Re: commit or not commit inside a cursor loop

Sybrand

I think we're arguing about something we agree on but what we refer to as the active transaction is different.

An ACTIVE transaction (i.e. one that has not committed yet) will hold onto it's rollback segement - we both agree. However an active session (A) could be using rollback information from another transaction (B) (that is making changes to blocks that A will need to read in a read consistent manner). B commits whilst A is still in progress. A can then suffer from ora-1555 as the rollback information for B is now suseptable to being reclaimed by another transaction and overwritten. Do we both agree on this. In the example we're talking about the cursor is effectively (A) and the committed updates are transaction (B).

Andy

"Sybrand Bakker" <gooiditweg_at_nospam.demon.nl> wrote in message news:ta5edv82jb9pers85q72je091imjenrgc7_at_4ax.com...
> On Fri, 30 May 2003 09:21:41 +0100, "Andy" <andy.spaven_at_eps-hq.co.uk>
> wrote:
>
> >The most important point I was
> >trying to get across was is that the comment "Data maintained for
> >read-consistency purposes in a transaction will NEVER been overwritten"
was
> >actually incorrect because a rollback segment being so used could be from
> >another transaction (and if that transaction has committed then the risk
of
> >ora-1555 raises it's ugly head).
>
>
> You don't seem to grasp a rollback segment can handle multiple
> transaction.
> Data in an *ACTIVE* transaction will *NEVER* get overwritten.
> Anyone saying something else just talks pure plain nonsense.
> The risk of ora-1555 starts when the active transaction has committed
> the cursor of the original program still needs data for read
> consistency purposes.
> I don't see why you can maintain the position my assertion is
> incorrect. It isn't. It is you who doesn't understand how rollback
> segments work.
>
>
> Sybrand Bakker, Senior Oracle DBA
>
> To reply remove -verwijderdit from my e-mail address
Received on Fri May 30 2003 - 04:36:33 CDT