ODBC stage loading Oracle funny characters

Post questions here relative to DataStage Enterprise/PX Edition for such areas as Parallel job design, Parallel datasets, BuildOps, Wrappers, etc.

Moderators: chulett, rschirm, roy

Post Reply
SBSA_DW2
Premium Member
Premium Member
Posts: 37
Joined: Fri Jul 25, 2008 1:24 am

ODBC stage loading Oracle funny characters

Post by SBSA_DW2 »

Hi Guys

Today a colleague asked me to look into a job for her that

(1) takes a flat file and loads using ODBC stage into an oracle database. This works fine, 100%, no errors, no warnings just the green green icons of success. The ODBC stage uses a defined DSN in the .odbc.ini file. 2 columns in the table, column1 is double(15) and column2 is varchar(50)

the data is in the table...

(2) Job 2 uses an ODBC stage reading from the table loaded in step (1). The double column comes out fine, but the varchar is filled with crazy characters.. characters you wouldn't trust with your sister.. shady.. real shady (unreadable cahracters).

When running the job to extract from the table the error: "APT_CombinedOperatorController,0: Invalid character conversion found converting to ISO-8859-1, substituting" appears.

I have searched the forum for the error and there is reference to NLS and code pages switching, but we don't have NLS enabled.

The kicker is we can load to the table but can't read from the table using the same DSN from .odbc.ini. :shock:

I'm going to look at including the oracle drivers installed in the LD_LIBRARY_PATH so that we can use the oracle plug-in stage over the odbc stage.

any suggestions welcome
chulett
Charter Member
Charter Member
Posts: 43085
Joined: Tue Nov 12, 2002 4:34 pm
Location: Denver, CO

Post by chulett »

What is the character set of both the source file and the target database? If you use another tool, say like Toad, to view the table are the values in the string fields still 'corrupted'?
-craig

"You can never have too many knives" -- Logan Nine Fingers
SBSA_DW2
Premium Member
Premium Member
Posts: 37
Joined: Fri Jul 25, 2008 1:24 am

Post by SBSA_DW2 »

thanks Craig
I'm not sure on the character set but the data is viewed fine via TOAD.

the plot thickens..

We have now created a Server version of the job on the same server, used the same ODBC settings and the data reads fine. This is a workable workaround but now the question.. what is the difference between the two stages and why would there be an issue..

thanks!
chulett
Charter Member
Charter Member
Posts: 43085
Joined: Tue Nov 12, 2002 4:34 pm
Location: Denver, CO

Post by chulett »

It's going to come down to character sets, so to take this any further you'd need to know what that is in your Oracle database and your PX job before anyone could help.
-craig

"You can never have too many knives" -- Logan Nine Fingers
SBSA_DW2
Premium Member
Premium Member
Posts: 37
Joined: Fri Jul 25, 2008 1:24 am

Post by SBSA_DW2 »

RESOLVED - turned out the dsenv and lib links weren't set up properly.
Post Reply