Re: Re-setting V$tables
- From: Mladen Gogala <mgogala@xxxxxxxxxxx>
- To: Sergey Popov <sspopov@xxxxxxxxx>
- Date: Wed, 11 Oct 2006 00:42:50 -0400
On 10/10/2006 09:43:55 PM, Sergey Popov wrote:
> Here is one that resets all of them and for all users:
>
> shutdown force
I would have never thought of that, especially because "force" is not a
legal option for the "shutdown" command. You probably wanted to imitate humor
by saying "abort"? Maybe "startup force"? May the force be with you.
>
> Creating a new session will almost take care of v$sesstat for you.
Almost? Are you sure? Are you really, really sure? Whaddya mean by "take care
of
v$sesstat"? Concrete shoes? V$SESSION will be sleeping with the fishes? Is there
a case in which I will get a new session and will not get a new statistics?
> It
> depend on triggers and client tool you use for this. They may do some
> after logon activity for you.
They may do some after logon activity for me? Like what? Sending my financial
records
to the IRS? I'm very picky about my database tools, I want them to execute my
commands,
no less, no more. That is why I use SQL*Plus and SQLDeveloper
Here's the deal: I have some long running applications which I want to profile,
without
restarting them. The applications are long running and not restartable without
cleanup.
Very surprising for developers, the application actually uses
DBMS_APPLICATION_INFO. I
need to create a profile and say which module is the most expensive one. I was
investigating
several possibilities:
1) Resetting statistics table whenever module boundaries are crossed and
thus collect the statistics with ease and elegance.
2) Creating a table to which I would store the values at beginning and at the
end, forcing me to write a SQL to calculate them. Unfortunately, I'll have
to do this.
3) Writing a Perl script which would produce the desired results by storing the
intermediate results in hashes. Unfortunately, this is more work then 2),
especially if I use analytic functions like "lead" or "lag".
I was trying to see whether there is a way of getting away with option 1) under
10.2.0.2.
Apparently not.
--
Mladen Gogala
http://www.mladen-gogala.com
--
//www.freelists.org/webpage/oracle-l
Other related posts:
