Re: 100 hard parses per second in 10.2 database

From: Mark D Powell <Mark.Powell_at_eds.com>
Date: Mon, 17 Aug 2009 09:18:26 -0700 (PDT)
Message-ID: <ed49415a-5f37-4d53-8a0f-e585b308423a_at_s31g2000yqs.googlegroups.com>



On Aug 17, 12:10 pm, Mark D Powell <Mark.Pow..._at_eds.com> wrote:
> On Aug 17, 3:31 am, ca111026 <ca111..._at_gmail.com> wrote:
>
> > This is complex application, I cannot test it myself. The tests for
> > 10.2.0.4 patch set were done by testing team, they didn't find any
> > issues, but when it went to production there were multiple problems.
> > We rolled back. More testing, second attempt also ended in rollback.
> > Obviously there is an issue with testing but this is not something I
> > can fix. The amount of time allocacted for testing of cursor_sharing
> > change was rather limited, testing team found some issues (basically
> > some statements using different plans) and this was it.
>
> I agree with Palooka, that trying to change cursor_sharing to FORCE is
> a bad idea.  There have been numerous bugs related to wrong results
> with this setting.  SIMILIAR is generally safer to use, but even using
> it we ran into serious trouble in the past.
>
> The application should be fixed to use bind variables.
>
> HTH -- Mark D Powell --

PS - I am pretty sure all our issues with cursor_sharing = SIMILAR had to do with performance and we probably could have fixed them if we had been given time.

  • Mark --
Received on Mon Aug 17 2009 - 11:18:26 CDT

Original text of this message