Re: Redo Copy Latch Contention

The problem is typically not to do with the lack of redo copy latches but because of an inefficient LGWR process. I suggest you take a look at Steve Adams site, and under tips look at Redo Latch Tuning at :- http://www.ixora.com.au/  

Basically LGWR takes out locks on all the redo copy latches before it can write to the redo logs.

HTH NeilC

Michel Gauthier wrote:
>
> Eric,
> I'm a little bit happy to hear someone observed the same behavior about that
> latch. I posted something about that last week (see my note here after
> yours), but nobody answered. I'd be glad to know which release and O/S you
> have bad "redo copy" latch contention. Mines are 7.3.2 on a HP3000, MPE/iX.
> There are very few installation of Oracle on those mini-computer and I first
> thought my bad "redo copy" latch behavior was caused by the specific
> implementation of Oracle on MPE.
>
> Michel Gauthier
> Donohue Inc.
>
> <miaemp_at_my-deja.com> wrote in message news:8383d8$tkf$1_at_nnrp1.deja.com...
> > The last several days, I have been testing the effects of the
> > LOC_SIMULTANEOUS_COPIES init.ora parameter to reduce Redo Copy Latch
> > contention we are experiencing. Originally it was set to 12. I
> > increased it to 24, the contention appeared to get slightly worse. I
> > decreased it to 8, the contention appeared to decrease back to what it
> > was when it was set at 12. Different performance tuning books have
> > stated that the above init.ora parameter should be at least set to the
> > number of processors on your system. Others have said that it should
> > be up to 2 times the number of processors of your system.
> >
> > We are experiencing 94% misses on the Redo Copy Latches. These misses
> > are primarily during nightly batch. Anyone have any ideas?
> >
> > TIA
> > Eric Peterson
> > Programmer/Analyst DBA
> > eric_petersonNO_at_mauricesSPAMMERS.inrg.com
> >
> > Please note that any views I have expressed are purely my own and not
> > that of my employer. To email me remove teh NO and the SPAMMERS.
> >
> >
> > Sent via Deja.com http://www.deja.com/
> > Before you buy.
>
> Michel Gauthier wrote (one week before...)
> >I use to monitor latches to see if there are some contention on some of
> >them. I can say that I'm proud of my database settings because, latches
> wait
> >ratios (misses divided by the gets) are almost all 0%, and when not,
> they're
> >less than 0.1 %... except for only ONE.
> >The redo copy latch wait ratio is always greater than 80% !!! And I simply
> >don't know why. I set up the log_simultaneous_copies to 6 (that is twice
> the
> >number of CPUs of our computer) as recommended.
> >Does someone have a suggestion?
> >
> >Michel Gauthier
> >Donohue Inc.
Received on Wed Dec 15 1999 - 10:02:10 CST