Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Shutdown Immediate Command

Re: Shutdown Immediate Command

From: <jeremiah_at_wolfenet.com>
Date: Fri, 19 Jan 2001 18:11:32 GMT
Message-ID: <94a00a$qi0$1@nnrp1.deja.com>

Actually, the points you are making have nothing to do with user skill level, or even the 'shutdown abort' issue. They have to do with application exception handling.

Any error, no matter if it results from a 'shutdown abort,' a constraint violation, or a user-defined exception, must be handled by the application appropriately. If errors are not properly handled by an application, then the application is BROKEN or BAD.

Many things can cause an operation to fail. No user who clicks "save" then gets a failure message saying "this was not saved" thinks they saved their data. It comes down to application quality. This has nothing to do with how you choose to shut down.

Note that a shutdown immediate could also kill a user's session, resulting in a failure message after clicking "save." Are you therefore also reccommending against 'shutdown immediate?' How about a CGI that fails to connect to a database that was 'shutdown normal' when the user presses 'save?' That would result in the user getting an error message. Should we not use 'shutdown normal' either?

--
Jeremiah

In article <3a6842fc_at_news.iprimus.com.au>,
  "Howard J. Rogers" <howardjr_at_www.com> wrote:

>
> "David Pomphrey" <High.Flight_at_btinternet.com> wrote:
> > The client won't be sent a 'statement processed' message until LGWR
> > indicates that is has finished writing to the online log group.
> >
>
> Presumably you are saying that until a client receives such a message,
we
> shouldn't really worry about data loss.
>
> Personally, I think we should be worried about anything that is lost
which
> the User reasonably *thinks* was committed, and bugger the messages
> displayed on the screen.
>
> > The client may issue a commit, but it will never hear that the
commit
> > was ever acknowledged - hence the client knows that the commit was
not
> > performed.
> >
>
> You obviously work with a set of Users who are 99.99% better trained
than
> any set of Users with whom it has been my misfortune to work.
Congratualte
> yourself if so, but recognise that mere mortals may have a humbler
> expectation: "if I click this button, I have saved my data".
>
> At which point, being told 'oh no you didn't, because some
obstreperous DBA
> decided to be "clever"' is not really comfort enough.
>
> > Easy really.
> >
>
> Yup, terribly easy: the risk remains with an abort of losing data
which
> everyone *thinks* (reasonably) has been committed.
Sent via Deja.com http://www.deja.com/
Received on Fri Jan 19 2001 - 12:11:32 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US