[Techtalk] backup - tar

[Travis Casey]

On Wednesday 01 September 2004 19:15, Maria McKinley wrote:

> Thanks everyone for all of your helpful advice. My short-term solution 
> is using tar, but not using compression and dumping error messages to 
> try to figure out what is going on.
> 
> I did find a useful web site
> 
> http://www.backupcentral.com/free-backup-software2.html
> 
> that lists all of the free software available for unix.

Well... that lists several pieces of free backup software, but nothing like 
*all* of it.  You might want to check out www.linux-backup.net as well.  
Also, O'Reilly's _Unix Backup and Recovery_ book has a lot of good advice.

> I was looking at amanda, which looks like about the best around, but I
> was confused by a statement they made:
>
>   AMANDA supports using more than one tape in a single run, but does not
> yet split a dump image across tapes.
>
> Can someone explain the difference between using more than one tape in a
> single run and splitting a dump image across tapes?  

Think of it this way.  If I remember how Amanda works right, a "dump image" 
is the backup image of a specific filesystem *or* set of directories and/or 
files that you've told Amanda to back up as a "unit".

Amanda cannot split a "unit" across tapes -- but it can back up more than 
one "unit" in a single run.

This has several implications.  Most importantly, if you have a filesystem 
that's larger than can fit on a single tape, you'll have to break it down 
into smaller units by specifying directories and/or files to back up.

Also of note, this can cause a waste of space on tape.  To take a worst-case 
sort of scenario, if you're backing up six 60 GB filesystems, and you have 
tapes that hold 100 GB, you're going to be wasting 40 GB on each tape.

> I would have 
> assumed these were the same things.  My situation is that all of the
> data I want backed up does not fit on one tape.  Right now I backup
> different parts of the system on different nights by cron (now it is
> spread out even more than before, since I have stopped using zip to
> compress!).  How much versatility do I have with Amanda?  Can I save
> half of /home in one night?

I *believe* Amanda will let you do this... but you'll have to specify the 
breakdown.  E.g., if you have users ann, fred, george, and stacy, you could 
tell it to back up /home/ann and /home/fred as one unit, and /home/george 
and /home/stacy as another, and schedule those differently.  If I'm 
remembering wrong, someone else please correct me!

The biggest problem here is that it's not automated... so if you add user 
jill, you'll have to manually add /home/jill to one or the other, or create 
a third "unit" for her.  And doing things manually is always subject to 
error -- for example, when I first was typing the above, I accidentally put 

"back up /home/ann and /home/fred as one unit, and /home/fred 
and /home/george as another"

Obviously, if I made that sort of mistake in actually setting things up, 
Fred would be happy that he has *two* backups of his files every night, but 
Stacy's not going to be happy with me at all when she loses some files!

> While I am asking for backup advice... We are about to switch to a 1/2
> TB raid array.  Right now we have an hp dlt 40/80GB backup tape system,
> but we are going to have to upgrade this.  Any advice on how to backup a
>   1/2 TB?

Arg.  I'm facing the same kind of problem myself, compounded by the fact 
that it's three 1/2 TB arrays, each of which is part of a database -- so 
they all need to be backed up in sync to have a valid backup!

There are tape drives now that will store 200 GB per tape.  There may be 
bigger ones than that, but I haven't heard of any that'll do 500 GB.  

Pretty much the only way to do it with tapes right now is to break it across 
multiple tapes -- which'll require either doing it manually, as above, or 
getting software that can do it for you.

The alternative is to back it up to other disks -- another, much larger RAID 
array.  Some vendors are now selling products meant for exactly this sort 
of thing, or you can do it yourself, by simply building a much bigger RAID 
(I'd recommend at least two times as big, more likely three) and using 
rdiff-backup, dirvish, or one of the scripts for using rsync for backups.

--
       |\      _,,,---,,_     Travis S. Casey  <efindel at earthlink.net>
 ZZzz  /,`.-'`'    -.  ;-;;,_   No one agrees with me.  Not even me.
      |,4-  ) )-,_..;\ (  `'-' 
     '---''(_/--'  `-'\_)