RE: [QUAR] Re: latch wait - cache buffer chain

From: <genegurevich_at_discoverfinancial.com>
Date: Wed, 25 Oct 2006 09:29:51 -0500
Message-ID: <OF3E7263C6.6FD45FF4-ON86257212.004F93F9-86257212.004FA1BE@discoverfinancial.com>

Mark, it is Oracle 9204. I'll look into the outlines

thank you

Gene Gurevich

                                                                           
             "Bobak, Mark"                                                 
             <Mark.Bobak_at_il.pr                                             
             oquest.com>                                                To 
             Sent by:                  <genegurevich_at_discoverfinancial.com 
             oracle-l-bounce_at_f         >, <oracle-l_at_freelists.org>         
             reelists.org                                               cc 
                                                                           
                                                                   Subject 
             10/24/2006 06:15          RE: [QUAR] Re: latch wait - cache   
             PM                        buffer chain                        
                                                                           
                                                                           
             Please respond to                                             
             Mark.Bobak_at_il.pro                                             
                 quest.com

What version of Oracle? In 10g, there are SQL Profiles. In previous versions, stored outlines...one of them may do the trick.

--

Mark J. Bobak
Senior Oracle Architect
ProQuest Information & Learning

There is nothing so useless as doing efficiently that which shouldn't be done at all. -Peter F. Drucker, 1909-2005

-----Original Message-----

From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org] On Behalf Of genegurevich_at_discoverfinancial.com
Sent: Tuesday, October 24, 2006 6:21 PM
To: oracle-l_at_freelists.org
Subject: [QUAR] Re: latch wait - cache buffer chain Importance: Low

Well, I have executed the same SQL without the index (supressed it by adding a +0 to a where clause) and it completed in 32 minutes. Still nothing to write home about , but better than it was. However the report is created by a reporting tool and I can't manipulate the SQL. Is there a way to rebuild that index or change some parameters based on my findings so that it does not slow down my SQL?

Thanks for any input

thank you

Gene Gurevich
Oracle Engineering
224-405-4079

genegurevich_at_disc

overfinancial.com

Sent by:
To

oracle-l-bounce_at_f oracle-l_at_freelists.org

reelists.org
cc

                                       frits.hoogland_at_gmail.com, "goran

                                       bogdanovic" <goran00_at_gmail.com>

             10/24/2006 05:01
Subject
             PM                        Re: latch wait - cache buffer

chain

Please respond to

genegurevich_at_disc

overfinancial.com

Thank you. Goran and Fritz.

I have now identified an index with a number of "hot" extents. I noticed that all the extents in the output of my SQL have the same child latch #. I am not sure whether this is OK. I am now trying to rerun my SQL without the index that was on the hot list to see if that helps. This index is parallelized and the SQL is executed with 8 readers and 8 writers. Should I change the value of freelist of my index based on that?

thank you

Gene Gurevich

             "goran
             bogdanovic"
             <goran00_at_gmail.co
To
             m>

genegurevich_at_discoverfinancial.com

             10/24/2006 03:33          frits.hoogland_at_gmail.com,
             PM                        oracle-l_at_freelists.org

Subject
                                       Re: latch wait - cache buffer

chain

take the p1raw from the v$session_wait for waiting session and this is your hladdr in the query you are looking for.

On 10/24/06, Frits Hoogland < frits.hoogland_at_gmail.com> wrote: You should find the hottest child latch yourself, and swap the value of
ADDR on line 12 with the value of hladdr of the child latch.

if you execute this query, no rows should come up, because the chance of
having an hladdr = 'ADDR' in x$bh is fairly low.

cheers!

frits

On 10/24/06, genegurevich_at_discoverfinancial.com < genegurevich_at_discoverfinancial.com> wrote: He everybody:

   I am trying to tune a report which has been running for some time and    has
   been timing out since last week.
   When I execute that report I see a large number of "latch free" wait    events
   with p2 parameter equal to 98.
   Based on v$latch that is a cache buffer chain latch. One of the reasons

for this event can be a hot block.
I have found the following query on the metalink to locate a hot block:

     1  select /*+ RULE */
     2    e.owner ||'.'|| e.segment_name  segment_name,
     3    e.extent_id  extent#,
     4    x.dbablk - e.block_id + 1  block#,
     5    x.tch,
     6    l.child#
     7  from
     8    sys.v$latch_children  l,
     9    sys.x$bh  x,

10 sys.dba_extents e
11 where

   12    x.hladdr  = 'ADDR' and
   13    e.file_id = x.file# and
   14    x.hladdr = l.addr and
   15    x.dbablk between e.block_id and e.block_id + e.blocks -1

16* order by x.tch desc

but it did not return anything. Is there anything else I need to look at

related to this latch?

Thanks for any insight

thank you

Gene Gurevich

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

--

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

--

http://www.freelists.org/webpage/oracle-l Received on Wed Oct 25 2006 - 09:29:51 CDT