Page 1 of 1

Strange behaviour in Transformer

Posted: Tue May 11, 2010 6:17 pm
by nilesh_kant
I have a job which reads from a sequential file and loads into a Oracle database through a tranformer. The job gets aborted saying "Abnormal termination of stage load..t1XfrmFilterRows detected" . There are no error messages only two warnings. No conversion done in transformer. Number of records are over 10 million.

Posted: Tue May 11, 2010 8:33 pm
by chulett
Reset the aborted job and let us know if a 'From previous run...' message gets added to the log. What function is the transformer serving?

Posted: Wed May 12, 2010 10:14 am
by nilesh_kant
I get only a single statement in the log saying Job has been reset and no othe logs. Jobs run normally when re-run.

Also, it is a server job and not a parallel job.

Posted: Wed May 12, 2010 10:59 am
by priyadarshikunal
Please post the warnings. Also check the messages in message handler for suppressed warnings.

Posted: Wed May 12, 2010 11:39 am
by nilesh_kant
Hi,

Abnormal termination of stage FactLLoad..t1XfrmFilterRows detected..

The above is the message I get.

Posted: Thu May 13, 2010 2:43 am
by priyadarshikunal
the message you posted generally comes along with some other warnings(or may be phantom errors which you don't have). Else, there is no way you can tell why the abnormal termination was caused.

Posted: Thu May 13, 2010 12:58 pm
by nilesh_kant
Yes I did get another warning. "DMDayLoad..DWFDay.inDWDay: DSD.BCIOpenW results of SQLColAttributes(PARENT_ROW_SEQ_NUM) gave MetaData mismatch
COLUMN.TYPE Expected = BigInt Actual = Numeric".

But this warning we get often and job doesn't get aborted.Only sometimes we get the job aborted.

What are phantom errors and any way to look at the error message.

Posted: Thu May 13, 2010 1:08 pm
by swarnkar
nilesh_kant wrote: What are phantom errors and any way to look at the error message.
The word Phantom is used to refer the background processes.

Posted: Thu May 13, 2010 1:24 pm
by asorrell
If you will look in the directory that holds your project, there will be a sub directory called &PH& - it will contain very small files that are used to hold output from the DataStage jobs that run in the background. Run your job again and let it abort, then go look at the very last file in that directory, it should contain additional errors related to your job.

Three notes:

1) The &PH& directory should be emptied periodically (manually) when nothing is running. IF you don't, it can cause system problems.
2) In UNIX the & is a special character - you'll need to quote the directory name to avoid problems, ie: cd "&PH&"
3) The projects are kept under the InformationServer/Server/Projects directory.

Posted: Mon May 17, 2010 12:18 pm
by nilesh_kant
Hi,

Any help on the below topic will be appreciated...

Please tell me the path in Unix where to look these files.

The job gets aborted sometimes not daily. It runs fine after a re-run.

Posted: Mon May 17, 2010 12:53 pm
by ray.wurlod
No idea. It's your server, not ours. You can use the Administrator client to find out - even if you are not the administrator; when you have a project selected the pathname of that project directory is displayed at the bottom of the window.