Rejected data is closely coupled to schemas (Chapter 2, Metadata and Schemas), as many of the input and output components will validate data according to a schema definition and then pass any incorrect data to a reject flow.
Reject flows thus allow non-conforming data to be collected and handled as per the needs of a project.
In some cases, depending upon the business requirement, rejects are not acceptable. In these cases, reject flows should be disabled and the job allowed to fail.
Whether a job dies on the first incorrect record, collects rejects in a file, or completely ignores rejects is a design decision that should be based upon the requirements for the process. Where possible, designers and developers ...