Re: Latch Contention on cache buffers lru chain

From: John Bechtel <john.bechtel_at_teldta.com>
Date: 1997/03/02
Message-ID: <3319B0BE.6A02@teldta.com>#1/1

> > When we upreved a couple of our instances from 7.1.4 to 7.1.6, we
> > immediately noticed an increase in the number of sleeps and latchwaits
> > when Oracle tries to aquire a latch on the cache buffers lru chain.
> > A parameter (spin_count) has helped alleviate some but not all
> > of the enormous performance degradation we now see. Any insight/feedback
> > would be appreciated. Reply to bill_sidloski_at_dg.com

Johnny Chan wrote:
>
> upgrade to 7.3, which includes a new init.ora param to set up multiple
> latches for the lru.
>
> Johnny Chan
> Independent Oracle Specialist

Unfortunately, Johnny is correct in that an upgrade is about the only thing that will solve this. I went around and around with Oracle "support" on this for several months, trying to change all kinds of parameters. Nothing seemed to make any difference. In 7.3, they have completely re-written the internal latching mechanisms, incuding allowing for multiple cache buffer latches to exist.

John Bechtel
TDS Computing Services Received on Sun Mar 02 1997 - 00:00:00 CST