Re: Digging in to trace files

Mladen Gogala wrote:
> On Thu, 04 Aug 2005 11:32:12 -0700, EdStevens wrote:
>
> > The file referenced in the above messages is the file in which the
> > messages are found. What is this telling me? Is this documented some
> > place that I've missed? Certainly an advanced search of MetaLink and
> > this ng's archives turned up nothing useful.
>
> You haven't searched enough. You are using MTS and your traces belong to
> the server processes. You cannot do that before 10G (see "trcsess"). To
> trace session in a meaningful way it has to be connected through a
> dedicated server. If not, you will get a piece of trace in each background
> process which executed your SQL. I don't know which of Tom Kyte's books
> are you reading but I am sure that he advises people not to use MTS
> without a real need in all of them. You should take a look at the
> "Efficient Oracle by Design". You have a lot to learn, young padawan.
>
> --
> http://www.mgogala.com

Mladen,

The Kyte book I referred to was Expert One-on-One. (I've just begun on Effective Oracle by Design). I know that with shared server the session will get traced into different files. And yes, I would prefer to not use shared server, but I believe this situation pretty much forced it. The db is supporting several apps, and one in particular has driven up to several hundred concurrent sessions and predicted over a thousand. Milsap (Optimizing Oracle Performance, pg 122) describes the process of gathering the specific session info from the several trace files. After reading it over again, he sort of hints that the markers I was asking about are used to point out where these 'file swaps' occur. Also, on pg 107, on a different subject, he has a trace file sample that includes the line '*** TRACE DUMP CONTINUED FROM FILE ***' which is close but not exactly the same as my '*** TRACE DUMP CONTINUED FROM FILE <filename> ***'. And he doesn't address the meaning of that particular line.

When I first saw these in my trace files, my first thought was that it was informing me of the 'trace file swap' (my term -- when the process changes dispatchers and trace files), but then I saw that it said, in effect '*** TRACE DUMP CONTINUED FROM <this very file> ***' Is that because the next request was simply, by luck of the draw, picked up by the same dispatcher as the previous one?

• Ed