DSXchange

Ford Prefect?

See your DBA, this is all Oracle.

TRUNCATE would have bypassed that trigger.

Sure, either use a Sequencer or Batch job to control your transformation job. The easiest to use is a Batch job, because it writes all of the job control for you. All you need to do is get that file into a variable. Here's one easy way: UNIXcmd="cat /wherever/yourfile" Call DSExecute(&...

Get your hands on the CD, there's a tutorial and boatloads of documentation comes with the client install. After that, experiment and play. We're open 24x7 if you need clarification.

kollurianu wrote: If i donot need the updates immediately , then i need not set transaction size to 1 or array size to 0 , is that correct.

Yes, and you can use row buffering and IPC stages.

Could be a bug, but have you recompiled the job incase of the previously stated mismatch?

The delete is absolute, it will work.

Sure, in your design replace OCI with HASH and that works given the previously stated conditions.

Samwise Gamgee

Hash files work the same way, no read or write caching and no row buffering or IPC between stages.

It's supposed to, it uses rollback space to perform the delete. Truncate doesn't and it's nearly instantaneous on non-partitioned tables. If the delete failed do to inability to get enough rollback, the query should have died causing the job to blowup. Are you using ODBC or OCI? If OCI, unless you d...

Off-hand, this part of the message gives me concern: While creating "DSColumnC" object of type "" The blank object type makes me think maybe some internal metadata is screwed up. This stage historically had some issues where derivations were lost. Also, the metadata is important. Just for giggles, d...

Commit after every row using 1 as the commit count. For upstream stages to wait for downstream stages to finish with a row, you must disable row buffering and not use IPC stages between the reference and the write.

Methinks your job design doesn't match the compiled code.

The clear table does a "DELETE FROM TABLE" command. A TRUNCATE would be better.

But, your data sounds like it updated existing rows or inserted.

Your requirement is that many, maybe millions of rows of data, across multiple tables, have to instantly appear. Give a large volume of data, there is no database mechanism available to do such a thing, so therefore no ETL tool can accomplish the same. For small volumes, you can group a small set of...