Re: Automatic checkpoint tuning on RAC
From: joel garry <joel-garry_at_home.com>
Date: Wed, 1 Jul 2009 16:02:43 -0700 (PDT)
Message-ID: <193bdc73-fd95-4bff-ac7b-9b4eb527c046_at_g1g2000yqh.googlegroups.com>
On Jul 1, 2:43 pm, Mladen Gogala <mla..._at_bogus.email.com> wrote:
> On Wed, 01 Jul 2009 21:22:56 +0100, Jonathan Lewis wrote:
>
> > Joel, thanks.for the reference.
> > Interesting suggestion - as you say.
>
> Jonathan, that is the reference I had in mind as well. There are, however,
> other recommendations as well:http://www.freelists.org/post/oracle-l/10gR2-RAC-and-FAST-START-MTTR-
> TARGET,1http://newappsdba.blogspot.com/2007/10/rac-crash-recovery.html
>
> Both of those seem to be more or less successful derivatives of the
> recommendation mentioned by Joel. So far, I don't see any problems with
> the automatic checkpoint tuning on RAC. Googling gave me a note by an
> old friend, Jamagadani Rajendra, who has set it to 300 on 9i RAC.
>
> --http://mgogala.freehos.com
Date: Wed, 1 Jul 2009 16:02:43 -0700 (PDT)
Message-ID: <193bdc73-fd95-4bff-ac7b-9b4eb527c046_at_g1g2000yqh.googlegroups.com>
On Jul 1, 2:43 pm, Mladen Gogala <mla..._at_bogus.email.com> wrote:
> On Wed, 01 Jul 2009 21:22:56 +0100, Jonathan Lewis wrote:
>
> > Joel, thanks.for the reference.
> > Interesting suggestion - as you say.
>
> Jonathan, that is the reference I had in mind as well. There are, however,
> other recommendations as well:http://www.freelists.org/post/oracle-l/10gR2-RAC-and-FAST-START-MTTR-
> TARGET,1http://newappsdba.blogspot.com/2007/10/rac-crash-recovery.html
>
> Both of those seem to be more or less successful derivatives of the
> recommendation mentioned by Joel. So far, I don't see any problems with
> the automatic checkpoint tuning on RAC. Googling gave me a note by an
> old friend, Jamagadani Rajendra, who has set it to 300 on 9i RAC.
>
> --http://mgogala.freehos.com
A bit more poking about finds http://www.oracleutilities.com/wiki/index.php?title=Obsolete_Parameters and looks like it was buggy before 10.2.0.3 per my oracle support.
Again, this is all academic to me, but I'll wildly speculate this has something to do with hacks getting the balance right between deciding a node is dead and a split-brain. So any problems would be load and configuration specific, with a very obscure problem topology. In other words, Murphy's gonna getcha.
jg
-- _at_home.com is bogus. http://ask.slashdot.org/story/09/07/01/190236/Staying-In-Shape-vs-a-Busy-IT-Job-ScheduleReceived on Wed Jul 01 2009 - 18:02:43 CDT