RE: AWR report html parser

From: Teehan, Mark <mark.teehan_at_credit-suisse.com>
Date: Thu, 12 May 2011 12:12:13 +0800
Message-ID: <E3CCEA4C11521E43A343441089E9E555059A9CA3_at_ESNG17P32003B.csfb.cs-group.com>


Somewhat off the point; but I no longer bother diff'ing AWR reports. I think that the AWR data collection is a great idea, but presentation of metrics like this in a report is a terrible idea.

I think that one of the most significant problems with analyzing AWR output is the rolling up of metrics for non-fixed periods: as the database workload shifts throughout the day, stats become hopelessly skewed (1: rman backups; 2: stats gather; 3: dbms_space_advisor; 4: the app itself...) and determining a dependable, repeatable workload profile becomes more trouble than it is worth.

That's why I use a central AWR repository that pulls content from DBA_HIST_.. From lots of databases into a central database where it is summarized, repartitioned, and presented using Apex charts. This differs a lot from Grid Control: long term trending of wait events and sql stats are a capacity management function that doesn't exist in GC, where the focus is very much monitoring.

Determining the ranking of wait events and whether number two is 50% of number one, or 30% of number one, is not an easy task on an AWR report but all becomes clear on a stacked area chart. Also comparing disparate metrics on a chart yields eureka moments now and again (wait event log_file_sync versus system statistic User Commits; or wait event enq: HW - contention versus system statistic Transaction Rollbacks.. for instance).

I love AWR; but I hate AWR reports. My advice is (license permitting) to access the awr views directly and start charting them. BTW once caveat on why AWR is not a great design: flushing of stats for SQL statements from the SQL area before they've been scraped to AWR - this is a problem on databases with daily cycles containing single-execution statements. Apart from that, I love AWR.

Rgds,
Mark
Singapore
www.markteehan.com

-----Original Message-----
From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Greg Rahn Sent: 12 May 2011 08:23
To: kylelf_at_gmail.com
Cc: ORACLE-L
Subject: Re: AWR report html parser

Slightly orthogonal answer:
This is why I have become a big fan of AWR extracts and not just getting reports.
http://download.oracle.com/docs/cd/E11882_01/server.112/e16638/autostat. htm#PFGRF94199

Then you can use the awr diff reports, etc. and you dont have to scrape/parse.

On Tue, May 10, 2011 at 11:55 AM, kyle Hailey <kylelf_at_gmail.com> wrote:
>
> I end up comparing a lot of AWR reports and all the reports tend to be

> HTML which is nice for browsing visually but not so nice for my
> scripts which diff the data. In order to diff AWR reports that are in
> HTML format,I use "HtmlAsText.exe" to convert HTML to txt then run my
> scripts on the text files.
>
> Does anyone know of any other tools for parsing HTML AWR reports?

--
Regards,
Greg Rahn
http://structureddata.org
--
http://www.freelists.org/webpage/oracle-l



=============================================================================== 
Please access the attached hyperlink for an important electronic communications disclaimer: 
http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html 
=============================================================================== 

--
http://www.freelists.org/webpage/oracle-l
Received on Wed May 11 2011 - 23:12:13 CDT

Original text of this message