Re: RawDevice??

Narayanan Olagappan <narayan_at_sprynet.com> wrote in article <3403AECF.659E_at_sprynet.com>...

> >
  As long as the
> > specific datafile is not too large to be a "file" in a filesystem (the
 old
> > 2 GB limit on some systems) you can back up the datafile (using dd) off
 to
> > tape, remove the link, and then restore the datafile to the dbs
 directory.
> > The control files will still remember the datafile exists in the dbs
> > directory (where the link use to reside) and you'll be able to re-start
 the
> > instance and open the database with no additional work required.
> >
> >
> >
>
> This is not true. In many of the Unix systems image of a raw device will
> not correspond to that of the standard file. You have to skip the raw
> device header before you make a standard file out of it.
>

I think it is true and I have never heard about "raw device file header". When
you read from or write to raw device you do it through device driver which takes care about physical location of data and all other things needed to successfully complete r/w operation. So what would be the role of "raw device
header" ? By the way, I remember that I have copied database files from raw
devices to file system using "dd" yet. And the database was working perfectly.
Anyway, one can easily check if there is any "header information" in raw device by following sample.

$ head < /dev/rdsk/rawdevice > so_called_header $ cat > /dev/rdsk/rawdevice
put some text here
it will be written to rawdevice and the "header" will be overwritten Ctrl d
$ head < /dev/rdsk/rawdevice > corrupted_header $ diff so_called_header corrupted_header

What do you expect, what will be the result of the latest command ?

Cheers

Mario Simic
IN2 d.o.o.
Zagreb, Croatia Received on Sun Aug 31 1997 - 00:00:00 CDT