Re: Trying to locate a cursor with very little information

From: Tanel Poder <tanel_at_poderc.com>
Date: Thu, 9 Jul 2009 17:49:37 +0800
Message-ID: <4602f23c0907090249v213dd3a9v5afa2c7e3c032bf3_at_mail.gmail.com>

Thanks for asking the question, Chen. You motivated me to finally write a blog entry about this ;-)

http://blog.tanelpoder.com/2009/07/09/identify-the-sql-statement-causing-those-wait-x-lines-in-a-top-truncated-sql-tracefile/

Basically you can take an errorstack level 3 or processstate level 10 dump of the target process with oradebug and the *cursordump* section contains the currently open cursors.

--
Tanel Poder
http://blog.tanelpoder.com


On Thu, Jul 9, 2009 at 3:10 PM, Cary Millsap <cary.millsap_at_method-r.com>wrote:



> Chen,

>

> It sounds like you've had some good fun. Thank you for letting me know how

> you got your "top-truncated" trace data. It all makes sense now.

>

> And *thank you*: we did redesign the method-r.com home page recently. We

> added the new Follow Us pane on the left, and we "orthogonalized" the banner

> links (Software, Services, Education, Support). The flash quotes have been

> there for a few months, though. The goal, of course, is to help make sure

> people understand that we exist (I meet people all the time who still don't

> realize that I severed my association with Hotsos in early 2008), and that

> we have lots of great ways we can help people make their software go faster.

>

>

> Cary Millsap

> Method R Corporation

> http://method-r.com

> http://carymillsap.blogspot.com

> http://twitter.com/cary_millsap

>

>

> On Wed, Jul 8, 2009 at 9:02 PM, Chen Shapira <cshapi_at_gmail.com> wrote:

>

>> > I agree with Kerry.

>> >

>> > I'm surprised, though, that there's no PARSING IN CURSOR #2 anywhere in

>> the

>> > trace stream preceding the FETCH #2 line. Whether there's a PARSE #2 or

>> not,

>> > I'd expect a PARSING IN CURSOR #2 to precede either the FETCH #2 or an

>> EXEC

>> > #2 in there somewhere. What circumstance do you think caused Oracle not

>> to

>> > emit the PARSING IN CURSOR section?

>>

>> Ah, I expect that Oracle did not emit the parsing section.

>> You would, however, miss the parse command if you never had access to

>> the trace file, and got your information from listening in to the

>> network, starting few minutes too late...

>>

>> Please don't ask how I ended up doing this :)

>> Obviously it is not a normal debugging scenario and I was perfectly

>> justified in refusing to look for the SQL. I was just very curious if

>> it *could* be done...

>>

>> BTW, Cary, Was there a redesign on Method-R homepage? I love the new

>> banner with the quotes :)

>>

>> Thanks,

>> Chen

>>

>

>



-- 
Tanel Poder
http://blog.tanelpoder.com

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

Received on Thu Jul 09 2009 - 04:49:37 CDT