Aggregator stage, IPC

[Bala R]

This is my job design

Code: Select all

                           Hashlookup
                              |
SourceOCI9i--> Agg--> IPC--> Xfm--> Agg--> IPC--> Xfm--> TargetOCI9i
 

When I thought i was going to increase some performance by increasing the buffer size in the IPC stage properties from 128 to 256, the second aggregator stage started aborting. The logged error message was ' Row out of sequence '. If this is about sort order, I had the data sorted when I select from the DB and had again set the same sort order in the aggregator inputs too. The two aggregators differ only by the grouping.
Somebody could point me what to understand from this, please.

Thanks for your time.

[ray.wurlod]

Ask this of Ascential/IBM. In particular ask whether rows are necessarily buffered in the order they are received.

[ray.wurlod]

Is the first Aggregator stage doing anything you couldn't do with a GROUP BY clause in the SELECT statement?

[Bala R]

Is the first Aggregator stage doing anything you couldn't do with a GROUP BY clause in the SELECT statement?

The source OCI stage and the first aggregator together do a vertical pivot. I had used CASE expression to shift the column value and then sum the rows in the aggregator to pivot.

[ray.wurlod]

That's cool

It's just an obvious thing to ask.