Re: V$active session history

From: Dragutin Jastrebic <orahawk_at_gmail.com>
Date: Thu, 26 Mar 2015 21:12:13 +0100
Message-ID: <CANGCQwmMYxCBKnGtQv_eRyhyFhRdJYO28qWOSnbYcARjg8xsNg_at_mail.gmail.com>

=>
2015-03-26 8:08 GMT+01:00 Tanel Poder <tanel_at_tanelpoder.com>:

> Yep, sampling, especially infrequent sampling has its limitations. But the
> problem is not really in V$SESSION itself (where ASH also gets its info
> from internally), but how people *use* it. ASH regularly samples the
> session state objects and even the infrequent short-running queries get
> eventually caught in a sample, so when aggregating ASH data over minutes
> and hours, the query will show up.
>
> V$SESSION tends to (often incorrectly) get used by just glancing at it
> once and assuming that whatever showed up there at that particular point in
> time must be what this problem session is stuck with.
>

<=
>

That was my point as well, sometimes DBAs do not use v$session correctly (for historical reasons probably, because before 10g most people had v$session_wait as mostly used and now they have all information joined in v$session )

A good experience for every DBA is to work at least once on some "near real time application" such as telecom front-office applications or stock exchange applications (and not only on classical ERP /inventory management applications)

In those environnements you learn to think in terms of these short but frequently executed queries that easily bring the server up to 60-70% CPU (even more), while a quick look at v$session shows you only idle sessions.

BTW, it's a pleasure for me to talk to Tanel Poder :-)

Thanks Tanel for all the contribution you are giving to Oracle DBAs community.

Dragutin

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

Received on Thu Mar 26 2015 - 21:12:13 CET