If you are loading data into a table in a multiple database partition nodegroup, the load utility requires that the files to be loaded are split and contain the correct header information. The load utility verifies the header information that the AutoLoader split operation writes to each data file to ensure that the data goes to the correct location.
If you are loading data into a table in a single database partition nodegroup, the files do not have to be split, even if the table is defined to have a partitioning key. In this situation, you would specify the noheader modifier for the load operation.
The load utility checks that the partitioning map used by the AutoLoader split operation is the same as that specified when the table is being loaded. If not, an error is returned. It also checks that the file partition is loaded at the correct database partition, and that the data types of the partitioning key columns specified during splitting match the current definition in the catalog. The nodegroup to which the table is loaded cannot be redistributed between the time that the data file is partitioned and the time that the parts are loaded into the corresponding database table. If redistribution has been done, the utility cannot load the partitioned data.
Although the load utility supports the following flat file formats:
AutoLoader can only be used to partition ASC and DEL files. PC/IXF files cannot be split, but can be loaded into a single database partition nodegroup using the noheader modifier for the load operation. If you want to load an IXF file into a multiple partition table, you must first load it into a single-partition table, and then insert the data into the multiple partition table using a SELECT from the single-partition table.
The LOAD ROWCOUNT clause is not supported in an AutoLoader operation. This parameter is only valid in a non-partitioned database environment. The LOAD SAVECOUNT clause is not supported if multiple splitters are used in an AutoLoader operation.
If a column that is part of the partitioning key is invalid or rejected, none of the data associated with that row is loaded. The row is not placed in the dumpfile, even if one has been specified; instead, a message indicating that the record has been rejected is written to the splitter log file. Be sure to check the splitter log file after the completion of the AutoLoader process.