Jeff Zhang wrote:
you should use the log4j rather than the apache common logging
Not necessarily. Hadoop code uses commons logging which defaults to going
to log4j, but lets you plug in different back ends via JVM properties.
This lets you do some tricks with log aggregation.
If changes to log4.properties don't seem to get picked up, it could be
some JAR has a copy of the file which is getting in the way. You can use
Ant's <whichresource> task to maybe track this down.
<whichresource resource="/log4j.properties" property="log4j.url"
classpathref="some classpath reference">
