RE: Looking for pointers in troubleshooting -> RAC performance degradation (single instance is fine)

  • From: "Pradeep Chetal" <Pradeep.Chetal@xxxxxxxxxxxxxx>
  • To: "Thomas Roach" <troach@xxxxxxxxx>
  • Date: Thu, 24 Dec 2009 09:29:32 -0500


Thomas,



There are lot of updates (not inserts) but on some of the tables that
are populated by sequences and others without sequences.



I will go through this and see if this helps.



Thanks,



-- Pradeep



________________________________

From: Thomas Roach [mailto:troach@xxxxxxxxx]
Sent: Thursday, December 24, 2009 9:08 AM
To: Pradeep Chetal
Cc: oracle-l@xxxxxxxxxxxxx
Subject: Re: Looking for pointers in troubleshooting -> RAC performance
degradation (single instance is fine)



Sorry to early.. meant that it might "NOT" be the right solution for you
:)

On Thu, Dec 24, 2009 at 9:07 AM, Thomas Roach <troach@xxxxxxxxx> wrote:

Hi,

Are you doing a lot of inserts/updates to tables that have indexes that
are populated by sequences? You should find the objects the contention
is on. To assist with this, take a look at this page.
http://www.ardentperf.com/2007/09/12/gc-buffer-busy-waits-in-rac-finding
-hot-blocks/

Basically what "could" be happening (what happened with me) was that I
had a bunch of sessions across nodes all fighting to put entries into
the same index blocks. Even on one node, this would be a contention
issue. What I did to fix my issue was identify the indexes with the
contention and made them reverse key indexes. With reverse key indexes
you lose the ability to perform certain operations like an index range
scan. So if you do queries where something is between this value and
that value etc, it might be the right solution for you.

Good Luck.






On Thu, Dec 24, 2009 at 8:28 AM, Pradeep Chetal
<Pradeep.Chetal@xxxxxxxxxxxxxx> wrote:



Folks,



We have an app that is running well in single instance but when put on a
2-node or 3-node RAC, we see pretty much similar performance. However,
if the CPU time is 55-60% in single instance, it becomes 20% in 3-node
RAC and top-5 itmed events become..



"gc buffer busy", "gc current block busy", "enq: TX- index contention



Any help in what to look for will be greatly appreciated.



Thanks,



-- Pradeep

Pradeep Chetal
Sr. Director - Infrastructure Architecture

________________________________

Mformation Technologies Inc.
Switchboard: +1 732 692 6200



Visit

WWW.MFORMATION.COM and see how the MFORMATION SERVICE MANAGER can help
you to transform your business

Please consider the environment before printing

This message and the information contained herein is proprietary and
confidential and subject to the Mformation Confidentiality  policy
statement,
you may request a copy by emailing info@xxxxxxxxxxxxxx





--
Thomas Roach
813-404-6066
troach@xxxxxxxxx




--
Thomas Roach
813-404-6066
troach@xxxxxxxxx





Visit WWW.MFORMATION.COM and see how the MFORMATION SERVICE MANAGER can help 
you to transform your business

This message and the information contained herein is proprietary and 
confidential and subject to the Mformation Confidentiality  policy statement,
you may request a copy by emailing info@xxxxxxxxxxxxxx

image/com

Other related posts: