Re: known bugs in 10.2.0.1 on aix 64bit
From: joel garry <joel-garry_at_home.com>
Date: Tue, 14 Sep 2010 12:08:20 -0700 (PDT)
Message-ID: <dd22056e-0883-43b3-bd64-62d8e46b8fdc_at_l32g2000prn.googlegroups.com>
On Sep 14, 11:26 am, "Gerard H. Pille" <g..._at_skynet.be> wrote:
> Gerard H. Pille wrote:
> > Can someone tell me about a bug in 10.2.0.1, which would cause excessive
> > undo with heavy dml on ASSM tablespaces? I found mention of this bug in
> > a post dating back to 2006, but no number.
>
> > I think this bug would be a good reason for upgrading, an application
> > used by my employer does excessive dml, and until today, I didn't
> > realize a more recent version could provide better performance.
>
> > Any other reasons for leaving 10.2.0.1 behind are also very welcome.
> > Should we go for 10.2.0.4 or 10.2.0.5 immediately?
>
> > Thanks,
>
> > Gerard
>
> ah, on supporthtml.oracle.com, I found 3610777 and 4288876, but one thing bothers me, about the
> fix for 4288876:
>
> Note: The fix for this bug introduces a potentially serious
> problem as described in bug 4597251. It is advisable
> for any customer with a one off patch for this bug
> (4288876) to replace it by the fix for bug 4597251 .
>
> where there is talk of block corruption. I prefer lesser performance to block corruption.
> Would I run this risk with one of the more recent versions?
Date: Tue, 14 Sep 2010 12:08:20 -0700 (PDT)
Message-ID: <dd22056e-0883-43b3-bd64-62d8e46b8fdc_at_l32g2000prn.googlegroups.com>
On Sep 14, 11:26 am, "Gerard H. Pille" <g..._at_skynet.be> wrote:
> Gerard H. Pille wrote:
> > Can someone tell me about a bug in 10.2.0.1, which would cause excessive
> > undo with heavy dml on ASSM tablespaces? I found mention of this bug in
> > a post dating back to 2006, but no number.
>
> > I think this bug would be a good reason for upgrading, an application
> > used by my employer does excessive dml, and until today, I didn't
> > realize a more recent version could provide better performance.
>
> > Any other reasons for leaving 10.2.0.1 behind are also very welcome.
> > Should we go for 10.2.0.4 or 10.2.0.5 immediately?
>
> > Thanks,
>
> > Gerard
>
> ah, on supporthtml.oracle.com, I found 3610777 and 4288876, but one thing bothers me, about the
> fix for 4288876:
>
> Note: The fix for this bug introduces a potentially serious
> problem as described in bug 4597251. It is advisable
> for any customer with a one off patch for this bug
> (4288876) to replace it by the fix for bug 4597251 .
>
> where there is talk of block corruption. I prefer lesser performance to block corruption.
> Would I run this risk with one of the more recent versions?
You'd run less risk on the recent versions than on the unpatched version IMO, but there are still ASSM bugs. If you google assm jonathan lewis you'll perhaps find posts about 10.2.0.4. Talk to support about what you've found and see what they say. My guess is these performance bugs mostly show up with high concurrency situations, or the specific cases people like Charles Hooper have posted about. I ran into one performance situation, made a note to myself to investigate a one-off patch, then forgot about it until you just reminded me. Corruption does tend to be more noticeable.
jg
-- _at_home.com is bogus. I once got a round tu it. http://www.flickr.com/photos/joel_garry/799193214/Received on Tue Sep 14 2010 - 14:08:20 CDT