Re: inconsistent query performance

Or just set statistics_level = 'ALL' at session level, run your SQL twice in that session (or as many times as you need to get both plans) and then collect SQLd360 with license N to get pretty much the same info as the trace (for the sake of SQL tuning at least) plus statistics, DDL, 10053, etc etc

Sorry Stefan, I had to do it :-)

On Wed, Jan 18, 2017 at 5:30 PM, Stefan Koehler <contact_at_soocs.de> wrote:

> Hey Bob,
> view on/of views is not necessarily a bad thing - it depends (as always)
> on various factors like rownum, aggregations, etc.
>
> However not quite sure if you have licensed diagnostic/tuning pack - if
> not SQLd360 will not provide a lot of historic information. In such cases
> you
> can rely on the good old kernel diagnostics and tracing infrastructure.
> Set the scope on sql_trace[] for your particular SQL ID and let it run until
> you have captured the good and bad runs - all you need :-)
>
> Best Regards
> Stefan Koehler
>
> Independent Oracle performance consultant and researcher
> Homepage: http://www.soocs.de
> Twitter: _at_OracleSK
>
> > "Stauffer, Bob" <RStauffer_at_fult.com> hat am 18. Januar 2017 um 20:49
> geschrieben:
> >
> > Hi Ken,
> >
> > It’s the exact same query and literals each time. And it’s based on
> views – which are based on other views. ;-) So that may be an issue, too.
> >
> > Anyway, I didn’t want to burden everyone with the details of the
> query. I was just hoping to get some avenues to pursue. Like I said in the
> > original post, I’m rusty with Oracle and needed some starting points.
> Thanks to you, Carlos, TJ, and Jeffrey. Your suggestions will get me
> > started.
> >
> > Y’all are a great resource.
> >
> > Bob Stauffer
> > DBA/Database Developer
> --
> http://www.freelists.org/webpage/oracle-l
>
>
>

--
http://www.freelists.org/webpage/oracle-l