The following services require an explicit initialization of the persistent state in the local filesystem:
- Hadoop NameNode (formatting a namenode via: hdfs namenode -format)
- ZooKeeper (formatting a local storage area via: zookeeper-server-initialize)
and the following ones require an initialization of the RDBMS database (which can reside on a local filesystem
via Derby or be hosted on a remote server such as Postgress, MySQL, Oracle, etc.):
- oozie DB (initialized via ooziedb.sh)
- possible Hive metastore
- possibly Sqoop metastore
In order to free the user from an explicit knowledge of what command to run under which account it is desirable
to have an init.d scripts for the above components support and extra command called 'init'.
Please let me know what do you all think.