On Wed, 2006-11-15 at 17:35 +0300, Andrew Edunov wrote:
> > This would verify that you don't have hugepages setup. In certainly
> > load conditions, this can have a significant impact on kernel overhead.
> >
> > The output of "cat /proc/sys/kernel/shm*" would also be interesting.
> > Are you using the SuSE oracle script to properly tune the environment
> > for running Oracle?
> >
> > Also, other suggestions about async IO are also worth pursuing.
>
> sunset:~ # cat /proc/sys/kernel/shmall
> 2097152
> sunset:~ # cat /proc/sys/kernel/shmmax
> 3294967296
> sunset:~ # cat /proc/sys/kernel/shmmni
> 4096

OK, those look fine, so I'm still going to shoot for hugepages being the
most likely problem. Here a link to a posting I had on the Redhat
taroon list that discusses a very similar problem a user was
experiencing. In his case configuring hugetlb was the thing which fixed
it.

http://marc.theaimsgroup.com/?l=taroon-list&m=112293119130924&w=2

The only thing that's a little odd about your postings is that it
doesn't appear you are having a high context switch rate, but I still
think it's work setting up hugepages. It's generally very simple to do.

Later,
Tom



--
To unsubscribe, email: suse-oracle-unsubscribe@(protected)
For additional commands, email: suse-oracle-help@(protected)
Please see http://www.suse.com/oracle/ before posting