DSXchange

Although I'm not sure what you are asking, why don't you use a VarChar(64) column which is more than sufficiently large.

So you can point the same job to a 9i or 10g source and see speed differences; but I believe it much more likely to be because of network bandwidth or database server load rather than due to DataStage. Have you checked this aspect? Also, if you are writing to the DB then it could be because you have...

If you open up the column import stage, are any attributes red or empty? This stage is a bit tricky to use and I can certainly imagine that some errors don't display with more details.

In TOAD you aren't limiting the display output, but there is no way that a valid DECIMAL(38,10) number such as '12345678901234567890' will fit into a VarChar(12).

Remove the HASHFILE1 lookup in a job copy, does that still hang? Have you turned on preload file to memory in the HASHFILE2 and is it a large reference file?

That message is always displayed as an effect of another error, it just means that the system is trying to clean up after itself after an error has occurred. Check your logs for the actual error.

The shared container itself doesn't have that facility directly. A BASIC transform stage allows you to call an after-stage shell script executed only after the last row is processed and could be added to your job, although I recommend against using BASIC transform stages.

Sequence jobs allow you to call shell scripts directly. Server and Parallel jobs usually use the before or after job commands to execute shell scripts.

DataStage doesn't have a CLOB datatype, but databases do. Which database and which interface are you intending on using CLOBs with? Some are supported, some aren't supported and work and some aren't supported and won't work?

I was too slow today and not thinking! Thanks for correcting that Ray.

I've marked my original post as non-premium. In Server the OCONV() function with the MB0X code is a simple way to solve your problem.

What values can you expect in the source FIELD-1? If just the digits 1-10 then a simple transform IF-THEN-ELSE construct would do the trick. Also, must this be a parallel job, it would be an almost trivial excercise in a server job or, at worst, a BASIC transform stage in PX job.

Have you done any repartitioning of the data in any of the stages leading up to the join stage? Note that you might be doing implicit repartitioning; and some stages change the output sort order and that might affect a 4 node but not a 2 node hashing partitioning.

The "load" method invokes a very different procedure than the insert/update/upsert method; for Oracle it executes the SQL Loader externally to DataStage. As part of your job log you should see references to the external files and also to the Oracle load logs. I can't recall in which direct...

Actually, turning off the row buffering will work - or give another error. If you get the same error then you still have buffering enabled.