DSXchange

Another thought.

Signals are the operating system's way to notify itself, and any executing process, that something has occurred. Some signals are innocuous, others indicate trouble. You will read on the forum about SIGSEGV (segmentation violation - an attempt to access memory that doesn't exist or not in your space...

"Phantom" is the term that DataStage uses to mean "background process" (in UNIX, they are specifically implemented as nohup processes so that they can continue to execute even though their parent process logs out). Apart from daemons (executables such as dsdlockd, dsrpcd) and agent processes (dsapi_...

Exactly how did you modify QETXT.INI file? That's what folks searching for this answer will want to know!

Well, at least now you can create a new dummy job, cut-and-paste to there, and retrieve when needed.

Take a look on your DataStage CD, in the Packages folder. If it comes with the software, it will be there.

Welcome aboard! :D Your answer is "it depends". If the resources are there to do a ten table join in the database, do that. It will mean that DataStage only has to process the result set, and processes no redundant data. In general this is the preferred approach. The other approaches are: (a) to hav...

This isn't DataStage's problem, it's yours. It would happen exactly the same if you were doing a lookup against a database. You have two key columns, you must provide two values. Where do you propose to get the second value from? If it's a constant, fill in its value (strings must be quoted) in the ...

Try using SQL*Loader rather than INSERT statements. Try using arrays and multiple rows per commit. (What are your current settings on these, and how did you choose them?) Switch the job to multi-instance, and have each instance run a separate subset of the data, if possible. The instances can run in...

If the Oracle column is a date column, you can just include the word SYSDATE in the SQL where the value for that column is being provided. No need, in that case, to deliver a value from DataStage, therefore not need to worry about conversion and so on. For example: INSERT INTO tablename (col1, c...

Array size 1 is so that the correct warning can be associated with the correct row. If you send 1000 rows, which one(s) generated the warning(s)? Better is an approach that avoids warnings altogether. Then the fact that a warning has occurred properly alerts all and sundry that there may be a problem.

You will only be partially OK. As I said, you also have to create a parameter in each of the jobs. That's more difficult to do in an export file, though not impossible. Exactly how it is done depends on the style of the export file, but each parameter is individually tagged, so all you have to do is...

I've intentionally stayed off this thread until now. Ken has been giving you not only good advice, but best practice too. There are many sites maintaining Type 2 slowly changing dimensions with DataStage hashed files; it does work; it works well, but it has to be done correctly. If you have seen som...

Since writing to a hashed file when its file system is full, could the DBwrite routine be modified to include a pre-check for sufficient available space in the group? This is actually in there already; it's what triggers the group extending into overflow. How difficult would it be to include a test ...

There's not much more I can add to that.

As a general rule, the more I/O channels you can provide, the better your throughput will be.

Make very sure your hashed files are on file systems that won't become full. They break if you try to update them when this is the case.