DSXchange

Install Oracle client on the DataStage server.
Configure Oracle client to "see" the database instance.
Done.

I always understood that the devil was in the details...

Try a search on just "code 134". There is a dozen hits, two of which are marked as Resolved.

No.

It's still more efficient for hashed files that don't need to contain 2GB of data to use 32-bit internal addressing, not least because the 32-bit pointers require less disk space. And there's three of those per record.

... but maybe not ALL trailing zeroes. Think it through. You will need to keep six numeric characters.

Run on a single node, or generate numbers that are unique across the domain. Search DSXchange for techniques and formulae.

Welcome aboard.

Is either of the SKG stages configured to use state file? If so, are they both attempting to use the same state file?

Good points.

You can perform a Reference Match using QualityStage and UNCERT as the match rule.

Bridges are installed on the client. Have you created your client machine as a Metadata Interchange Server?

As well as being multi-instance the job must also have been successfully compiled.

No, Array Size sets the maximum number of columns that can be generated (pivoted) onto an output row. If there are fewer input rows than this nulls will be generated. By this means all output rows have identical record schema. You can deal with that downstream to suit your own requirements.

You still need the "#" markers for environment variables, and these need to have been loaded into the job as parameters. For example:
File path =#$EVENT_XML_PATH#/#$XML_NAME#.xml

Have you read the help on Array Size? I think you'll find it informative.

You haven't mentioned whether this is a horizontal pivot or a vertical pivot that you're trying to accomplish.

Fork join will work without creating a deadlock. To ensure this DataStage inserts buffer operators on both inputs to the Join stage, which you can see in the job score.