[NMLUG] Topic for meetings.

nmlug@swcp.com nmlug@swcp.com
Tue, 6 Nov 2001 21:42:15 -0700 (MST)

> This far my own solution would be to have an old crappy box sitting across
> the network (probably in a different room) with two high-capacity drives

At work, for a while (but not any more), we used a similar solution--
except our other box was located in Tokyo -- ie, we backed up our
Albuquerque server to our Tokyo server. I, personally, felt the idea was
inefficient, but it did have a certain bizarreness value, plus the cozy
feeling that even if the entire state of New Mexico were reduced to
scorched rubble, at least our data would be safe.

> you have a way to be notified when something goes wrong. The hardest part is
> going to be writing all the shell scripts to make something like this work
> and work well.

The basic concept is crontabbing something like this:

for i in (`find / -newer /.lastmirror`) {cp $i /mnt/$i} ; touch /.lastmirror

IE, basically use "/.lastmirror" as a timestamp of when you last copied
files over, and re-copy anything that's changed since then. Refinements
include handling deleted files, handling files that change while the
script above is running, handling symbolic links, excluding certain
directories (like /tmp) from being backedup/mirrored, reporting errors.
The script I finally wrote does a pretty good job of handling all that
(though directories that are symlinks are especially painful).

An easier solution (if you control both machines) is something like rsync.

It might also be neat to consider a redundant, reliable,
non-excessive-bandwidth-consuming, secure, encrypted scheme to backup data
to free webhosts (geocities, freedrive, hypermart, tripod, et al).

- Sarang

To UNSUBSCRIBE send a message to nmlug-request@swcp.com
with only the word unsubscribe in the body.  More
information can be found at www.nmlug.org/info.html