RE: latch wait - cache buffer chain - Solved?
- From: "Cary Millsap" <cary.millsap@xxxxxxxxxx>
- To: <genegurevich@xxxxxxxxxxxxxxxxxxxxx>
- Date: Thu, 26 Oct 2006 23:39:16 -0500
Maybe, ...maybe you were encountering some unlucky hash collisions. When
two or more popular blocks hash to the same cache buffers chain (cbc),
you end up with a higher amount of cbc latch competition than you want.
It might be that when you created your new index, the file ids and block
ids of the new index didn't hash to the same cache buffers chains as
your old ones, and you removed an unlucky contention.
Problem is that now you'll never know, unless you still have the old
index lying around. You could have looked in x$bh to see if you had
pairs (or more) of particularly active blocks on the same cbc.
Cary Millsap
Hotsos Enterprises, Ltd.
http://www.hotsos.com
Nullius in verba
Hotsos Symposium 2007 / March 4-8 / Dallas
Visit www.hotsos.com for curriculum and schedule details...
-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of
genegurevich@xxxxxxxxxxxxxxxxxxxxx
Sent: Thursday, October 26, 2006 10:56 AM
To: oracle-l@xxxxxxxxxxxxx
Subject: Re: latch wait - cache buffer chain - Solved?
I seem to have solved my issue by building another index and the same
query
that was not coming back
and running for 30min with a suppressed index is running in 3 to 4
minutes
now. What I wonder now is whether
this was a good solution in the long run or did I just do a bandage
without
addressing the core issue. If anyone
have any thoughts I would appreciate them
In the meantime, thanks to Goran, Frits, Mark , Thomas and Eagle fan for
thier help
thank you
Gene Gurevich
Oracle Engineering
224-405-4079
"Thomas Day"
<tomday2@xxxxxxxx
m>
To
Sent by:
genegurevich@xxxxxxxxxxxxxxxxxxxxx
oracle-l-bounce@f
cc
reelists.org oracle-l@xxxxxxxxxxxxx
Subject
Re: latch wait - cache buffer
chain
10/25/2006 01:39
PM
Please respond to
tomday2@xxxxxxxxx
I just look up "Ask Tom" and he says that there's only one reason to
used a
reversed index -
"why you would: you are using OPS and need to remove a hot spot from an
index on a table every node inserts into. Period. thats the only
reason."
So I guess that it wouldn't be a worthwhile fix to this problem anyway.
--
//www.freelists.org/webpage/oracle-l
--
//www.freelists.org/webpage/oracle-l
Other related posts:
