z2 initializer for hadoop jobs. This class runs within the hadoop task context,
that's why it may not reach out to other stuff not
available on the classpath at its execution time (which excludes anything but hadoop stuff)
This is a singleton within a hadoop task. As reducers may run as combiners within map tasks,
they will use the identical runner and hence job component etc. when nested.
There are the following combinations:
- External task, just Mapper or Reducer
- External task, Combiner right after or between Mappers
- Local task, Mapper, Combiner, Mapper, Reducer
In all cases, the z2 environment should not be stopped (as re-initialization
may not always work within the process), while the hadoop process
runner should actually be de-initialized after each run (in order to not hold
on to outdated or wrong job implementations).
In addition, the entry point is always on of