Re: Documentation for HHWM and LHWM?

  • From: "Charles Schultz" <sacrophyte@xxxxxxxxx>
  • To: "goran bogdanovic" <goran00@xxxxxxxxx>
  • Date: Tue, 6 May 2008 11:37:37 -0500

Right, I understand that, and read those explanations in various
incarnations from other authors. But what has Oracle documented? The
sections on ASSM only reference a single HWM.

On Tue, May 6, 2008 at 10:29 AM, goran bogdanovic <goran00@xxxxxxxxx> wrote:

> hi,
>
> It has to do with how oracle process new allocated blocks. In case when
> hwm is moved forward, oracle does not immediately format all of the 'new'
> blocks allocated...oracle make it on their first use. This means, when full
> reading a segment, oracle somehow need to know if this blocks are formatted
> and eligible to be read or not...means nothing interesting in them.
> For that purpose there is low hwm and hwm (or hhwm). For all the blocks
> which are between low hwm and hwm, when full scanning a segment, oracle have
> to check assm bitmap block to see which blocks can be read and which can be
> skipped.
>
> hth,
> goran
>
>
>
>
> On Tue, May 6, 2008 at 4:43 PM, Charles Schultz <sacrophyte@xxxxxxxxx>
> wrote:
>
> > Broswing through various blogs, whitepapers and other presentations, I
> > have seen folks like Tom Kyte and Lutz Hartman reference the HHWM and LHWM
> > of ASSM tables, however I have been unable to find where they are officially
> > documented. Perhaps this is a limitation of tahiti and metalink (metalink
> > gave me one hit for bug 6416822). I have looked through the Concepts,
> > New Features and DBA books in the documentation set, learning a lot in the
> > process, and still have a lot to go back over. *grin* Additionally, I
> > appreciate that dbms_space gives us access to detailed information about
> > used blocks, but I find it a little odd that the HWM is not specifically
> > included, much less HHWM and LHWM. Forgive me if I missed it, but did Oracle
> > ever document these ASSM high water marks?
> >
> > Thanks to all of you who have worked hard to expose various internal
> > tidbits and publish helpful hints.
> >
> > --
> > Charles Schultz
>
>
>


-- 
Charles Schultz

Other related posts: