[SATLUG] backup recommendations

[Al Castanoli]

You have a lot more faith in disk drives than I do.  I do back up to
disk, but after restoring an entire datacenter from tape after Hurricane
Andrew over ten years ago, I still back up to tape.  For examble: I'm
currently backing up around 50GB per night on just one of my servers to
tape, and intend to continue doing so, even though the machine I'm
backing up has a failover server that updates every five minutes.  I've
had occasion to recover data off those tapes four or five times a month
to show what was on the server at an earlier date.  I'm using a tape
changer and don't have to worry about overwriting older data when the
robot has removed a tape from the tape drive.  To get the same
functionality from disks, I'd have to invest in several more terabytes
of space for disk drives that rarely approach their MTBF ratings these
days.

Long story short - backing up to remote disk drives is fast, fairly
reliable, and much easier than backing up to tape if you're writing your
own backup scripts rather than using some pointy clicky gui thingy for
backups, but that doesn't mean backing up to disk is the right solution
in all cases.

For the Linux servers I have running near the Pacific coast, I use mtx,
mt, dump, and restore from the command line to do manual tape backups,
and script those commands into my cron driven backups.  Except for the
terminal I ssh into those servers with, the servers are firewalled off
from contacting other machines, and the customer will not allow
exceptions. so offsite rdist syncing or some other hard drive backup
routine is not possible.

Al Castanoli

On Fri, 2007-03-02 at 17:05 -0600, Daniel J. Givens wrote:
> Tapes are unreliable and unnecessarily expensive. An off-site VPN, disk to disk
> backup would be a better solution. Backup the logs, document root, configs and
> databases (if applicable) regularly. Better yet, store your configs in an SVN
> repo and keep that backed up. Only do full system backups occasionally, and I
> would say if you are using stock packages, then just backup your package list
> and leave the distro provided stuff out of the backups all together.
> 
> What I do for my clients who are limited in funds and storage space is backup
> the critical areas regularly. The standard stuff that comes from the distro is
> going to be able to be recovered with a quick re-install on new hardware. So in
> case of a critical failure, you pop in your install disk to the new server,
> install, update and then restore your configs, websites from SVN and databases.
> 
> The other option is to use LVM with /var and /usr/local as separate logical
> volumes. Back those up as snapshots, then once you have your freshly installed
> system, you copy snapshots back to the newly created LVs on the new system.
> 
> If you have more money to play with and two or more datacenters, go more
> enterprise. Setup a two backup (file) servers in a failover cluster hooked to a
> SAN. Have this setup at both datacenters with a VPN between the two. Do said
> backups as I described before to both backup servers. You have high speed
> recovery possible should you lose a server in the datacenter, or if you lose a
> whole datacenter. If you lose both, you're screwed.
> 
> I'm a little hurried in writing this as I'm about to run to the store with my
> wife. If something doesn't make sense or needs more explanation, please let me
> know and I'll address it.
> 
> Best of luck!
> 
> 
> 
> Hector Bojorquez wrote:
> > Thanks so much for all this great information.. .
> > I am looking at installing an older  tape drive --- Sony AIT SDX-D500c
> > 
> > I want to be able to backup to a tape nightly.... take the tape off site..
> > etc.
> > Ideally I would like to be able to do an entire system restore with such a
> > backup.