Find errors logged from oracle batch job...

Question

Hi,
I have a batch job that calls a procedure.. code is as under
begin
  sys.dbms_job.submit(job => :job,
                      what => 'begin
                                   delete_stale_data;
                                    end;',
                      next_date => to_date('29-06-2006 05:33:28', 'dd-mm-yyyy h24:mi:ss'),
                      interval => 'sysdate+(30/1440)');
  commit;
end;Now, how can i get to view the error messages if
1) procedure runs into an exception and hence job is not executed
2) if job runs into some problem due to which it is not completed..
Thanks,
JP

Answer 1

"Viewing errors" is the wrong term to use - it assumes that there are devices like STDERR and STDOUT in Oracle that is written to when an error occurs.
There are no such devices in Oracle. Oracle itself will record system related errors (and messages) to the alert log of the database instance. Custom PL/SQL code can also write to the alert log using an undocumented call - but it is a very poor idea as the alert log is for the DBA to deal with system errors. Not with someone's application errors.
The correct method is to implement an application logging PL/SQL interface (package) that allows applications to log errors and warnings and messages (including debug stuff). The basic method of this API implementation is to create a log table and write messages (from applications) to this table using autonomous transactions. The API call can also record the system date/time, the current PL/SQL stack trace, client session details, etc.
Using such an API, you would schedule the job as follows:
Method 1:
The job becomes broken after 16 repeated failures (the exception is re-raised in order for DBMS_JOB to deal with it).
sys.dbms_job.submit(job => :job,
what => 'begin
delete_stale_data;
exception when OTHERS then
APPLOG.CriticalError( 'Error occured running DELETE_STALE_DATA', SQLCODE );
raise;
end;',
next_date => to_date('29-06-2006 05:33:28', 'dd-mm-yyyy h24:mi:ss'),
interval => 'sysdate+(30/1440)');
Method 2:
The job will never break. The exception is supressed and DBMS_JOB will not know that any errors occurred:
sys.dbms_job.submit(job => :job,
what => 'begin
delete_stale_data;
exception when OTHERS then
APPLOG.CriticalError( 'Error occured running DELETE_STALE_DATA', SQLCODE );
end;',
next_date => to_date('29-06-2006 05:33:28', 'dd-mm-yyyy h24:mi:ss'),
interval => 'sysdate+(30/1440)');