If binary columns will be used, or if character data columns will be output to the file system, you must be sure the output directory exists.

Unless overridden by the outputDataDir setting described previously, the default output directory will be the incoming directory parallel to Forge’s working directory. For example, the following directory tree shows this incoming directory, assuming forge_input is configured as the Forge working directory:

Note that the incoming directory is accessible at ../incoming relative to the pipeline files within forge_input/.

On initialization, the Advanced JDBC Column Handler will check to see if the outputDataDir is writable. If it is not, the Advanced JDBC Column Handler will throw a warning into the Forge log, similar to the following:

WARN    09/24/14 14:00:49.958 UTC		FORGE      {forge,baseline}: (com.endeca.soleng.itl.jdbc.AdvancedJDBCColumnHandler): outputDataDir ../incoming not writable

Because the Advanced JDBC Column Handler does not require a writable directory if only character columns are used and character data is not spooled to disk, Forge will continue processing when it encounters a non-writable outputDataDir. However, if binary columns are used or if character data is explicitly spooled to disk, Forge will not return any valid data for those columns, and will log additional warnings to the Forge log for each row and column in the database it is not able to process. These warnings will appear similar to the following:

WARN    09/24/06 14:02:34.828 UTC		FORGE      {forge,baseline}: (com.endeca.soleng.itl.jdbc.AdvancedJDBCColumnHandler): IOException while dumping MyBinaryData column