Re: DB RAID Setup

The first thing I would recommend is to look at v$event_histogram.
This will give you a breakdown of where the times are falling into.
Averages from AWR, etc, can be very misleading in this case.

select wait_time_milli, count(*)
from v$event_histogram
where event='log file sync'
group by wait_time_milli
order by 1

WAIT_TIME_MILLI         COUNT(*)
--------------- ----------------
              1                1
              2                1
              4                1
              8                1
             16                1
             32                1
             64                1
            128                1
            256                1
            512                1

(output for reference only)

You could also do a weighted average (wait_time_milli * count(*)) to
get a "biggest bucket" view.  Sometimes a few large values can be very
problematic.  On very busy systems these will stall and cause an issue
where the system doesn't recover until load drops significantly.

Second, there are a bunch of bugs that relate to log file sync.
7452373
6352003
6193945
5896963
5147386
5087592
Work with Oracle support on these.  Some are RAC specific.
These are all fixed in 11.1.0.7 and 10.2.0.5.  One offs are available I believe.

Good luck!

On Tue, Dec 9, 2008 at 1:17 PM, Martin Brown <martinfbrown@xxxxxxxxxxx> wrote:
> I have a similar question regarding raid 5 and what I perceive to be a
> "potential" problem. I am experiencing a high number of logfile sync waits
> that appear most afternoons as the daily load starts to increase.  We're
> running 10.2.0.3 RAC with a 8 node cluster. We use a 3par storage system for
> the database. My architect has configured both raid 1 and raid 5 diskgroups
> and he duplexes the online redo logs and controlfiles to raid5.
>
> Now, he insists that the logfile sync waits are not related in any way to
> the raid5 configuration and the 3par storage solution is
> state-of-the-art. But the only remedy I can find is to move to faster SSD.
> Is it possible that the high afternoon activity coupled with the raid5
> configuration is the root of my problem?
>
> I've looked at other stuff too. The log switches are 2 - 3 an hour. The log
> buffer is never full. Anyone have any other suggestions as to where to look?

-- 
Regards,
Greg Rahn
http://structureddata.org
--
http://www.freelists.org/webpage/oracle-l


Other related posts: