DSXchange

If your Upsert specifies "Insert before Update" the insert will always work because there is nothing preventing it. And you get duplicate keys. Try changing the order to "Update before Insert".

I doubt that this is possible using Notification Activity. However you could use other activities (Execute Command or Routine) to encrypt the email body and then invoke sendmail() to send it.

In the past I've always doubled up - the Invocation ID governs the instance name, and I also pass the same value as an explicit job parameter. This approach works.

Make sure that this entry appears in the project's uvodbc.config file. In your UniVerse stage, specify localuv as the DSN.

That will be interesting to see. AES-128 encryption should be impossible to break in a reasonable amount of time using the kind of hardware that most Information Server sites use.

Short answer - a different total machine load on Friday.

My guess is that the practice questions you faced were from the 8.0 test but you have passed the 8.5 or 8.7 test.

How are you describing the "UniVerse" database in .odbc.ini and uvodbc.config ?

If Craig's suggestion doesn't work, try an explicit Failure trigger.

Welcome aboard.

If you are using version 8.7 or later and if you have Information Analyzer licensed then you can use the Data Rules stage to perform a matches_regex test on the URL. This is by far the easiest way to check URLs, as there are so many valid variants.

In fact it almost certainly breaches the licence conditions of the author's Information Server installation (the part about reverse engineering).

I make that statement with no knowledge whatsoever of the tool.

Presumably this utility does not work with passwords encrypted strongly using the encrypt.sh (encrypt.bat) utility?

Looks like it may be ignoring the Field Delimiter property. It shouldn't be.
Why not try defining the format as variable length (all strings are VarChar) and using a Transformer or Filter stage to trap the rejects? This has the advantage that the filtering is done in parallel.

"Suddenly", eh? What changed?

The whole point of parallel jobs is that you do NOT need to redesign when running on a different number of nodes.