The current mr_job table, jobconf columns contains the content of jobconf for some rows, and other rows contains the
path of the jobconf.
In the sequence file, jobconf is the field for location of the jobconf path. The content of the jobconf is
JOBCONF-JSON. In database table, jobconf is the content of the jobconf. Unfortunately, the new automatic mapping will
detect sequence file "jobconf" field is a match with "jobconf" column in the database. Hence the
last update wins.
To work around this problem, we need ability to undefine sequence file field to match to database column.
Personally, I think indirect data translation mapping is the wrong thing to do. The mapping should be as straight forward as possible.