 | | | issue with wait on log buffer space. Oracle recommends enable write cache?? | issue with wait on log buffer space. Oracle recommends enable write cache?? 2006-01-31 - By Anand Rao
Hi,
what was the size of your log_buffer and what is the current size? you are
only inserting about 32000 rows, so it shouldn't be a big deal. how frequent
is this data load job?
'log buffer space' waits are generally caused due to a small log buffer
relative to the amount of redo you are generating.
there is no direct reason why db_writer_processes has to be
increased.....for 'log buffer space' waits. If you have configured Async IO
on the OS side, then you should use disk_async_io = TRUE. Are you using RAW
volumes for your redologs and datafiles? else, it doesn't make sense to turn
on async IO.
I can't remember too many problems with Solaris 8 Async driver, but you are
better of checking it with your sysadmin as well searching Metalink for any
async IO related patches.
If you are using UFS or VxFS for redo and datafiles, then what you need to
do is reduce your Filesystem Buffer Cache. Are you using Direct IO or some
new version of this 'near RAW performance Filesystem booster package' ? you
need to consider all these filesytem config options before deciding to use
Async IO.
you need to see if your Disk IO thoroughput is able to handle the workload
generated. you may have very few and slow disks causing LGWR to be slow
writing data from the log buffer to the redolog files.
use a 1MB log buffer and see what you get. it all depends on your redo rate.
regards
anand
On 31/01/06, Jaco Polet <jaco.polet@(protected)> wrote:
>
> Hello,
>
> I have a general issue with loading data into a Oracle 9.2.0.4 database on
> solaris 8. I reduced it down to a simple statement which takes 1 second on
> my PC (oracle 9.2.0.6) but takes 20 seconds on the 9.2.0.4 database on
> solaris.
> I traced it and the trace showed a 19 second wait on log buffer space.
> (see trace output)
> I increased the value of log_buffer and set disk_async_io=FALSE
> (increasing the db_writers) but this didn't make a difference
> On metalink I found note 263652.1 saying that this was a general issue on
> all platforms which could be fixed by switching on write cache. (It doesn't
> state that is is fixed in a next version)
> It sounds strange to me that Oracle recommends to write the redo log to
> cache. Is this correct and if so how do I switch this on (the redologs are
> on local disks) ? I am also surprised that when searching the internet I
> don't get any hits on this although it looks like a general issue. What am I
> missing?
>
> Hope someone can help me with this... Jaco Polet
>
> The formatted trace:
>
> insert into t_jpo select * from test_tabel
>
> call count cpu elapsed disk query
> current rows
> -- ---- -- --- -- ----- -- ---- -- -- ---- -- -- ---- -- -- ---- --
> -- ---- --
> Parse 1 0.01 0.00 0 0
> 0 0
> Execute 1 0.90 20.08 0 2920 7664
> 32114
> Fetch 0 0.00 0.00 0 0
> 0 0
> -- ---- -- --- -- ----- -- ---- -- -- ---- -- -- ---- -- -- ---- --
> -- ---- --
> total 2 0.91 20.09 0 2920 7664
> 32114
>
> Misses in library cache during parse: 1
> Optimizer goal: CHOOSE
> Parsing user id: 1340 (PERF)
>
> Rows Row Source Operation
> -- ---- -- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---
> 32114 TABLE ACCESS FULL TEST_TABEL (cr=1182 r=0 w=0 time=96619 us)
>
> error during execute of EXPLAIN PLAN statement
> ORA-00942 (See ORA-00942.ora-code.com): table or view does not exist
>
> parse error offset: 109
>
> Elapsed times include waiting on following events:
> Event waited on Times Max. Wait Total
> Waited
> -- ---- ---- ---- ---- ---- ---- ---- -- Waited -- ---- --
> -- ---- ----
> log buffer space 43 1.00
> 19.18
> SQL*Net message to client 1 0.00
> 0.00
> SQL*Net message from client 1 2.49
> 2.49
> *****************************************************************************
***
>
>
Hi,<br><br>what was the size of your log_buffer and what is the current size?
you are only inserting about 32000 rows, so it shouldn't be a big deal. how
frequent is this data load job?<br><br>'log buffer space' waits are generally
caused due to a small log buffer relative to the amount of redo you are
generating.
<br><br>there is no direct reason why db_writer_processes has to be increased..
...for 'log buffer space' waits. If you have configured Async IO on the OS side,
then you should use disk_async_io = TRUE. Are you using RAW volumes for your
redologs and datafiles? else, it doesn't make sense to turn on async IO.
