You can write reject files to the Data Integration Service machine or to the Hadoop cluster. Or, you can defer to the Hadoop connection configuration. The Blaze engine and Spark engine can write reject files to the Hadoop environment for flat file and HDFS targets.
If you configure the mapping run-time properties to defer to the Hadoop connection, the reject files for all mappings with this configuration are moved based on whether you choose to write reject files to Hadoop for the active Hadoop connection. You do not need to change the mapping run-time properties manually to change the reject file directory.
For example, if the reject files are currently moved to the Data Integration Service machine and you want to move them to the directory configured in the Hadoop connection, edit the Hadoop connection properties to write reject files to Hadoop. The reject files of all mappings that are configured to defer to the Hadoop connection are moved to the configured directory.
You might also to choose to defer to the Hadoop connection when the connection is parameterized to alternate between multiple Hadoop connections. For example, the parameter might alternate between one Hadoop connection that is configured to move reject files to the Data Integration Service machine and another Hadoop connection that is configured to move reject files to the directory configured in the Hadoop connection. If you choose to defer to the Hadoop connection, the reject files are moved depending on the active Hadoop connection in the connection parameter.
You cannot create the reject file directory at run time. Depending on the option that you choose, the configured reject file directory must exist on the Data Integration Service machine or in the Hadoop environment before you can store reject files.