Re: Experience with setting _kgl_latch_count > 31

From: John Hurley <johnbhurley_at_sbcglobal.net>
Date: Wed, 19 Aug 2009 07:00:46 -0700 (PDT)
Message-ID: <e6902156-9326-48cc-9b6a-b555a0ba177c_at_r18g2000yqd.googlegroups.com>



On Aug 19, 4:35 am, ca111026 <ca111..._at_gmail.com> wrote:

snip

> Would you be able to share real-life experience with setting
> underscore parameter
> _kgl_version_count to value more than 31? This parameter controls
> number of
> child latches for library cache latch.
>
> It is recommend to increase the value of _kgl_latch_count if instance
> experiences issues
> with library cache, for example:
>
> http://www.ixora.com.au/q+a/library.htm
>
> However Metalink Note 420308.1 states that as a result of bug 5508574
> the instance may hang if the value of parameter is more than 31.
>
> We are looking at increasing value of this parameter for database
> running on
> server with 24 CPU. Default value of the parameter is 29 (next prime
> number after CPU count).

With all the postings you are putting out here lately it seems like perhaps you might want to bring in some consulting help. If you have the budget to pay maintenance/support on a 24 cpu machine then you want to proceed slowly and cautiously in thinking about changes.

Mr. Jonathan Lewis is one person that perhaps you might consider bringing in to look at what is going on in your system. The guys at Pythian are getting a good reputation. Perhaps even Tusc or Oracle consulting.

If this is the same system you are running that is having issues with hard parses then I think your shared pool size is like 5 gig? If you reduce that ( what seems to be excessively large ) size you may reduce the impact you are having on library cache.

Your installation seems to be going down the path of always increasing everything ... that's not always ( or perhaps even often ) a good path.

Experimenting with changing oracle underscore parameters is dangerous. Received on Wed Aug 19 2009 - 09:00:46 CDT

Original text of this message