Re: "log file parallel write" wait times get increased a lot after upgrading from 10203 to 10204
- From: Eagle Fan <eagle.f@xxxxxxxxx>
- To: Tanel Poder <tanel@xxxxxxxxxx>, oracle-l@xxxxxxxxxxxxx
- Date: Wed, 1 Dec 2010 10:19:52 +0800
Thanks Tanel. It makes sense.
I run "truss -p <lgwr process>" on another 10203 database for a few seconds
and there were 483 "kaio(AIOWRITE" in the log file.
But statspack's 15 minutes report only show 20 waits in it. Obviously oracle
has some problem of recording this wait events in 10203.
On Wed, Dec 1, 2010 at 8:31 AM, Tanel Poder <tanel@xxxxxxxxxx> wrote:
> It's more likely that you hit an instrumentation bug which didn't record
> any log file parallel write wait events (in most cases) in 10.2.0.3 and the
> bug was fixed in 10.2.0.4. I don't know the bug# but I've seen the same
> issue, also for db file parallel writes.
>
> So you're not waiting more, you just know about this waiting now. Having
> only 1193 log file parallel writes per day in a busy system doesn't sound
> correct anyway...
>
> --
> Tanel Poder
> http://tech.e2sn.com
> http://blog.tanelpoder.com
>
>
>
> On Tue, Nov 30, 2010 at 7:07 AM, Eagle Fan <eagle.f@xxxxxxxxx> wrote:
>
>> hi:
>>
>> I noticed that after upgraded from 10203 to 10204, the number of waits of
>> parallel write got increased a lot.
>>
>> We are using lgwr async IO (_lgwr_async_io=true and disk_asynch_io=true).
>> OS is Solaris 10.
>>
>> SNAP_TIME TOTAL_WAITS TIME_WAITED
>>
>> ----------------- ----------- -----------
>>
>> 20100425 00:00:00 1193 20
>>
>> 20100426 00:00:00 1187 19
>>
>> 20100427 00:00:00 83753743 3122627 -- after upgraded from 10203 to
>> 10204
>>
>> 20100428 00:00:00 94783396 3353594
>>
>> 20100429 00:00:00 96792543 3320591
>>
>> 20100430 00:00:00 98197875 3245199
>>
>> 20100501 00:00:00 98144618 3162137
>>
>> I checked several databases, all of them had the same behavior.
>>
>> I tried to find some docs to explain this on metalink, but I didn't find
>> anything.
>>
>> Did anybody also have such change before? What's the reason behind this?
>>
>> And I'm curious what the result is if _lgwr_async_io is false.
>>
>> Thanks in advance.
>>
>> --
>> Eagle Fan (www.dbafan.com)
>>
>
>
--
Eagle Fan (www.dbafan.com)
Other related posts:
