DSXchange

Have you checked for the user ID that runs DataStage jobs? Look in the second event in the job log ("environment variables").

What else IS there?

To ensure referential integrity the ONLY technique (irrespective of how you code it) is to verify that the value occurs in the parent (dimension) table.

The last part of Chapter 2 of Parallel Job Developers Guide outlines the restrictions.

The environment variables in Administrator are listed in the DSParams file in your project directory on the server. Edit this file to add SHLIB_PATH and your issue will be resolved.

There is no a priori knowledge of the size of the hashed file when "allow cache" is set at design time. If the hashed file is too big, then a warning is issued, and the hashed file is accessed on disk. But this can not be ascertained until run time.

There's nothing in the BAT file to tell it to come back, so why would it?

The problem is that it can't find an AT command. This is weird, because AT is internal to DOS.

Sure it's slow - update always is, particularly if there are indexes and constraints.

Why not use bulk loader?

If the column(s) on which you are sorting are indexed, then it will be faster to sort in the database. This is because index values are stored in sorted order, so to "sort" simply involves scanning along the leaf nodes of the B-tree index (or straight through a bitmap index).

In Director, the server time is displayed on the right-hand end of the status bar. This can be useful if the client and server are in different time zones.

Your use of "stuck with" implies a negative attitude. By all means code it yourself, if you're prepared to maintain your code indefinitely into the future, and can realize at least the flexibility and performance that is possible with DataStage. DataStage has now been around for more than 10 years. ...

The solution is to prevent it (the error), by not providing an invalid argument (property) value for connecting to DB2.

Welcome aboard.

And you want to perform this split with a TX map? Surely it would be easier using a UNIX command such as split or awk - search the forum for examples.

You can disable combination at an individual stage level, rather than globally for the job. That might ameliorate your performance woes a little.