Re: Database backup

Στις Κυρ 19 Ιαν 2003 20:23, ο/η Nigel J. Andrews έγραψε:
> On Sun, 19 Jan 2003, Alexandros Perdikomatis wrote:
> > Óôéò Êõñ 19 Éáí 2003 04:59, ï/ç D'Arcy J.M. Cain Ýãñáøå:
> > > Redirecting to help list as it seems more appropriate.
> > >
> > > On Saturday 18 January 2003 09:17, Alexandros Perdikomatis wrote:
> > > > Question: Am I too stupid to use tar?
> > >
> > > I'm sure not but you may misunderstand what tar does. Tar is for
> > > storing a collection of files into an archive. It expects to find the
> > > list of files on the command line in one form or another. If you don't
> > > write to a file first then all you have is a stream of data, not a
> > > file.
> > >
> > > > Some time I had no space on the filesystem to copy all the data in
> > > > command mode (pg_dump) so I thought to pipe pg_dump to tar using a
> > > > scsi tape archiver as output (/dev/rmt0).
> > > > Why doesn't work?
> > >
> > > What exactly was the problem? You may just have a device selection
> > > issue. What system is this on?
> >
> > Thanks for the reply.
> > I think I understand what you mean, I just thought that instead of a file
> > input, we could have a pipe input.
> > It runs on a SuSE linux 6.4 (you may ask why this old version, I updated
> > what was necesary, I like it becouse I have full greek support) and the
> > archiver is an old conner DDS-2 scsi tape streamer which is known on the
> > operating system as /dev/st0 by the kernel whith a symbolic link to
> > /dev/rmt0. When I pg_dump on a file, of cource there's no problem. I
> > restore also correctly.
> > But I would like to pipe this procces. For example: if we type the
> > command: # tar cvf /dev/rmt0 *
> > we're going to have all the files in the tree below to the tape.
> > How can I pipe the output of the pg_dump to it?
>
> Well, you could do:
>
> pg_dump ... > /dev/rmt0
>
> or if you want to set block size:
>
> pg_dump ... | dd of=/dev/rmt0 obs=1k
>
> Basically, you're trying to make it too complicated, tar will archive one
> file but would you really use something like:
>
> tar -cf archived-file.tar unarchived-file
>
> unless the unarchived-file name was generated automatically and it just so
> happened that only one file existed at that time?
>
> If you really do want something tar try cpio instead.

Thanks for the information, I should think about piping to dd... thanks again.
The idea of cpio looks more interesting becouse we could have a multivolume
archive I suppose.

one more time thanks,

Alexandros