I'm working with Hadoop 1.2.1 to create a series of chained map reduce jobs which will be run regularly in our production environment. At the moment, we are using graylog2 to get centralized access to logs and I would like to have the logs from my job sent to our log server.
I've added the gelfj jar file to /usr/share/hadoop/lib and modified the /etc/hadoop/task-log4j.properties to output logs to graylog but so far am not seeing anything arrive on the graylog2 side. Can anyone confirm that using an alternate log appender is possible for hadoop jobs and what config file should be updated to do so?
Figured this out.
In the hadoop configuration directory (/etc/hadoop in my case) there is a file named task-log4j.properties. Overriding the appender "log4j.appender.TLA" affect the log output from tasks.