[svlug] Network based backup/restore.

Michael C. Robinson plug_1 at robinson-west.com
Tue Oct 6 14:28:21 PDT 2015


I am facing three problem areas trying to NFS root boot Linux from
Scratch:

1) It takes time to compile all the utilities and programs a Linux
system needs from scratch and longer if you decide to add X and Gnome.

2) Wasted time on the systemd version and got essentially nowhere when
it came time to NFS root it.  Mistakes are easy to make compiling all
the software and installing it which leads to more lost time.

3) A valid criticism of NFS root, it isn't secure.  The network I'm
using it on is local area though and this problem isn't as serious.
This problem seems fixable, do NFS over IPSEC?

One way to potentially save time and avoid compiling literally dozens
of programs is to NFS root a binary Linux distribution instead.   But
which one will I use and do I have to go pre systemd?  I know there is
RIP Linux which is Slackware based for example.  There is Fedora, but
I'm stuck potentially using an old non systemd version.  Debian might
be a good choice, Debian current stable evidently supports non systemd
configuration.  Or perhaps CentOS 6.x is a good choice in the short
term, because 6.x is getting very old.

Another problem, while external hard drives and burnable DVDs or using
a memory stick is popular, this isn't scalable and typically not
conducive to using a network.  Hard drives can be dropped, memory
sticks and DVD-Rs can be hard to organize and later find.  Using
removable media for backups, you can easily shelve it for too long and
it goes bad.  There is also the issue of securing removable media and
having someone on site to feed it through.  

I am trying to learn how to do backup/restoration work on a large
network, say I get a job where Linux needs to be installed unattended
to hundreds of computers in a beowolf cluster or something similar.
More advanced, imagine installing to a heterogeneous network where 
some computers are Linux based and others use non Linux systems 
such as Freedos, Windows, etcetera.  Installing the OS attended isn't
an option in say a computer lab setting.  Too expensive to hire a
hundred people or more.  You don't want to answer anaconda prompts a
hundred or more times to install Linux on a hundred or more computers
either.

I've looked at Mondo Backup, there is continuous work to get it working
on new releases of popular Linux distributions and Mondo for all the
great work put into it has some major drawbacks.  If for some reason
someone wants to one-up: Fedora, Slackware, Debian, CentOS etcetera by
releasing a new binary Linux distribution that cherry picks the best
from all of these, there is the problem of porting Mondo Backup to it.

Mondo Backup and even network booting Linux for Backup/Restoration has
the drawback of not being able to backup data in real time on a live
system.  Think programming on a Linux system and CVS is on the NAS.
Think personal files on individual computers need to be backed up in
real time.  Is Bacula the answer for data backups?  I'm concerned that
Bacula is proprietary or requires special licensing :-(  I have to
learn it too.

Desires for doing backup/restoration work:

1) Windows NT, even the latest version, backs up to the NAS and can be
restored push button from the NAS with all of a person's programs and
on drive data.  The software keys are easily retrievable if a person
decides to do a fresh install and reinstall their programs.  I am using
FreeNAS9.  PXE boot should be sufficient to restore Windows.

2) No need for removable media.  Laptop friendly, no need for a second
or larger hard drive.

3) Able to back up a MacOS X based system to the NAS and restore it
from the NAS.  Not just MacOSX, insert custom OS'es and other OS'es.

4) Real time back up of CVS repositories and personal programs with, as
-needed, license keys.  Back of of VirtualBOX machines where these
backups are optionally restored in the recovery process.  Note that
a CVS repository should be on network attached storage.

5) Enough information about old backups to decide if you want them back
and diffs if appropriate to save space on the NAS over storing multiple
full backups where the backups are only minimally different.  There
should be protection against deleting the backup that all the diffs
depend on.

In short, I'm looking for a bare metal recovery solution and a real
time backup solution and there are very different use cases.  Since
I am learning on a small personal network, KISS is important.  Keep
it super simple.



More information about the svlug mailing list