<br><br>I can't remember too many problems with Solaris 8 Async driver, but you
are better of checking it with your sysadmin as well searching Metalink
for any async IO related patches.<br><br>If you are using UFS or VxFS for redo
and datafiles, then what you need to do is reduce your Filesystem Buffer Cache.
Are you using Direct IO or some new version of this 'near RAW performance
Filesystem booster package' ? you need to consider all these filesytem config
options before deciding to use Async IO.
<br><br>you need to see if your Disk IO thoroughput is able to handle the
workload generated. you may have very few and slow disks causing LGWR to be
slow writing data from the log buffer to the redolog files.<br><br>use a 1MB
log buffer and see what you get. it all depends on your redo rate.
<br><br>regards<br>anand<br><br><div><span class="gmail_quote">On 31/01/06, <b
class="gmail_sendername">Jaco Polet</b> <<a href="mailto:jaco.polet@(protected)
.com" target="_blank" onclick="return top.js.OpenExtLink(window,event,this)">
jaco.polet@(protected)</a>> wrote:</span><blockquote class="gmail_quote" style
="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding
-left: 1ex;">
<font size="2"> </font>
<div>Hello,</div>
<div> </div>
<div>I have a general issue with loading data into a Oracle <a href="http://9.2
.0.4/" target="_blank" onclick="return top.js.OpenExtLink(window,event,this)">9
.2.0.4</a> database on solaris 8. I reduced it down to a simple statement which
takes 1 second on my PC (oracle
<a href="http://9.2.0.6/" target="_blank" onclick="return top.js.OpenExtLink
(window,event,this)">9.2.0.6</a>) but takes 20 seconds on the <a href="http://9
.2.0.4/" target="_blank" onclick="return top.js.OpenExtLink(window,event,this)">
9.2.0.4</a> database on solaris.</div>
<div>I traced it and the trace showed a 19 second wait on log buffer space.
(see trace output)</div>
<div>I increased the value of log_buffer and set disk_async_io=FALSE
(increasing the db_writers) but this didn't make a difference</div>
<div>On metalink I found note 263652.1 saying that this was a general issue on
all platforms which could be fixed by switching on write cache. (It doesn't
state that is is fixed in a next version)</div>
<div>It sounds strange to me that Oracle recommends to write the redo log to
cache. Is this correct and if so how do I switch this on (the
redologs are on local disks) ? I am also surprised that when searching the
internet I don't get any hits on this although it looks like a general issue.
What am I missing?
</div>
<div> </div>
<div>Hope someone can help me with this... Jaco Polet</div>
<div> </div>
<div>The formatted trace:</div>
<div> </div>
<div>insert into t_jpo select * from test_tabel</div>
<p>call count cpu
elapsed disk
query current
rows<br>-- ---- -- --- -- ----- -- ---- -- -- ---- -- -- ----
--- -- ---- -- -- ---- --<br>Parse
1 0.01
0.00 0
0
0 0
<br>Execute 1 0.90
20.08
0 2920
7664 32114<br>Fetch
0 0.00
0.00
0 0
0
0<br>-- ---- -- --- -- ----- -- ---- -- -- ------
- -- ---- -- -- ---- -- -- ---- --
<span></span> <br>total 2
0.91 20.09
0 2920
7664
32114</p>
<p>Misses in library cache during parse: 1<br>Optimizer goal: CHOOSE<br>Parsing
user id: 1340 (PERF)</p>
<p>Rows Row Source Operation<br>-- ---- -- ------
-- ---- ---- ---- ---- ---- ---- ---- ----<br> 32114 TABLE ACCESS
FULL TEST_TABEL (cr=1182 r=0 w=0 time=96619 us)</p>
<p>error during execute of EXPLAIN PLAN statement<br>ORA-00942 (See ORA-00942.ora-code.com): table or view
does not exist</p>
<p>parse error offset: 109</p>
<p>Elapsed times include waiting on following events:<br> Event waited on
Times Max. Wait Total Waited<br> -- ---- --
-- ---- ---- ---- ---- ---- -- Waited -- ---- -- -- ----
-- --<br> log buffer space
43  
;
1.00 19.18<br> SQL*Net
message to client
1
0.00
0.00<br> SQL*Net message from client
1 2.49
2.49<br>***********************
*********************************************************
<br></p>
</blockquote></div><br>
|
|
|