| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Mailing Lists -> Oracle-L -> RE: no-routine log file sync
Hi Diego,
DBWn no longer waits in 'log file sync' waits for LGWR to sync the redo for recent
changes to write batch blocks, as it
used to under Oracle7. Instead it sends an asynchronous request to LGWR to sync the
required RBA and then moves the
affected blocks to something called the deferred queue. Blocks on the deferred queue
are included at the head of the
next write batch, by which time the LGWR sync should be well and truly over.
Group commits do not reduce the number of 'log file sync' waits; they only reduce the
number of 'log file parallel
writes'.
There are plenty of reasons why the numbers of 'user commits' and 'log file sync'
waits will not be exactly the same,
however, the correlation should be strong and if not it might be worth investigating
why.
@ Regards, @ Steve Adams @ http://www.ixora.com.au/ @ http://www.christianity.net.au/
-----Original Message-----
Sent: Friday, 19 October 2001 4:41
To: Multiple recipients of list ORACLE-L
Hi Steve,
According to Pablo's question:
"If I recall correctly a log file sync wait is the time
it takes to write log buffer blocks to disk when a
user issues a commit. And this event "only" increments
when a user issues a commit, am I correct ?"
What I think is that a 'log file sync' is not only producced by a
'user commit'.
I think that the difference between these 2 values might be because
'user commits' does no take in consideration 'piggyback' writes due to
multiple commits (so it counts 7 commits and only one log file sync, for
instance) and syncs needed by the DBWR (who will also wait in log file
sync) and are not counted in 'user commits' statistic.
Am I forgetting some other cause? or I'm missing something else?
TIA
Diego
> Hi Pablo and list,
>
> The idea is that commits are routine, and you should expect approximately
1 'log file sync' wait per commit, and the
> duration of that wait should be approximately the same as the average 'log
file parallel write' time. My thinking was
> that if there are more 'log file syncs' or if they take longer, then that
indicates a "non-routine" performance problem.
> However, this is no longer as reliable as it used to be under Oracle7.
Under 8i (and possibly 8.0) if LGWR is using
> asynchronous I/O, then the 'log file parallel write' time may not include
all of the I/O time.
>
> @ Regards,
> @ Steve Adams
> @ http://www.ixora.com.au/
> @ http://www.christianity.net.au/
>
>
> -----Original Message-----
> Sent: Thursday, 18 October 2001 2:16
> To: Multiple recipients of list ORACLE-L
>
>
> Hi list,
> I don't quite understand this Steve Adams script
> (in fact, it's just a part of the script
> resource_waits.sql):
>
> select
> 'non-routine log file syncs',
> e.total_waits,s.value,
> round(e.average_wait * greatest(e.total_waits -
> s.value, 0)) val,
> e.average_wait
> from
> sys.v_$system_event e,
> sys.v_$sysstat s
> where
> e.event = 'log file sync' and
> s.name = 'user commits'
> order by 2 desc;
>
> Why does it substract user commits to total log file
> sync waits?
>
> If I recall correctly a log file sync wait is the time
> it takes to write log buffer blocks to disk when a
> user issues a commit. And this event "only" increments
> when a user issues a commit, am I correct ?
>
> Can anybody explain this problem?
>
> thanks, Pablo
>
-- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: Steve Adams INET: steve.adams_at_ixora.com.au Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051 San Diego, California -- Public Internet access / Mailing Lists -------------------------------------------------------------------- To REMOVE yourself from this mailing list, send an E-Mail message to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in the message BODY, include a line containing: UNSUB ORACLE-L (or the name of mailing list you want to be removed from). You may also send the HELP command for other information (like subscribing).Received on Thu Oct 18 2001 - 14:39:28 CDT
 |
 |