Re: high log file sync wait

From: Suya <suya.huang_at_gmail.com>
Date: Tue, 4 Dec 2012 17:18:13 +0800
Message-ID: <CAJ-jFMbzUSpLLM=kSOeVndsME=3Dtd2kz3S4tGiBOe7juXA6Kw_at_mail.gmail.com>



I compared the system statistics we got before with the current one: the first one is the metric before, the second is the metric we have now, and the third value is the percentage increased.

redo blocks written
 66,777,507
 45,417,168
 -32%

in our previous test, the application TPM is higher, thus the redo blocks written is higher, this is reasonable. however, the following metrics seems abnormal:

redo blocks checksummed by FG (exclusive)  4,122,706
 8,741,271
 112%

redo synch long waits
 9,759
 106,017
 986%

 redo synch time
 26,950,595
 35,266,236
 31%

 redo synch time (usec)
 269,505,943,171
 352,662,539,336
 31%

 redo write broadcast ack count
 39,923
 96,618
 142%

 redo write broadcast ack time
 254,688,838
 686,018,713
 169%

 redo write broadcast lgwr post count
 1
 1,142
 114100%

anyone has any clue about what does the above metrics mean and where is the bottleneck of a LGWR SYNC wait?

2012/11/26 Suya <suya.huang_at_gmail.com>

> Hi,
>
> I have a four-node RAC system, the workload runs against the database
> waits on log file sync heavily, average wait time is 82.41 ms, account for
> 75% DB time. The system CPU resource is rich enough, average 70% idle time.
>
> similar testing was done previously, which has log file sync wait with
> average 40 ms, and because the CPU resource shortage, this time, we added
> two logical CPU to the system. but see that the log file sync wait event
> was even worse.
>
> any ideas on what else I should check in the AWR report?
>
> Thanks!
> Suya
>

--
http://www.freelists.org/webpage/oracle-l
Received on Tue Dec 04 2012 - 10:18:13 CET

Original text of this message