hadoop-site.xml in this directory to apply different parameters.

Notes

• Hadoop requires that everyone in the cluster can reach everyone else in the cluster by a TCP connection.
• This plugin requires slave.jar from Hudson 1.292 or later, or else you'll get "undefined option -cp" error on slaves.
• Scheduling of the work in Hudson and that in Hadoop are independently done, without consideration to each other.

Changelog

Version 1.3 (2009 Sep 2)

• NameNode now listens on all interfaces.
• Fixed a cygwin compatibility problem (report)

Version 1.2 (2009 Jul 20)

• Fixed a bug with recent versions of Hudson (1.316, for example) where a failure to bring name node causes a failure to start Hudson
• Improved the plugin so that it automatically figures out the network topology.

Version 1.1 (2009 May 15)

• Hadoop wasn't seeing hadoop-site.xml and hadoop-default.xml.

Version 1.0

• Initial release