Scott H wrote: > > I am working on methods for recovery of a Linux > server, and I wanted to run this by the group for > comments or recommendations, since I'm new to > this. I have a server which acts as an email > relay, storing no mail locally. It also stores > no other data, besides the configuration data for > its mail systems. I plan to set up an automated, > regular complete backup for it, but I was also > considering this: Say I have among my partitions > on this box: > > sda2 / > sda3 /backups > sda5 /var/log > sda7 /var/spool > > Let's say I set up mirrordir to keep /backups > (sda3) mirrored from / (sda2) maybe once per > night or week. Then, if I manage to mess up the > OS (as I recently did, running tripwire and > up2date at the same time - don't ask) and the > machine won't boot, I wouldn't have to resort to > a rebuild and the tape backup. I could boot the > box to my LNX-BBC CD (or a Tom's root boot), > change fstab to mount sda3 as / and sda2 as > /backups, and reboot the box. As long as > mirrordir hadn't yet copied whatever corruption > was in the OS, I should be able to boot and go > just as things were the last time mirrordir ran, > right????????? Better ways to do this? > > Scott Hi Scott, Here's what I did at a web hosting company. I had another drive installed that mounted on /backup at midnight (maybe later, depending on server load at midnight). My backup script copied /etc, /root, /home, and /var. That was it. Everything else could be rebuilt from the master image. You could follow this same scheme (if using Red Hat). If you hose your OS, or the drive fails, reinstall using your kickstart config file (located in /root), then restore your data from the other drive. I use this method for my firewall, and I tested it (bare metal recovery). The scripts to do this are on my site (georgetoft.com - linux - cool shell scripts). Your idea is pretty slick, though - switch partitions in fstab, reboot and go. Keep in mind what happens if your drive fails. Regards, George