You can choose to store reject files on HDFS for the Blaze and Spark engines.
Reject files can be very large, and you can choose to write them to HDFS instead of the Data Integration Service machine. You can configure the Hadoop connection object to write to the reject file directory.
Grant permission to the following users:
Hadoop impersonation user
Mapping impersonation users
If the blaze user does not have permission, the Blaze engine uses a different user, based on the cluster security and the mapping impersonation configuration.