hadoop-common-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Apache Wiki <wikidi...@apache.org>
Subject [Lucene-hadoop Wiki] Update of "HowToConfigure" by OwenOMalley
Date Fri, 30 Jun 2006 22:26:51 GMT
Dear Wiki user,

You have subscribed to a wiki page or wiki category on "Lucene-hadoop Wiki" for change notification.

The following page has been changed by OwenOMalley:
http://wiki.apache.org/lucene-hadoop/HowToConfigure

New page:
= How To Configure Hadoop =

Hadoop is configured with a set of files. The files are loaded in order with the lower files
taking priority over the higher ones:

|| '''Filename''' || '''Description''' ||
|| hadoop-default.xml || Generic default values ||
|| mapred-default.xml || Site specific default values ||
|| job.xml || Configuration for a specific map/reduce job ||
|| hadoop-site.xml || Site specific value that can not be modified by the job ||

== Look up path ==

Configuration files are found via Java's Classpath. Only the first instance of each file is
used. The $HADOOP_CONF_DIR is added by the bin/hadoop script to the front of the path. When
installing Hadoop on a cluster, it is best to use a conf directory outside of the distribution.
That allows you to easily update the release on the cluster without changing your configuration
by mistake.

== Hadoop-default.xml ==

This file has the default values for many of the configuration variables that are used by
Hadoop. This file should never be in $HADOOP_CONF_DIR so that the version in the hadoop-*.jar
is used. (Otherwise, if a variable is added to this file in a new release, you won't have
it defined.)

== mapred-default.xml ==

This file should contain the majority of your customization of hadoop. Useful variables are:

|| '''Name''' || '''Meaning''' ||
|| dfs.block.size || size in bytes of each data block in DFS ||
|| io.sort.factor || number of input files to each level in the merge sort ||
|| io.sort.mb || size of buffer to sort the reduce inputs in ||
|| io.file.buffer.size || number of bytes used for buffering io files ||
|| mapred.reduce.parallel.copies || number of threads fetching map outputs for each reduce
||
|| dfs.replication || number of replicas for each DFS block ||
|| mapred.child.java.opts || options passed to children task jvms ||
|| mapred.min.split.size || minimum number of bytes in a map input split ||
|| mapred.output.compress || Should the reduce outputs be compressed? ||



== job.xml ==

This file is never created explicitly by the user. The map/reduce application creates a JobConf,
which is serialized when the job is submitted.

== hadoop-site.xml ==

This file overrides any settings in the job.xml and therefore should be very minimal. Usually
it just contains, the addresses of the NameNode and JobTracker, the port and working directories
for the various servers.

Mime
View raw message