On Mon, 2005-08-15 at 09:05, James B. Byrne wrote:
We also are looking at tape systems. The idea is that we will end up with a four stage backup process. 1. Daily automated backups of each server over the network to a local high capacity linux server devoted to hosting the backup volumes.
Backuppc (http://backuppc.sourceforge.net/) is excellent for this and it's compression and duplicate file linking scheme will greatly increase the on-line capacity.
- Daily automated moving of certain
critical, volatile backup sets from those on-line local stores via ssh to an off-site server performing the same function.
Depending on network capacity and your backup window, you might simply run an independent copy of backuppc offsite, using rsync over ssh for the copy. This has the advantage of continuing to work if your main backup server has a problem.
- Daily semi-
automated (somebody has to change the disc) stores of less critical and only somewhat volatile data onto DVD kept on-site in a fireproof vault.
Backuppc has a manual command to write the last backup of a single host out as a normal compressed tar image either to a tape drive or as files split to a certain size.
- Semi-automated archiving of the rest onto tape on a
weekly basis and moving these off-site (the backup job is automatically scheduled but the media has to be manually handled).
If you are archiving these for a long time, using the archive function to write tapes is probably the best approach with backuppc. However, if the object is just to get copies offsite, you might be able to use external drives. The huge number of hardlinks makes file-based copying impractical so you may have to raid-mirror or do LVM snapshots and image copies of the snapshot to make the copy in a reasonable amount of them. The space-savings of the storage scheme makes this a nice approach. I'm doing it with a 250 Gig drive that typically has around 100 Gigs used, but it holds a week's backup runs of 28 machines and the uncompressed data would total over 600 gigs. I can plug the external drive copy into my laptop which also has backkuppc installed and have instant access to any of the files from any of that week's runs.
The total volume of data handled by all these these processes is not prohibitively large. We are speaking of tens of giga-bytes (~40GB) over all.
Depending on the rate of change and your bandwidth to your offsite location, you might be able to let an offsite instance of backuppc run a complete independent copy and not bother swapping anything around. Using rsync for the copy method means that only the changes are going to be moved on each run. Five of the 28 machines I mentioned are in remote offices.
--- Les Mikesell lesmikesell@gmail.com