Re: ORCLALRT.LOG doesn't contain timestamps!

I treat the term 'flushed' to mean 'moved permanently away'. This is the use of the word when talking about e.g. Unix file system cache flushing (i.e. putting everything on disk) when unmounting filesystems.

Now you say the buffer cache is flushed at a checkpoint (i.e. also at a red log switch).

What really happens is that all modifications made to data blocks are copied to the datafiles (by copying to disk the affected blocks).

But the data blocks (modified, yes) still stay in memory (or are you trying to tell us that Oracle scrubs blocks from memory as soon as they are modified and written out to disk at a checkpoint!!

I would be shocked if that was true!

So what we have here is a misunderstanding of terms.

You use 'flush' to mean 'changes-to-data-blocks-are-written-to-disk'

I use 'flush' to mean 'moved-permanently-from-one-place-to-another'

The moral of this thread is (which we can all learn from) is :-

communicate clearly and simply with the minimum use of imprecise abbreviations and colloquialisms.

David P.

Oracle Certified DBA.

trea wrote:
>
> I do have a copy of the Generic Online Docs on CD and read through them last
> nite. While the shared pool does not get flushed each time the redo logs
> switch the buffer cache in fact does get flushed. Buffer cache being 50% of
> the SGA, at least in importance. So half right. I placed a call to Oracle
> support and questioned them about the wording in the Tuning Handbook and
> they agreed that it was a very bad choice of words in the book. Thanks for
> bringing this to my attention. By the way, you were incorrect in saying
> that the buffer cache did not get flushed....so you too were half right.
>
> Trea
>
> DNP <High.Flight_at_btinternet.com> wrote in message
> news:38F0F711.30ED_at_btinternet.com...
> > I would seriously advise getting a hold of the 'Generic Oracle
> > Documentation CD-Rom.
> >
> > The issues here are more complicated than the book you're quoting
> > implies. Even although it comes from Oracle Press, it is not a proper
> > manual (unlike the contents of the Cd-Rom above).
> >
> > If you were to see the paper copies of the Oracle Manuals from the
> > Cd-Rom they would propably fill a small shelf at your bookshop!
> >
> > Perhaps Oracle Press feel its better to give people some advice, albeit
> > while simplifying other things. I can understand this up to a point, but
> > I don't think you can rely on their books for real technical detail.
> >
> >
> > David P.
> >
> > Oracle Certified DBA.
> >
> > ====================================================================
> >
> >
> >
> >
> > trea wrote:
> > >
> > > DNP <High.Flight_at_btinternet.com> wrote in message
> > > news:38EFF150.5B1B_at_btinternet.com...
> > > > trea wrote:
> > > > >
> > > > >
> > > > > everytime your redo logs switch your SGA is flushed. this clears
 out
 your
> > > > > db block buffers, shared pool, etc.
> > > >
> > > >
> > > > What a load of rubbish - modified buffers from the buffer cache are
> > > > written (copied, really) to the datafiles on disk as part of a
> > > > checkpoint which occurs at an impending log switch.
> > > >
> > > > One the checkpoint is complete, the contents of the red log buffer are
> > > > then written out to the redo log files.
> > > >
> > > > The shared pool is not touched! No buffers are removed from the buffer
> > > > cache. No data dictionary information in lost.
> > > >
> > > > Why can't some people read the manuals before the give advice???
> > > >
> > > >
> > > > David P.
> > > >
> > > > Oracle Certified DBA.
> > > >
> > >
> > > David;
> > >
> > > Begging your pardon but here is the quote from the manual where that
> > > statement came from:
> > >
> > > Book: Tuning Oracle
> > > Publisher: Oracle Press
> > > Page: 134
> > >
> > > "The problem with redo logs is that when a log switch occurs, the SGA is
> > > flushed...by flushing the SGA, you loose the benifits of keeping hot
 spots
> > > in memory. It is a very expensive process to retrieve information off
 of
> > > disk, compared to fetching information out of memory."
> > >
> > > That is just a portion of the topic in the manual. It goes on to say
 that
> > > you should make your redo logs a lot bigger and size them so they switch
> > > only durring lunch and off-hours etc. Also to set the
> > > LOG_CHECKPOINT_INTERVAL to a size greater than the redo logs file size
 and
> > > the set the LOG_CHECKPOINT_TIMEOUT parameter to 0 so that only when the
 redo
> > > log files fill up and cause log switch will you have a checkpoint.
> > >
> > > If I am missing something here or this information is incorrect please
 let
> > > me know. I would greatly appreciate it.
> > >
> > > Thanks;
> > >
> > > Trea
Received on Mon Apr 10 2000 - 00:00:00 CDT