Restarting Transformation

If a transformation run fails (for example, because of a system or process failure), the user can restart BDT with the -r option so that the utility resumes processing from the last committed batch. Such a restart is a more efficient and faster way to resume transformation, rather than rerunning from the beginning.

The following requirements must be met to successfully restart a transformation run:

• The destination must be empty at the time of the first run.

• There cannot be any new records in either source file or table.

To restart a BDT transformation at the point of failure, use the -r option:

bdt -r <transformation_id>

Where, by default transformation_id is the ID attribute in the policy file. If there are duplicate policy IDs in the internal BDT database, find the proper transformation ID from the failed transformation log file entry.

The transformation_id can be explicitly provided with bdt -t in the failed run.

bdt -t <transformation_id>

BDT launches the new transformation job to process the remaining records from the point of the failed transformation. If the number of records in the committed state and the number of actual written records are not identical, BDT cannot restart the transformation. This is a rare case where the system failed just after a record batch write but before writing the commit state.