Re: LOG_BUFFER setting question

"Jan Gelbrich" <j_gelbrich_at_westfalen-blatt.de> wrote in message news:arv8ts$mcmqe$1_at_ID-152732.news.dfncis.de...
>
> "Paul Brewer" <paul_at_paul.brewers.org.uk> schrieb im Newsbeitrag
> news:3de28833$1_3_at_mk-nntp-1.news.uk.worldonline.com...
> > "Jan Gelbrich" <j_gelbrich_at_westfalen-blatt.de> wrote in message
> > news:arte04$lmbbd$1_at_ID-152732.news.dfncis.de...
> > > Hello, I would like to ask on more question:
> > >
> > > The 8i online CD says:
> > > --------------------------
> > > LOG_BUFFER
> > > Parameter type: Integer
> > > Parameter class: Static
> > > Default value: Operating system specific. Maximum: 500K or 128K *
> > > CPU_COUNT, whichever is greater
> > > Range of values: Operating system dependent
> > > --------------------------
> > > We have 4CPU, so maximum is 524288 ?
> > > Does it mean that the actual setting of 2097152 (2M) is a mistake ?
> > > It was set as we found 'redo log space requests' from a script report,
> > > and the script suggested to increase log_buffer (from 1M already).
> > >
> > > If 128k*CPU count is the maximal value, what can I do to decrease
'redo
> > log
> > > space requests' still occurring ?
> > >
> > > SQL> select substr(name,1,25) Name,
> > > 2 substr(value,1,15) "VALUE (Near 0?)"
> > > 3 from v$sysstat
> > > 4 where name = 'redo log space requests'
> > > 5 /
> > >
> > > NAME VALUE (Near 0?)
> > > ------------------------- ---------------
> > > redo log space requests 2988
> > >
> > > This definitely far beyond 0 and also increasing over the last months
> ...
> > >
> > > My system:
> > > Oracle EE 8.1.7 on AIX 4.3.3 (4 CPU), C/S environment with 200
clients,
> > but
> > > only 10 active (avg) at the same time.
> > > Oracle Forms&Reports Runtime Environment on clients.
> > >
> > > ThanX in advance,
> > > Jan
> > >
> > How often are the logs switching?
>
> In normal times, switches every 5-8 hours, on heavy duty times (once a day
> for a period of 1 hour) every minute sometimes.
> When I first realized this, I thought this is a mess, I should increase
the
> redo log sizes (3 groups of 40M).
> Then I saw the history of all switches over the last 2 years, its figures
> where almost constant: about 50 switches on weekdays
> and about 5 on weekends. Exceptions only at currency conversion to EUROs.
>
> I admit that I am not satisfied with this situation, but unfortunately, I
> cannot do just what I would like to improve the db,
> because I am not the only one in the team ... would You, too, suggest to
> increase log group sizes ? And _then_
> increase log_buffer - even if it would be over its recommended maximum of
> 500K then ?
>

I'd definitely increase the size of your redo logs to something like 1GB each, and make sure to schedule a job which forces a log switch at a rate that you are happy gives you protection from instance crashes (specifically, unduly long instance recoveries).

The log buffer almost never needs to be more than a couple of meg at most. I doubt you need to increase it.

Regards
HJR
> Yours,
> Jan
>
>
Received on Wed Nov 27 2002 - 02:54:32 CST