RE: streams capture performance
Date: Thu, 16 Oct 2008 16:29:54 +0300
Message-ID: <48f7421f.0309d00a.21e8.723c@mx.google.com>
Madhu,
Do you have any spilling under V$BUFFERED_QUEUES for you capture queue?
I've seen a situation where streams_pool_size is NOT dynamically resized even though ASMM is enabled
Hazem
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org]
On Behalf Of GovindanK
Sent: Thursday, October 16, 2008 6:02 AM
To: Madhu Sreeram
Cc: oracle-l_at_freelists.org
Subject: Re: streams capture performance
Madhu
Here are a few more. Is there a separate ARCH process for switching the logs to arch or is the LGWR over burdened? Do you have any table with > 125 columns?
Recommended Patches for Streams 437838.1
Bug 6017440 - Streams capture / CDC / apply slow mining DMLs for tables with
REFRESH ON COMMIT Materialized Views
Doc ID: Note:6017440.8
Bug 5881229 - AV capture is slow Doc ID: Note:5881229.8
Bug 5370578 - Capture process slow on startup on 'control file sequential
read'
Doc ID: Note:5370578.8
Bug 5135907 - Streams capture slow when object definition is not in
dictionary
Doc ID: Note:5135907.8
Streams Capture Process Takes a Long Time To Start With State 'Dictionary Initialization' and Wait Event 'Control File Sequential Read' Doc ID: Note:406673.1
Streams Capture Process Running Extremely Slowly With 'CPU Time' as the Top
Wait Event
Doc ID: Note:458214.1
HTH GovindanK
On Wed, 15 Oct 2008 17:38:54 -0500, "Madhu Sreeram" <madhusreeram_at_gmail.com> said:
Yes the note#335516.1 as well as many other documents in metalink under the
"streams" knowledge browser.
Note#365648.1 seems to be relevant only for apply. I have looked at the
output of Health check specified in #273674.1 but nothing obvious.
I have only one capture process running (there is just one capture queue),
however I set the parallelism to 4 (using: dbms_capture_adm.set_parameter)
but found no improvement with the default (=1) .
The latency has stayed mostly the same and I do see that log miner is
processing archives constantly.
-Madhu S
On Wed, Oct 15, 2008 at 7:23 PM, GovindanK <gkatteri_at_fastmail.fm> wrote:
When you said Best Practice Docs did you mean Metalink Note # 335516.1 ?
Did you look at Note:365648.1 , 273674.1 ; How many capture process are running? Does the difference always stay at 12 hrs or it picks up sometime?
GovindanK
-- http://www.freelists.org/webpage/oracle-lReceived on Thu Oct 16 2008 - 08:29:54 CDT