Re: suspicious standard Oracle Linux start/stop script

Bloody wireless keyboards....

"Howard J. Rogers" <howardjr2000_at_yahoo.com.au> wrote in message news:3eda73d8$0$23749$afc38c87_at_news.optusnet.com.au...
> OK, Joel. You win... I couldn't resist one, very last, post.
>
> Because I couldn't resist pointing out the utter absurdity of your
argument.
>
> > Does it make a difference? If you need media recovery, how can you
> > avoid instance recovery?
>
> So, now we see that you really don't see the difference between instance
> recovery and media recovery. Never mind that the ocde paths are utterly
> different. That one is done automatically by SMON, and the other done
> manually by your Server Process. That one involves reading only the logs
> written to by LGWR, but that the other involves reading logs created ab
> initio by ARCH.
>
> All of which is why posting a list of bug reports which refer to media
> recovery issues has nothing, repeat nothing, to do with instance recovery.
> Unless you are suggesting that the same bit of code which tells SMON how
to
> do stuff also happens to instruct your Server Process how to behave?
>
> >If any of them affect instance recovery,
>
> Big if. Why don't you post specific instance recovery-related bugs,
instead
> of trawling the deep seas of Metalink and coming up with Nada relevant?
>
> > then the simple act of doing a shutdown abort as opposed to shutdown
> > immediate stuffs you. In fact, the one with the plus sign is instance
> > recovery, and is noted mostly by people getting stuffed on their
> > standby databases, which as we all know, are in continuous recovery
> > mode. One could expect many of the standby bugs to show up in
> > recovery.
>
> "One" might expect that, if "one" supposes that the code which governs
> Standby Database behaviour has the slightest thing to do with how instance
> recovery is handled on a non-standby database. But that's a mighty big
> supposition, and is (a mere incidental, I'll admit) totally wrong.
>
> >
> > HJR made it into a red herring by saying I don't know the difference
> > between media and instance recovery.

Well, the evidence is in: you don't. You appear, from your very own post, to think that if anything reads the current redo log and applies it to your datafiles that ipso facto means you've just done an instance recovery. But it just ain't so, Joel: different processes do the various actions. Which means different Oracle code. Which means a bug in one isn't going to affect the other.

>>In fact, he made most every
> > reference I made into red herrings like that, rather than answering
> > the objections.

Well, if you had any meaningful objections, I've yet to hear of them. You post a list of metalink messages which have *zilch* to do with instance recovery. And you accuse *me* of throwing red herrings around!

>>I could have specifically answered each "bollocks"
> > with a metalink reference, but Morgan asked me to stop publicly, so I
> > stopped.
> >
> > There is nothing wrong with scripts doing a shutdown abort followed by
> > a bounce, _as long as there is something in there to notify someone
> > right away if it doesn't work_ and the DBA takes responsibility for
> > any problems.

Hang on. So now there's "nothing wrong with doing a shutdown abort". A moment ago, you were saying, "If any of them [ie, bugs] affect instance recovery, then the simple act of doing a shutdown abort as opposed to shutdown immediate stuffs you."

Which is it? Are there bugs which mean the mere act of doing shutdown abort stuffs you? Or is it ok to do a shutdown abort, so long as you know immediately something goes wrong?

The crux of this argument is here, I think. Either you believe there are bugs which affect instance recovery, or you don't. If you do, then why in God's name are you using Oracle in the first place?? Any database which cannot recover from a simple power failure doesn't deserve to be a repository for important data. Access springs to mind. But if you believe (as I do) that there are no identified instance recovery bugs, then shutdown abort holds no particular terrors, apart from the one about the loss of the current redo log.

>>I still maintain it is extremely bad advice to tell
> > newbies that there is nothing different between shutdown abort and
> > shutdown immediate.

Why? If you believe that shutdown abort is so dodgy as to be unreliable, then switch to DB2 or SQL Server or some other RDBMS that doesn't suffer from this dreadful bug-riddled behaviour.

If, on the other hand, the best you can do is to post a list of bugs which relate to media recovery, but claim that they probably also affect instance recovery, then you don't know your onions from your brassicas, and your argument holds no water.

>>That's just macho posturing to prove you aren't
> > afraid of Oracle's recovery mechanism.

There is more than one Oracle recovery mechanism. That seems to be the point you're missing. And I'll say it again: if you are "afraid" of Oracle's recovery mechanisms, you probably shouldn't be using Oracle to store your data in the first place.

>>Given the O8 track record on
> > recovery, that's like jumping between the tops of two elevators in a
> > shaft.
> >
> > And now we're going to go through a whole development cycle again for
> > 10i.

Yup. Mis-state the actual issue in the first place (instance recovery=media recovery therefore a bug affecting one must affect the other), and then throw in an unreleased version, just to keep the FUD pot stirring.

It's sad, really.

In a sea of ignorance, which is sad enough on its own, *wilful* misrepresentation of simple facts is even sadder.

And that really is it from this particular neck of the woods.

HJR Received on Sun Jun 01 2003 - 16:58:56 CDT