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: Async I/O

Re: Async I/O

From: Alex Hudghton <alex.hudghton_at_capgemini.co.uk>
Date: Wed, 24 Mar 1999 09:02:57 +0000
Message-ID: <36F8AA41.AE4612AA@capgemini.co.uk>


Raw devices are required on DEC UNIX for Parallel Server because the devices cannot be shared by more than 1 node if they are 'normal' filesystems (until the next major release of the OS - 5 I think)

Regards

Alex

Stephane Faroult wrote:

> Hans-Peter Sloot wrote:
> >
> > Hello Stephane,
> >
> > I don't think that your statement is true.
> > Oracle will be clever enough to open the redo log files
> > with the O_SYNC flag.
>
> I would say it depends on operating systems. I distinctly remember
> (well, a long time ago : t'was v5 then) that on some ports the
> documentation was explicitly stating that the only way to guarantee
> integrity was to use raw devices. I have not studied it in detail, but I
> am ready to bet that it is a similar reason why raw devices are required
> on DEC/Unix for intance with the Parallel Server Option. And even if
> redo log files are synchronously written (right, the example was perhaps
> not the most appropriate), and control files too, there is some need of
> synchronization between data, redo and data files, since the data file
> header contains the system commit number which enables Oracle to require
> a recovery if needed on startup. Be certain that if the Oracle support
> discourages something, it means there is a potential risk, however small
> and unlikely.
>
> --
> Regards,
>
> Stéphane Faroult
> Oriole Corporation
> ------------------------------------------------------------------
> http://www.oriolecorp.com, designed by Oracle DBAs for Oracle DBAs
> ------------------------------------------------------------------
Received on Wed Mar 24 1999 - 03:02:57 CST

Original text of this message

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