Re: [Dar-support] backing up directly to tape

[Denis C. <dar...@fr...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 17/02/2015 12:02, Gour wrote:
> Hiya,

Hello,
> 
> I'd like to use Dar instead of Amanda for backing up directly to
> LTO-2 tapes.
> 
> By looking at docs I see the following:
> 
> 'If the filename is "-" *and* no slicing is asked for (no -s
> option) the archive is produced on the standard output allowing the
> user to send the resulting archive through a pipe (or into a tape
> device).'
> 
> Now I wonder how is one supposed to e.g do full backup which should
> span 3-4 tapes, iow. should one pipe the whole backup directly to
> the tape (how?)

piping one command's output to a pipe is for the whole life of that
program, this is the shell that binds the stdout to something like a
special device inode or a plain file or an anonymous pipe, for example.

> and dar will somehow handle spanning to several tapes or it's 
> required to define slice size?
> 
> In the latter case I wonder how to tell dar to directly write to
> each tape sequentially and not create all the slices and then cp
> them to the tapes?

Good question! Well this is something that is missing to dar, at first
I thought outputing dar to the split command would have solve the
issue but nop, split does not help sending again data to a
file/special device once end of file was met.

I think about adding a dar independent program for now (let's call it
dar_split) that would copy its stdin to its stdout and suspend when it
reached end of file (on output), having the user able to change the
tape and let the output continue on a new tape. A program that could
also work the other direction, by gathering different tapes in input
also suspending when reaching eof of file (on input) and outputting to
stdout. So piping dar output to dar_split should probably address your
need and reading dar in sequential-read mode from dar_split would let
you access your archive without having to copy from tape to disk first.

This type of program seems feasable and quit simple, see attachment, I
have not tested it much but it seems to work as expected.

In consequence, if dar archive are not larger than one tape you don't
need anything more than dar, else you would require dar_split piped to
the dar output or to dar input (with dar used in --sequential-read),
The good point too is that you don't have to calculate the size of
slices to use:

  dar -c - -R some/where -z | dar_split > /dev/tape

  dar_split < /dev/tape | dar -t - --sequential-read

If that matches your need I will integrate that program (after more
intensive testing) into dar for next major release.

> 
> 
> Sincerely, Gour
> 
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
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=ANiB
-----END PGP SIGNATURE-----