Sql Tracing question
- From: Joop Gijsbers <jg_dba@xxxxxxxxx>
- To: "List, Oracle-l Freelists" <oracle-l@xxxxxxxxxxxxx>
- Date: Thu, 27 Sep 2007 16:52:04 +0200
Because of an application problem - unexpected crashes with Forte
Composer - i did today tracing activities on the (Oracle 9.2.0.8)
database on an AIX 5.3 node.
The first "problem" is that there is no (ORA-/TNS-)error at all and the
second problem is even more that it is very difficult to identify the
Unix Process ID, or the SID or Serial# of the session,- the crashes are
unexpected and very soon after starting the session and there are many
users at the system in their session can occur the crash - so i tried
some tracing on instance level.
I tried:
alter system set TRACE_ENABLED=TRUE
but this generated no extra logging at all (maybe not strange, because
there is no error);
I tried also sql*net tracing:
TRACE_TIMESTAMP_SERVER = ON
TRACE_LEVEL_SERVER = 16
TRACE_UNIQUE_SERVER = ON
that generated a lot of information , but it gives the forte specialists
not the relevant information (as the sql statement at the moment of the
crash of the composer)
Next step is reproducing the problem on a isolated system - this tracing
was on production - , so that i can isolate and recognise the Unix PID
and so the PID and Serial#, so that i can trace that session.
But is there another way to trace the bulk of all instance/database
statements, that i overlooked?
I hope i made my question clear, i am not sure :-)
Much thank for any comment.
Joop
--
http://www.freelists.org/webpage/oracle-l
Other related posts: