quarkman <quarkman_at_myrealbox.com> wrote in message news:<oprt5c1rq5zkogxn_at_haydn>...
> On 19 Aug 2003 01:33:16 -0700, Ed <ed_zep_at_ntlworld.com> wrote:
>
> > Hi.
> >
> > Does anyone know of a list of Oracle disaster/recovery scenarios with
> > their solutions. What I have in mind is a flow diagram with something
> > like
> >
> > "control file missing" -----> solution
> >
> > Ideally it could fit on a single sheet. This is sounding a naive
> > question, but should the worst happen it would be great to not have to
> > look through manuals.
> >
> > Does anyone know of a document like this?
> >
> > Thanks.
> >
> > Ed.
> >
>
>
> No, and sorry to say, I wouldn't tell anyone about it even if I did. You
> simply can't do recovery with flow diagrams. There's a billion and one
> possible variations, and no flow diagram is ever going to cover all
> eventualities.
>
> Seriously: the only way to master backup and recovery is to understand the
> mechanisms Oracle uses to ensure database synchronisation. If you can
> visualise where your data files are at, what agrees with what, and what the
> application of redo does to things, then it really becomes relatively
> straightforward. I'm not saying that *you* are trying to avoid this sort of
> low-level understanding by producing flow charts, but if your junior were
> to try and follow a generic flow diagram without understanding what is
> going on, it'll be a recipe for total disaster except in the simplest of
> cases.
>
> There might well be legal implications to producing such a document, too.
> If I lost hundreds of thousands of dollars because a flowchart-follower had
> just trashed my database, I'd be gunning for the flowchart producer.
> Probably.
>
> If you really, really want mindless recoveries, then use RMAN. The commands
> don't need searching in the manual (at least not in 9i: restore database;
> recover database; ...works nearly every time!).
>
> Seriously: try and understand what is going on during normal database
> operations, and then it becomes painfully obvious what needs to happen
> during recoveries. The commands aren't that taxing, either: recover
> database, recover datafile, recover tablespace and recover database until
> TIME | SEQUENCE | CANCEL. That's about it, really, unless you are getting
> into the really exotic stuff... and then no amount of flowchart is going to
> help you.
>
> By the way, you'll never need anything very complex to know how to deal
> with a "control file missing" situation, because if you multiplex your
> control files, as you're supposed to do, the solution is simply to copy one
> of the duplicates back into the slot vacated by the missing one. ;-)
>
> Regards
> HJR
Had a bad CPU board once that rebooted the server several times before
it was finally fixed. In each case, it trashed all three copies of
the control file. I have real good notes on how to recover with a
backup control file :).
Received on Thu Aug 21 2003 - 15:18:45 CDT
