[tlug] Saving Backups in One File versus copying directory structure . . . . . . . . . . . . . . . . . . . (was Re: Remote Backups and Command Line Stuff)

[Jim <jep200404@example.com>]

Stuart Luppescu <s-luppescu@example.com> wrote:

> If you're going to do a backup, 
> I would think you'd want the original files, 
> rather than a tar file on the remote machine. 

It depends. 

There can be good reasons for doing either or both. 

When I do such a backup, I typically _want_ it to be in very few 
files, so I often save the backup in something such as a tarball. 

Typically, I want the tarball in one file, 
and I want a little file with the MD5SUM of the tarball. 

The machine on which the backup is stored, 
might not be able to store the information about the files 
accurately. E.g., if I back up to a Windows machine, 
many of the permissions and ownership information might 
go poof. This issue can arise even with Linux. Consider the 
label/value pairs of stuff that can be might be used by something 
such as SELinux. (Sorry, I forget the nomenclature for the label/value 
stuff.) Not all versions of Linux can handle the label/value 
information about files. Also, Considering who owns what files, there might 
be security issues regarding ownership when the files 
are copied over to even another Linux box. The "jim" user 
on one box might not want "jim" on another box accessing the 
original "jim"'s files. If only root has access to the tarball, 
this is less of an issue. 

A tarball can avoid some of the i18n filename issues that us 
folks have to worry about. If I move many files from a modern 
Linux box, to a Win95 box, to a Mac OSX box and back to a 
modern Linux box, what are those non-ASCII characters going to 
look like? Backing up to a single file (such as a tarball) 
greatly reduces the issue. 

A compressed backup also has the obvious benefits of saving space. 

   Check out the attached tarball

When I back up to CD or DVD, and I have enough room, 
I will often have the big honking tarball and MD5SUM file, 
plus a copy of the easily browsable whole directory structure 
with a bazillion files. So I can peruse the CD as ordinary files 
to easily find the stuff I want, and I have the pristine 
preservation of the meta-info in the tarball. 

Sometimes I backup a whole partition or whole drive as one file, 
to preserve every crumb of metainfo. 

   [root@example.com /]# dd if=/dev/hda | gzip | nc 192.168.0.2 1050

   [root@example.com root]# nc -l -p 1050 >hda.gz

> If you're going to do a backup, 
> I would think you'd want the original files, 
> rather than a tar file on the remote machine. 

It depends.