Re: TAF and FCF together ?

From: DA Morgan <damorgan_at_psoug.org>
Date: Fri, 15 Sep 2006 21:34:00 -0700
Message-ID: <1158381237.715218@bubbleator.drizzle.com>

joeNOSPAM_at_BEA.com wrote:

>> With a TAF failover the query restarts with the same SCN. Thus the
>> result set is exactly the same.
>> --
>> Daniel Morgan
>> University of Washington
>> Puget Sound Oracle Users Group

>
> Sorry, my newsreader sometimes makes it very difficult
> to bottom post!
>
> My concern is that the SCN does not necessarily indicate the
> same DBMS datum from DBMS to DBMS over time, for instance
> if the table has changed between the failure of a connection and a
> eventual reconnect to the other DBMS.

Of course it does. That is precisely what Oracle's Multiversion Consistency model is all about. This is Oracle Concepts and Architecture 101. Oracle uses the SCN, which equates with a timestamp, and the rollback/undo to reconstruct the data as it was at specific moment in time.

There are members of your team working with RAC here in the Puget Sound region, perhaps you are one of them, if so you might want to contact me off line. At least one member of your team took my class at the U and it seems there are quite a few misunderstandings on the part of BEA about RAC.

> The oracle TAF document
> warns that on occasion, a query will fail, and I was interested
> in expanding the example to actually exercise TAF's requery ability,
> and wondering if these failures are always obvious.

Run the demo downloadable from the PSOUG website. That is precisely what it reports and precisely what we use it to demonstrate.

Look in the code following the comment that begins:

• There are 7 possible failover event

> I think we both agree that TAF is being deprecated, but there are
> some, obviously that are still being lured by the siren-call of it's
> promise.

Hardly a siren call. It is a valuable technology not available from any other RDBMS vendor. The only reason I prefer FCF is its relationship with FAN and the fact that it works as well with SQL*Plus as with OCI, or Java thick or think or, .NET.

> If we expanded the example to more thoroughly test
> TAF, and if it exposed more of the Oracle-documented problems,
> it might serve to sway these tempted souls from the rocks or
> show how to handle them.
>
> Joe

Perhaps I am mistaken but you don't need an expanded example. You just need to understand, far better, the one you have already.

-- 
Daniel Morgan
University of Washington
Puget Sound Oracle Users Group

Received on Fri Sep 15 2006 - 23:34:00 CDT