Dataloader/MP Reference Manual
Table Of Contents
- What’s New in This Manual
- About This Manual
- 1 Introduction to DataLoader/MP
- 2 DataLoader/MP Components
- 3 Running DataLoader/MP
- 4 Specifying File-Related Options for DataLoader/MP
- 5 Creating a Customized Version of DataLoader/MP
- User Exits
- User Exit Descriptions
- BUILDKEY
- CHECKARG
- CONVERTIT
- DELETEIT
- DONEWITHTRANSACTION
- EXITSDESCRIPTION
- GETNEXTRECORD
- INITIALIZE1
- INITIALIZE2
- INSERTIT
- INSTRUCTIONS
- MISC1, MISC2, MISC3, and MISC4
- NEWTRANSACTION
- NEXTINDIRECTFILE
- SKIPPING
- STATISTICSTIME
- TERMINATING
- T0330U00_DEFAULTEXITS_C
- T0330U00-DEFAULTEXITS-COBOL
- T7900D41_DEFAULTEXITS_C
- T7900V00-DEFAULTEXITS-COBOL
- UPDATEIT
- Default User Exits
- DataLoader/MP Library
- The MAKE Routine for NM DataLoader/MP
- The MAKE Routine for Nonnative Mode DataLoader/MP
- 6 DataLoader/MP Examples
- 7 Recovery Strategies
- A Error and Warning Messages
- B Processing Flowcharts
- C C-Only Error Functions
- Index

DataLoader/MP Examples
DataLoader/MP Reference Manual—424148-003
6-6
Multiple Source Parallel Loading
The following commands use these parameters:
•
-E gives each of these downstream DataLoader/MP processes different files (err1,
err2, ...) in which to record their execution time errors.
•
-I directs each DataLoader/MP process to get its input from the initial
DataLoader/MP process, $dbl0.
•
-X directs the processes to call a series of user exits indicated by the string “CIU”:
°
“C” directs the process to call the user exit CONVERTIT.
°
“I” directs the process to call the user exit INSERTIT, which attempts to insert
the converted record into the database. If the insertion is successful, the
processing for that record is complete, and the process reads the next record.
°
If the insert is unsuccessful, “U” indicates that the next user exit, UPDATEIT,
should be called to update the corresponding row in the database. If the
update is successful, processing for that record is complete, and the process
reads the next record. If the update is not successful, a message is sent to the
error file indicating that no user exit was successful with this record, and the
process reads the next record.
2>dbl1 /NOWAIT/ -E=err1 -I=$dbl0(recform=fb<recsize=200>) &
-X=ciu
3>dbl1 /NOWAIT/ -E=err2 -I=$dbl0(recform=fb<recsize=200>) &
-X=ciu
4>dbl1 /NOWAIT/ -E=err3 -I=$dbl0(recform=fb<recsize=200>) &
-X=ciu
...
5>dbl1 /NOWAIT/ -E=errn -I=$dbl0(recform=fb<recsize=200>) &
-X=ciu
Multiple Source Parallel Loading
This highly parallel load scenario with write operations uses only two simple exits: one
to build the key from a raw record (the BUILDKEY exit in the initial DataLoader/MP
process) and one to convert a raw record to a suitable format (the CONVERT exit in
downstream DataLoader/MP processes).
LOAD requirements involve the following:
•
You must start with one initial DataLoader/MP process for each data source.
•
You must have one SQLCI LOAD process for each partition and one
DataLoader/MP process that is the input to each LOAD process.
Figure 6-3 shows multiple input streams. Suppose you are replacing all of the data in a
partitioned database.
In this example, you know the key partition boundaries and give them to the initial
DataLoader/MP process in a simple edit file. You would include a user exit in each
downstream DataLoader/MP process. The user exit converts incoming records to the