DSXchange

An educated guess. There's a foreign key constraint called ITEM_FOREIGN_KEY and a column in the insert statement called ITEMID.

There is so too! A domain user is for the IBM Information Server domain, also referred to as the deployment domain.

Entries are probably still buffered for writing when you're executing your after-job subroutine. Add a SLEEP of at least two seconds duration. Even better, use a Routine activity after the Job activity in a job sequence; in an after-job subroutine you can never catch the "job finishing" message.

Trim() can never generate "".

I would use one job, with the Join stage performing a left outer join (with X as the left input). Downstream of the Join stage a Filter or Transformer stage would segregate the input into two streams, one receiving all rows (unconstrained), the other receiving only rows where the key column from Y i...

That's why we have Code tags!

The burning question is which user ID and password do you have to give when the -domain option is used? I don't know. Experiment.

Get the end users (management) to complain about the slowness of the ETL process, and let it be known that SELECT privilege to the table(s) would be a solution.

date_from_timestamp() in a Modify stage, TimestampToDate() in a Transformer stage. Both are documented in the Parallel Job Developer's Guide

What a wonderfully useful Subject! (Not)

Forget rows/sec - it is one of the most meaningless metrics ever devised. What you are doing in writing to this hashed file is random writes to 14,765,651 different disk addresses. You can't guarantee that these are contiguous, either. Striping, RAID and other high-availability technologies only exa...

ITEMID value supplied does not occur on the ITEMS table.

Definitely use QualityStage. This is precisely the kind of thing it does, with the added benefit of configurable levels of uncertainty in the matching phase. You can use NYSIIS and/or Soundex (forward or reverse) or not, as pleases you.

Good luck finding a tool. Dynamic metadata is not supported in DataStage server jobs.

There are workarounds, of course, such as generating entire SQL statements into job parameters. But not on a row-by-row basis, which is what your question seems to require.