Re: RAC global cache

All the normal RAC rules apply to a database where cluster_database is set to true. I am not aware of any single instance RAC optimisations (nor can I think of any reason), which means the normal steps are followed. Probably the callout on the interconnect is not made, because the RAC cluster knows there is only one instance (every instance has to join the cluster explicitly). I haven't tested this explicitly, though.

This should mean some gc waits can be seen, but because there is only one instance, waits like global cache cr request should be minimal.

Because the 'resource mastering' shifts from two nodes to one node upon bringing down one node/instance, this can result in some waits during the remastering. (this phenomenon can be detected in the alert log, if I'm not mistaken)

frits

On Thu, Jun 5, 2008 at 10:38 PM, Oliver Jost <Oliver.Jost_at_ca.fujitsu.com> wrote:

> Hi All,
> When I bring down one instance in a 2- node RAC environment (CRS still
> up, AMS still up, Solaris 9, Oracle 10.2.0.3) I still see wait events
> relative to global cache activity? I'm wondering if anyone might clear this
> up for me. When a block isn't found in the local cache is LMS still
> busy checking across the interconnect? Do we still get global cache cr
> request and global cache cur request waits while the living instance
> determines whether the request will be fulfilled?
>
>
> Thank you,
> Oliver
>

--
http://www.freelists.org/webpage/oracle-l