Rollback segments explode with few records

[dba_fabio]

Hello to all, this is my first post here ...

this is my problem, i have an Oracle 9.0.2 on Sun Solaris with an instance that load data from 2 different source.
First source is a SQLLOADER script that every 10 minutes load about 300.000 records.
Second source is a Client (datastage) that via network load about 50.000 (30Mb) records 1 time a day.
The rollback segment is able to load the 300.000 records without any problem in a single segment of about 200Mb.
The problem is that sometimes the rollback segment that load the 50.000 records explode (9Gb).
I have to tell that the origin data for the 50.000 records no contain errors or strange caracters, no concurrent transaction working on the same tables during load.
My question is, that is a case in your experience in which a situation like mine repeated ?

thanks a lot in advance ...

[davey23uk]

what query are you running, id suggest very bad code

[ixion]

Sqlldr is probably committing every x number of records. So the RB segment space doesn't need to be so large since you get a new transaction every so often.

I'm betting your datastage application does not commit till the very end. In that case you need to check the alert log for a space error on the Rollback segment tablespace. (ora 1650 )... Increase the size of the Rollback segment tablespace and/or add datafiles as needed.