DSXchange

Not sure about concerns, seeing as how you don't really have a choice. I'll have to let someone with experience in that arena answer, I've done very little SQL Server work over the years. I believe about the only thing you wouldn't have not being on Windows would be a direct bulk load capability. No...

You would have a "dedicated SQL Server stage" if your DataStage server was installed on Windows. For UNIX you would need to use ODBC.

Why? If they've been added because someone thought they would magically improve performance, that's not what they are for. I'd suggest removing them from the job, 2 million records is not a large amount and I can't imagine turning off the troublesome buffering would make any kind of significant chan...

Are the two "link" stages actually doing anything in the job?

So it seems you may have found this post if you know about skipping Domain authentication. Did you read their resolution and try to verify that your permissions "all the way down" are correct?

Get the third field of a "_" delimited string, followed by the first field of a "." delimited string.

Try passing in a full set of credentials: server, userid, password, domain...

OK, I was thinking a mismatch of the total count was all that was needed, but if it is counts of matching and non-matching records per table then as you said it's not possible without some amount of metadata.

And how did you answer the questions?

So, "compare tables counts" rather than "compare tables data" it would seem. What have you tried? Seems like parameterized table names, selecting a count from both the source and target tables, joining those two inputs and comparing the results would be fairly straight-forward.

qt_ky wrote:It sounds like your partitioning may not be correct.

Agreed. Anytime I hear "different results" running the same data through multiple times I suspect faulty partitioning. And I too would be curious what happens when the job runs on a single node.

Yah, there's plenty of information out there for the google'n, including articles like this on Tookbox. Vincent McBurney blogs there and has all kinds of DS goodness available including his three part Dickens inspired "Ghosts of DataStage" posts, which the linked article references.

I made that assumption but have no way to test it.

That's true... it does seem a little backwards.

Ah, yes... an empty string is not the same as a null. You could probably combine those two checks into a single one for length=0 but what you've got is fine as well.