[CoLoCo] Best backup solution (AMANDA)

[Kevin Fries]

On Wed, 2008-03-12 at 08:24 -0700, Jeffrey LePage wrote:
> On the subject of backup, Michael Robbert wrote:
> >>I use AMANDA at work, but that is really overkill 
> >>for home use...
> 
> I'm setting up some stuff at a local school, and now
> they want a backup solution.  I immediately thought of
> AMANDA.  
> 
> We have some Windows (2000, XP, Win2k3 server) and
> some Linux.  More Linux is on the way.  We have about
> 30 computers total.  We may peak around 100.  I'm not
> sure how many computers would actually need backup.  
> 
> Can anyone comment on the issues I would be facing if
> I were to set up Amanda?

I had mixed results with Amanda.  I tend to use a program that is in the
repositories called BackupPC.

Instead of going to tape, it stores to a hard drive.  There are serveral
advantages to the BackupPC solution:

First, files stored on BackupPC will not be stored twice if two people
try to store the same file.  What happens is that computer one tries to
store "Corp Memo.pdf" in the backup.  BackupPC does a checksum against
it, and stores it in the general pool of files.  It then makes a hard
link to the file inside the directory holding the backup for machine 1.
Later, machine 2 has the same file for backup.  BackupPC understands
that and will not store the file again, but instead, make a hard link to
the file in the directory holding the backup for machine 2.

Second, your entire backup history is online, not near-line.  Again,
take that memo file that was backed up... The next day, both computer 1
and computer 2 will get a new directory created to hold that days
backup, and new hard links will be created to the file in the pool.
This is four hard links, and only one actual copy of the file...
compression without compression, and backup for day 1 and day 2 are both
stored in an accessible form.

Third, Like Microsoft's shadow copies, all files can be made available
to end users or other assignees to restore their own files.  All files
are online storage, therefore no tapes need to be loaded.  This makes
files available at all times.  BackupPC will use Apache to make
management, and restoration available via a web page.

Fourth, Because of the way files are stored in a central pool, and hard
links are used to show which files belong to which backup set, every day
is presented on the web management page as a full backup.  Therefore,
you do not need to remember which day that file changed.  Example,
weekly full backups are done every friday.  On thursday you realize you
accidentally erased that important memo you received earlier in the
week.  BackupPC, I go to the backup for last night and restore it.
Amanda, I go to the system admin and tell them I lost the file, and need
it back, I got it Monday or Tuesday I think.  The Sys-Admin then needs
to look through the logs for each day to figure out what tape it is on,
then load that tape to restore the file.

Fifth, by using Samba, RSH, and SSH, BackupPC can backup any type of
file from any OS.  All users see the same interface regardless of OS,
making technical support quite easy.

Five bullets should be enough to make my point.  If you want more, just
ask.  I have used both systems, and have found nothing in Amanda that is
not implemented far better in BackupPC.

Oh, and if you are concerned about offsite, combine BackupPC with FAM,
an always on Internet connection, and RSYNC to have the backup
replicated off site automatically.

The final point (OK, I guess this is number six, but it needs to be
said), if you put your backup on a RAID device that has LVM partitions,
you can keep expanding your backup to meet your needs.  Tapes require
tape changers to extend beyond one tape.  With everything online,
expandable, and easily replicable offsite, take a look at BackupPC.
With 30 desktops and growing, you will be really glad you did.

-- 
Kevin Fries
Senior Linux Engineer
Computer and Communications Technology, Inc
A Division of Japan Communications Inc.