DSXchange

The number of rows sent along the link is not necessarily the number of rows that end up in the hashed file, because duplicate key values effect a destructive overwrite.

That's not the complete error message.
Please post the complete error message.
Are you REALLY on version 4?

This is true only if the lock is owned by a non-defunct process.

There's always retrophrenology.

It depends on context. There are contexts where I would prefer a sequential file over a hashed file. Can you be more specific about the circumstances in which you assert that a hashed file is to be preferred?

I don't need one!

The variants on the "lookup failed" rule indicate that you are getting no matches. The Continue rule mimics a left outer join, hence the returned NULL fields.

Can you execute this query successfully outside of DataStage with appropriate value substitutions?

Not quite.

What I mean is a new column, generated in a Column Generator stage (that is, not in the file itself). The column name can be anything you like; the value in it must be 'X' for every row.

It might be the clash of having CDR_DIR defined in both locations. Try removing it from the Windows environment variables and relying solely upon the ones defined in Administrator (from which $PROJDEF asserts that it should get the value if not overridden).

It's not partitioning. Look at the score. Note that DataStage has inserted some tsort operators (and probably some buffer operators also) on the inputs. So, if you don't specify sorting, DataStage will insert sorting. You might prefer a Sort stage so you can tell it "don't sort (previously sorted)" ...

By "they" I mean operations staff, who may shut things down in order to take backups, etc.

Multi-instance jobs are intended for almost precisely the situation you're in, perhaps also for continuously-running jobs or jobs that use the RTI components (and might be invoked from lots of places). I agree that non-selective log purge is an issue, but you can code around this or just rely upon a...

No. Determine the owner of the lock and its cause. Eliminate the cause.

There is also an UNLOCK command, or the DS.TOOLS menu. Search the forum for more details.

If DB2 UDB server is on the same machine as DataStage server, set Array Size to 1. Otherwise set Array Size to some multiple of (packet_size / row_size). Is there a large number of deletes? If not, set rows per transaction to 0, so that all the deletes form one transaction.

I prefer uvbackup because it gets all those files, like DSParams, that export misses.