Re: write complete waits

From: John Thomas <jt2354_at_gmail.com>
Date: Mon, 28 Sep 2015 19:47:54 +0000
Message-ID: <CAOHpfbGbDy5O3UD=NmMpJnA5wvGV8GA++C+3pnm+DAmdPc29NQ_at_mail.gmail.com>

Seems a reasonable explanation, concurs with the 10g documentation as well. So probably nothing to do with AIO, though that's also suggested elsewhere.

http://docwiki.embarcadero.com/DBOptimizer/en/Write_complete_wait

If you have scope to increase your buffer cache by increasing sga_target or memory_target, maybe that will help. Other advice has increasing number of DBWR processes, though that's possibly a pre-11g, pre-AIO thing .

Who is it who said test, test, and test again?

Appreciate you're probably talking about a production system so fiddling about and testing may not be easy, but you probably have no alternative if you have a real problem. If the problem is just that the number/duration of this wait type has gone up and you don't have a response time problem, don't fix it.

Regards

JT

On Mon, 28 Sep 2015 at 20:03 Matt Anderson <matta576_at_gmail.com> wrote:

> According to Doc ID 1665736.1, this event can't be tuned or avoided. This
> document is specific to writing to securefiles. Are you using securefiles?
>
> Matt
>
> On Mon, Sep 28, 2015 at 12:46 PM John Thomas <jt2354_at_gmail.com> wrote:
>
>> My understanding its that oracle processes can get on with other things
>> but still need to hear from the AIO driver that the write competed
>> successfully.
>>
>> If that's wrong, I'd like to hear a correct description. Not in the
>> office and on mobile so I'm just half speculating in case it helps .
>>
>> Regards
>>
>> JT
>>
>> On Mon, 28 Sep 2015 19:41 Matt Anderson <matta576_at_gmail.com> wrote:
>>
>>> Why would there be a wait event for an asynchronous write to complete?
>>> Isn't not waiting for completion the whole point behind asynchronous writes?
>>>
>>> Matt
>>>
>>> On Mon, Sep 28, 2015 at 12:00 PM John Thomas <jt2354_at_gmail.com> wrote:
>>>
>>>> Just the filesystem io one which defines asynchronous, direct io or etc
>>>> as supported by whatever platform you are on.
>>>>
>>>> AIO driver status might be another thing to check.
>>>>
>>>> Regards
>>>>
>>>> JT
>>>>
>>>> On Mon, 28 Sep 2015 18:05 Michael Calisi <oracle455_at_gmail.com> wrote:
>>>>
>>>>> No parameter changes recently? Any particular Oracle parameter you
>>>>> are thinking about?
>>>>>
>>>>
>>>>> On Mon, Sep 28, 2015 at 12:25 PM, John Thomas <jt2354_at_gmail.com>
>>>>> wrote:
>>>>>
>>>>>> Is that not a wait for asynchronous write completion? Haven't changed
>>>>>> parameters related to filesystem io or similar recently, have you?
>>>>>>
>>>>>> Regards
>>>>>>
>>>>>> JT
>>>>>>
>>>>>> On Mon, 28 Sep 2015 16:39 Michael Calisi <oracle455_at_gmail.com> wrote:
>>>>>>
>>>>>>> A new event suddenly start to appear in my Top 10 Events.
>>>>>>> Wondering what may be the cause and how to avoid.it Any
>>>>>>> information will be helpful.
>>>>>>>
>>>>>>> Thanks,
>>>>>>>
>>>>>>> Top 10 Foreground Events by Total Wait Time
>>>>>>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>>>>>> Tota Wait % DB
>>>>>>> Event Waits Time Avg(ms) time Wait
>>>>>>> Class
>>>>>>> ------------------------------ ------------ ---- ------- ------
>>>>>>> ----------
>>>>>>> DB CPU 14.5 18.1
>>>>>>> write complete waits 115 13.8 120204 17.3
>>>>>>> Configurat
>>>>>>>
>>>>>>
>>>>>

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

Received on Mon Sep 28 2015 - 21:47:54 CEST