DSXchange

First thing that comes to my mind in a parallel job is using RCP. Any specific reason to prefer server jobs? Is there an equivalent property to RCP in server jobs?

What happens when you try to insert the records? Did you check if there is a record in the table with the same keys as the update record?

I believe you will find the scripts under IBM/WebSphere/AppServer/bin. I think the script is called "startServer.sh" and one more "Status.sh" to check the status of the WAS.

Uncheck the box which says something like "Log warnings for stages that dint finish ok" in the sequence properties.

Looks like a problem with the external c++ routine to me. What does this routine do? Copy paste the code here.

Could it be a dead lock on the record in parallel mode? Try hash partitioning the data on the keys and run it in parallel.

The 2GB limit on the DataSet is imposed by the operating system constraints and not a DataStage constraint (mostly for the data files in the resource disk). Number of rows is never a problem, and the size is also not a constraint as long as it is not constrained by the OS.

Modify stage does not create extra processes. It runs on the output link of the previous stage. So it would be slightly faster than a transformer. AFAIK, the RCP option in each stage also creates a modify operator to propagate or drop the columns.

BugFree, You would get the result but it will be 6 records instead of 3 . Ashish, if you are sure that the records are already in the required order then you can generate a surrogate key for each of the links using a Row generator stage and then join the data on this field.

Is that the complete error message?

Search on "will partition despite the preserve-partitioning flag on the data set on input port" for the first warning. As for the second warning you can use the property in the Aggregator stage called "Default to Decimal output". And change the precision to match the output field.

DataStage inserts a buffer operator on each input link to the funnel stage. So for 100 input links, 100 buffer operators, 100 processes. If it is a continuous funnel try using the APT_BUFFERING_POLICY set to "NO_BUFFERING".

Yes. Transform routines in Basic Transformer. Before/After subroutines in Job properties or in basic transformer before/after subroutine tab.

Any specific reason why you don't want to use the Pivot stage? A custom Build Op stage can be used, but involves a lot of coding.

In the designer, right click on the job name and select delete.