DSXchange

Have you tried TrimB(TrimF(L_Agg_Out.street)) ?

The only "solution" involves managing your expectations (in terms of elapsed time) better and realizing that rows/sec is an almost entirely meaningless metric in this context.

Rows/sec doesn't matter if you're not watching the job. So don't watch it.

I can only suggest that, if you really did "complete the installation and configuration of DS8 as per the installation document" then the document is deficient.

Read Vincent's blog on the issues he had getting DS8 to install.

I, for one, plan not to be an early adopter of this version!

It's a guideline, not a rule. Try to match, rather than overload, your hardware capabilities.

It's possible; it's not best practice.

Parameterized schema files for the source files, and rely on RCP and parameterized table names for the loads.

No guarantees - you may need to tweak some other aspects.

This is expected behaviour if write cache is used. Running data into the cache is very fast. Once all the rows are processed, the cache must be flushed to disk. However, the clock keeps running during this process, so the rows/sec figure (which I have already decried as meaningless) appears to fall....

Never exceed 16KB (separation 32) unless your disk controller moves more than that volume at a time.

COUNT.SUP (or COUNT SUPPRESS) suppresses display of the "1 records listed." message.

We do not do "urgent" or "as early as possible" here. If you want "urgent" sign up with your support provider for premium service and learn the true cost of "urgent". What you seek is definitely possible, and was answered in another thread as recently as today. Essentially you need two Hashed File ...

First prove that it is a hashed file stage that is generating the error. Hashed File stages are passive stages, therefore do not generate processes, therefore can not terminate, abnormally or otherwise.

Second, it's hashed file, not hash file.

How long is a piece of string? Pick a number that sounds small. For separation 1, say, < 50 bytes would be small. Roughly, if you can store more than nine records in 80% of a group, the records are small. It is not really quantifiable more exactly than that.

You can also use COL.SUP but *technically* the full words are defined for SQL. COUNT.SUP might also be useful.

Non-concurrent does not prevent others from using different sequences. Concurrent allows multiple processes to retrieve keys from the same sequence.

So why have you marked the topic as Resolved?

If it IS resolved, please post how it was Resolved, to aid future searchers.

Column Generator stage, cycle with only the one value '-'.