DSXchange

Ray does tend to be a pretty good guesser, though.

Is a warning or other message being logged? Do you have constraints on the output link? What exactly do you mean by "rejected"?

No-one can know the why unless you provide more information on what you are doing or seeing, right now we'd only be guessing.

vinsashi wrote:reset and rerunning.

Re-read what I posted. Only reset it if you want it to restart from the beginning.

No.

What is the scale on your target field?

Interesting... two inline stages is a very cool idea for that situation, not sure I would have thunked of that.

The message is pretty clear. Have you attempted to "ensure" what it asked you to ensure?

The appropriate rtype would handle that, up or down. Use that in the right function for the job, DecimalToDecimal() perhaps. Check the Parallel Job Developer's Guide for Type Conversions.

Build them separately and then concatenate them together afterwards in the proper order. You may also be able to build the records manually as one varchar field and make sure they write out in the proper size and format.

Well, "too many" is a bit of a red herring as it is all about how much space their names take up rather than strictly the number of files.

We're not privy to the inner workings of the stage, hence the suggestion to take this to support.

Communication area for the "PHantom" (background) processes all jobs run as. They're just text files, poke around and see for yourself what's in them.

Well... we're be here when you come back. I sincerely doubt that would be all the fix you need but time will tell.

Right, in a RDBMS table it makes absolutely no difference where indexed fields are in the table, including the PK fields. As noted, the only thing that seems to care is the wetware, the people.

No, there's no rule as to where or in what order they need to be declared in the table. That's a people thing.

So, one job running on 6 nodes brought your server to its knees? And 2 hours to process half a billion records is 'poor performance'? Me, I would be happy with the 2 hours, especially considering the amount of hardware you (apparently) have to throw at the problem.