[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Q's on backup to CD-R

on Tue, Apr 23, 2002, Ross Boylan (RossBoylan@stanfordalumni.org) wrote:
> I'm trying to do a fairly comprehensive system backup to CD-R.  The
> amount I have to backup is much larger than the free space on the
> system, which in turn is larger than a CD-R.

CD-Rs are too small to be practical for any reasonably sized drive these
days -- 20 - 120+ GiB.  It's like backing up to floppies on a 100 MiB
system.  Tedious.

For spot archival, it's a decent idea, but real backups call for:

  - Tape.
  - Networked backup.
  - Tape.
  - Removable "fixed" backup (e.g.:  drive caddies).
  - Tape.

Tape's a bitch, but the cost, reliability, standardization, and
capacity, weigh strongly in its favor.  Still crazy after all these

DVD-R loses on several counts including size (2-4 GiB still isn't big
enough), and in particular, lack of standardization.


> Also, though I've always used tar, I understand it's a very fragile
> backup scheme and that afio (or maybe cpio--couldn't quite tell) is
> safer for archiving.  Any comments?

Internal structures are such that if your media or files are damaged,
you've got a better chance for recovery with afio/cpio.  That said, I
find the interfaces to both far more awkward than tar.  I'd bet on a set
of _differential_ backups from a known good comprehensive backup.
Preferably one or more cycles back:

Scheduled something like:

    week 0 day 0:  comprehensive backup
      w 0 d 1-7:  differential since w 0 d 0 (you're playing catch-up)

    week 1 day 0:  comprehensive backup
      w 1 d 1-7:  differential since w 0 d 0 (you're playing catch-up)

    week 2 day 0:  comprehensive backup
      w 2 d 1-7:  differential since w 0 d 0 (you've finally caught up)

    week 3 day 0:  comprehensive backup
      w 2 d 1-7:  differential since w 1 d 0 (you're moving forward)

From here forward, you're doing differentials since two prior
comprehensives.  Why?  Because you now have your pick of any of three
comprehensive backups, and (with diminishing recency), *** THREE WEEKS
*** of daily differentials which will get you at least _some_ of what
you lost.

Tower of Hanoi schemes and such can minimize tape use and archival time,
but the recovery process is twisted, and if _any_ backup is lost from
the set, you're hosed.  This enlightenment comes by way of a friend who
held backup tapes in his hand while wading through the server room,
explaining to PHB what their ToH backup scheme did and didn't optimize

My focus is this:  backups are a data recovery mechanism.  As such,
optimize for data recovery, and make a realistic assessment of your
costs elsewhere.  You know the drill:

   - DDS-4 40 GiB tape library:  $7000
   - Media cost:  $40
   - Nightly backups:  $2.50 for Starbucks.
   - Recovering lost data:  Priceless.

> By the way, here are two helpful sites I've culled from previous
> messages:
> http://kmself.home.netcom.com/Linux/FAQs/backups.html
> http://www.linux-backup.net/app.gwif.html
> The problem is a lot of the stuff doesn't address the peculiarities of
> CD-R.
> Final question, about what to back up.  Karsten Self's nice page
> (first reference) says some parts of /var are good to backup.  Which
> ones?  

The ones you need ;-)

If you'll drop lower on that page my script includes an environment
variable $backupdirs, including: 

     /var/backups /var/lib /var/log /var/www

> Also, I have a nagging feeling that some of /lib or /usr/lib
> has some customizations (vague memories of editing or adding files
> there, maybe for browser add-ons...).  

Mozilla adds its plug-ins there.  If they can be installed at all.
Broken IMVAO.


Karsten M. Self <kmself@ix.netcom.com>        http://kmself.home.netcom.com/
 What Part of "Gestalt" don't you understand?
   GNU/Linux & BSD:  We *are* the way out.

Attachment: pgp2guLbMeWvN.pgp
Description: PGP signature

Reply to: