Like the GangliaContext for Hadoop, sends metrics to Graphite
Compile:
$ jar -cf GraphiteContext.jar -C org/ .
Installation:
In your hadoop-env.sh file (usually in /etc/hadoop/conf/), add the location of the GraphiteContext.jar file into the HADOOP_CLASSPATH
example: export HADOOP_CLASSPATH="/home/brandon/GraphiteContext.jar"
Configuration:
In your hadoop-metrics.properties file, add the following for all metrics
mapred.class=org.apache.hadoop.metrics.graphite.GraphiteContext
mapred.period=60
mapred.serverName=@Your Graphite Server@
mapred.port=2013
jvm.class=org.apache.hadoop.metrics.graphite.GraphiteContext
jvm.period=60
jvm.serverName=@Your Graphite Server@
jvm.port=2013
dfs.class=org.apache.hadoop.metrics.graphite.GraphiteContext
dfs.period=60
dfs.serverName=@Your Graphite Server@
dfs.port=2013
ugi.class=org.apache.hadoop.metrics.graphite.GraphiteContext
ugi.period=60
ugi.serverName=@Your Graphite Server@
ugi.port=2013
metric.path=@Your Path@ can be used to specify the path in Graphite. Defaults to Platform.Hadoop
Restart Daemons:
$ sudo /etc/init.d/hadoop-tasktracker restart
$ sudo /etc/init.d/hadoop-tasktracker restart
$ sudo /etc/init.d/hadoop-tasktracker restart
$ sudo /etc/init.d/hadoop-tasktracker restart