=== Flume Environment Variables This section describes several environment variables that affect how Flume operates. The +flume+ script in +./bin+ uses these system environment variables. Many of these variables are set by the +flume-daemon.sh+ script used when flume is run as a daemon. [horizontal] +FLUME_PID_DIR+ :: The directory where a flume node or flume master will drop pid files corresponding to the daemon process. +FLUME_CLASSPATH+ :: The custom java classpath environment variable additions you want flume to run with. These values are prepended to the normal Flume generated CLASSPATH. WARNING: The +flume+ script overwrites the standard CLASSPATH when it is executed. +FLUME_LOG_DIR+ :: The directory where debugging logs generated by the flume node or flume master are written. +FLUME_LOGFILE+ :: This sets the suffix for the logfiles generated by flume node or flume master. +FLUME_ROOT_LOGGER+ :: The log4j logging setting for the executing command. By default it is "INFO,console". +ZOOKEEPER_ROOT_LOGGER+ :: The log4j logging setting for a master's embedded zookeeper server's logs. By default it is "ERROR,console". +WATCHDOOG_ROOT_LOGGER+ :: The log4j logging setting for the logging the watchdog that wraps flume nodes and flume masters generates. By default it is "INFO,console". +FLUME_CONF_DIR+ :: The directory where the +flume-site.xml+ and +flume-conf.xml+ files flume node and flume master will use reside. This defaults to +./conf+ if a +./conf/flume-conf.xml+ file is found, or to +/etc/flume/conf/+ if it +flume-conf.xml+ is found there. +HADOOP_HOME+ :: The directory where Hadoop jars are expected to be found. If not specified it will use jars found in +/usr/lib/hadoop+ or +./lib/+. +FLUME_DEVMODE+ :: If this value is set to "true" the +./libbuild+ jars which include ant jars required to compile JSP servlets will be included in the CLASSPATH. +FLUME_VERBOSE+ :: If it this is toggled, the flume script will print out the command line being executed. +FLUME_VERBOSE_JAVA+ :: If this is toggled along with FLUME_VERBOSE, the "-verbose" flag will be passed to the JVM running flume. +LD_LIBRARY_PATH+ :: Populate with a : separated list of file paths to extend the java.library.path aspect of Flume. This allows you to include native libraries in the java lookup path. Example usage would be with Lzo where you need to extend the path with the Lzo C libraries contained in your Hadoop install.