Re: Commit -> write to data file immediately???

Niall Litchfield wrote:
>
> That will teach me to post after a long day (and maybe never to post on
> checkpoints again).
>
> Last try to write something sensible.
>
> Checkpoint = flush dirty buffers to disk. I.E. DBWR writing.
>
> LOG_CHECKPOINT_TIMEOUT - Does exactly as Bruno describes for the purpose
> Bruno describes (read I was wrong sorry)
> LOG_CHECKPOINT_INTERVAL - Similar to the above, but specifies the interval
> in operating system blocks not seconds.
>
> I believe that the documentation says that DBWR only wakes up every 3
> seconds - However there was some posting earlier on saying that physical
> writes were happening every 3 seconds - verified by system monitoring tools.
>
> AS an aside it is probably worth taking more or less everything I say at the
> moment with a grain of salt. I am currently in the middle of learning
> --
> Niall Litchfield
> Oracle DBA
> Audit Commission UK
> "Bruno Jargot" <bjargot_at_club-internet.fr> wrote in message
> news:lc77ctk9vhctr1d7lpapc3ba70cs876eps_at_4ax.com...
> > On Thu, 29 Mar 2001 17:13:05 +0100, Niall Litchfield wrote:
> >
> > >"Norman Dunbar" <ndunbar_at_lynxfinancialsystems.co.uk> wrote in message
> > >news:F43E6BAE5BB5D411A44C00805FBE740D6DF36B_at_apps.lynx-fsc.co.uk...
> > >> >> Anyway, the most important misunderstanding above is the 3 second
> > >> >> timeout. DBWR will wake up every 3 seconds regardless of anything
> > >> >> else, but won't necessarily write anything at that point.
> > >>
> > >> Interesting. My course notes clearly state that there will be a
> > >> checkpoint 'n' seconds after the last one (but only if
> > >> LOG_CHECKPOINT_TIMEOUT is set to 'n' seconds.
> > >
> > >This sounds like the misuderstanding I was inadvertently responsible for
> > >posting to this group a while back.
> > >
> > >DBWR does write every 3 seconds. This is not a checkpoint (in the sense
 of
> > >writing the redo log buffer to disk). The LOG_CHECKPOINT_TIMEOUT controls
> > >how often LOG writer writes. (unless of course I've got it all terminally
> > >confused again).
> >
> > No.
> >
> > The checkpoint is when the buffer cache is flush in the datafiles and
> > not when the logwriter is writing in the redolog.
> >
> > And the LOG_CHECKPOINT_TIMEOUT controls how often there will be a
> > checkpoint. The goal of this parameter is to avoid, when the redolog
> > is big, having too much data in the redolog that are not written in
> > the datafiles and then reduce the time to recover if the database
> > crash.

and to make things worse, these two parameters changed at 8.1.5 (I think )now so that (and excuse the bad explanation) it only asks dbwr to spit enough stuff that it would now satisfy the parameter in question

hth
connor

-- 
===========================================
Connor McDonald
http://www.oracledba.co.uk (mirrored at
http://www.oradba.freeserve.co.uk)

"Some days you're the pigeon, some days you're the statue"