The DB2 LOAD process
Thursday, March 1, 2007 at 22:37 During the load phase , the load utility scans the input file for any invalid data rows that do not comply with the table definition; for example, if a table column is defined as INTEGER but the input data is stored as "abcd". Invalid data will not be loaded into the table. The rejected rows and warnings will be written to a dump file specified by the dumpfile modifier. Valid data is then written into the table. At the same time, table statistics (if the statistics option was specified) and index keys are also collected. If the savecount option is specified in the load command, points of consistency are recorded in the message file. Consistency points are established by the load utility. They are very useful when it comes to restarting the load operation. You can restart the load from the last successful consistency point.
During the build phase , indexes are produced based on the index keys collected during the load phase. The index keys are sorted during the load phase, and index statistics are collected (if the statistic option was specified).
In the load phase , the utility only rejects rows that do not comply with the column definitions. Rows that violated any unique constraint will be deleted in the delete phase. Note that only unique constraint violated rows are deleted. Other constraints are not checked during this phase or during any load phase. You have to manually check it after the load operation is complete. Refer to Section 12.5.7, Validating Data Against Constraints, for more information.
During the index copy phase , index data is copied from a system temporary table space to the original table space. This will only occur if a system temporary table space was specified for index creation during a load operation with the read access option specified (see section 12.5.2.5, Locking Considerations During a Load).
DB2 


