Re: Is using OPTIMAL in RBS ORA-1555 prone?

damorgan wrote:
>
> Come on Howard who has that much free time?
>
> Modify the default profile to automatically kill based on CPU usage. That'll
> teach the little beggars. And you can continue to do real work.
>
> Daniel Morgan
>
> "Howard J. Rogers" wrote:
>
> > Why on Earth would you ever get an 'unable to extend' error message, unless
> > your rollback segments were growing in size? (Bear with me, I'm leading up
> > to something here...!)
> >
> > And if your rollback segments are growing in size, it means that either you
> > haven't made them big enough in the first place or your users are royal
> > pains in the butt who refuse to commit their transactions in a timely
> > fashion.
> >
> > The cure here is not to dabble with Optimal, but to invest in appropriate
> > storage for rollback segments that are sufficient unto the task that is
> > required of them, and/or to beat your Users around the earlobes until they
> > start behaving.
> >
> > The second option is usually more fun (for the DBA). Even funnier -start
> > killing off transactions that have been pending for a long time. That
> > usually teaches them quite fast.
> >
> > Regards
> > HJR
> > --
> > ----------------------------------------------
> > Resources for Oracle: http://www.hjrdba.com
> > ===============================
> >
> > "Syltrem" <syltrem_at_videotron.spammenot.ca> wrote in message
> > news:A4tc8.24$a04.147_at_tor-nn1.netcom.ca...
> > > But then if I don't I'll eventually get an ORA-1650 unable to extend
> > > rollback segment
> > >
> > > I'll have to manually shrink my rollback segments, won't I?
> > > I just can't let them grow forever...
> > >
> > > What is your approach with this?
> > >
> > > Thanks
> > > --
> > >
> > > Syltrem
> > > http://pages.infinit.net/syltrem (OpenVMS related web site - en français)
> > > To reply to myself directly, remove .spammenot from my address
> > >
> > > "damorgan" <dan.morgan_at_ci.seattle.wa.us> a écrit dans le message de news:
> > > 3C714542.22B587A0_at_ci.seattle.wa.us...
> > > > Yes. And you are also more likely to generate needless overhead. It is a
> > > syntax
> > > > I never use.
> > > >
> > > > Daniel Morgan
> > > >
> > > >
> > > >
> > > > Syltrem wrote:
> > > >
> > > > > If using OPTIMAL setting in the rollback segments, am I more likely to
> > > have
> > > > > ORA-1555 snapshot too old?
> > > > >
> > > > > That is, Oracle will shrink the rollback segment at the start af any
> > new
> > > > > transaction, so it may thus drop information required by that other
> > > > > long-running process.
> > > > >
> > > > > Even so, you can never tell when a rbs extent will be reused so how
> > can
> > > you
> > > > > make 100% sure you will not get this error? Users get it every couple
> > of
> > > > > weeks and I can't do an alter session use rollback segment xxxx in
> > this
> > > > > application.
> > > > >
> > > > > Any suggestions appreciated.
> > > > >
> > > > > --
> > > > >
> > > > > Syltrem
> > > > > http://pages.infinit.net/syltrem (OpenVMS related web site - en
> > > français)
> > > > > To reply to myself directly, remove .spammenot from my address
> > > >
> > >
> > >

I've always preferred killing at random rather than based on silly metrics like cpu and/or IO

(Maniacal laugh .... )

-- 
==============================
Connor McDonald

http://www.oracledba.co.uk

"Some days you're the pigeon, some days you're the statue..."