Re: log_buffer sizing

  • From: Tanel Poder <tanel@xxxxxxxxxxxxxx>
  • To: kp0773@xxxxxxxxx
  • Date: Mon, 15 Aug 2011 21:04:52 +0100

You can't tune log buffer on 11g (and some 10g versions) anymore as the
buffer size is calculated automatically and is related to your SGA granule
size.

The undocumented way for increasing the log buffer is to increase the SGA
granule size by increasing the _ksmg_granule_size parameter, but you
probably don't need to do this.

I would look into this only if I saw the "log buffer space" wait event
taking significant part of the response time and even then, before
increasing the buffer, I would drill down deeper into the root cause of the
problem - why can't LGWR cope with the redo volume. Snapper on LGWR session
is a good starting point. The main reasons are:

1) Slow IO - writes into redolog file can't be done fast enough to cope with
the incoming redo volume
2) LGWR not getting enough CPU time for doing its work

--
Tanel Poder
New seminars:
http://blog.tanelpoder.com/seminar/

On Mon, Aug 15, 2011 at 8:54 PM, K R <kp0773@xxxxxxxxx> wrote:

> sorry for not mentioning earlier  it is  on 11.2.0.2  running on solaris
>
>
> On Mon, Aug 15, 2011 at 12:37 PM, Venkat Krish <venkat.lear@xxxxxxxxx>wrote:
>
>> what version? Hopefully you are employing  batch commits.
>>
>> from 11.2 
>> docs<http://download.oracle.com/docs/cd/E11882_01/server.112/e16638/memory.htm#PFGRF94347>
>>
>> *A reasonable first estimate for such systems is to the default value,
>> which is:*
>> * *
>>
>> *MAX(0.5M, (128K * number of cpus))*
>>
>> * *
>>
>> *On most systems, sizing the log buffer larger than 1M does not provide
>> any performance benefit. Increasing the log buffer size does not have any
>> negative implications on performance or recoverability. It merely uses extra
>> memory.*
>>
>>
>> You definitely should look for more convincing evidence (redo related wait
>> events) before you go ahead and start tuning log_buffer size
>>
>> Regards,
>> Venkat
>>
>>
>> On Mon, Aug 15, 2011 at 3:22 PM, K R <kp0773@xxxxxxxxx> wrote:
>>
>>> All ,
>>>
>>> my online redo logs are 3G . This is a batch intensive datawarehouse
>>> application.  What should i be keeping my log_buffer .
>>>
>>> currently it is set at 29425664  and it just seems too low .
>>>
>>> thanks
>>> Kart
>>>
>>
>>
>

Other related posts: