Re: 'alter session set sql_trace=true' bug or feature ?

From: Holger Baer <holger.baer_at_science-computing.de>
Date: Tue, 23 Dec 2003 13:42:19 +0100
Message-ID: <bs9d7c$pvi$1@news.BelWue.DE>

VC wrote:
> Hello Niall,
>
> ["Niall Litchfield" ]
>

>>for l_record in (select sysdate from all_objects) loop
>>
>>null;
>>end loop;
>>dbms_output.put_line( (dbms_utility.get_time - l_start)/100
>>||'seconds....' );
>>end;
>>/

>
>
> It it a joke ? What has measuring 'select sysdate from all_objects'
> performance got to do with my original test ? The original query performs a
> merge join on a real table and the trace(sql_trace) was supposed to show at
> what stage in the join Oracle spends most time.
>
> I am sure that the test would perform beautifully if you removed the query
> entirely.
>
>

Sorry if I missed something in this thread sofar, but what is actually your problem? If you turn on tracing, then your goal is not to measure elapsed wallclock time for the complete query, but rather create a trace file with explain plan, waits etc.

So while an massivly increased response time might bother you, it doesn't render sql_trace useless.

Except, and this hasn't been mentioned in this thread (or i missed it), when the elapsed time in sql*plus is the same as you find in the resulting trace file.

Regards,

Holger Received on Tue Dec 23 2003 - 06:42:19 CST