On 03/06/11 00:44, Mark Wagner wrote:
On Thu, Jun 2, 2011 at 16:11, Neil Harrisneil@tonal.clara.co.uk wrote:
Tape is -- still -- your friend here. Flip the write-protect after writing, have two sets of off-site tapes, one copy of each in each of two secure and widely separated off-site locations run by two different organizations, and you're sorted.
The mechanics of the backup are largely irrelevant. What matters are the *policies*: what data do you back up, when do you back it up, how often do you test your backups, and so on. Once you've got that sorted out, it doesn't really matter whether you're storing the backups on tape, remote servers, or magic pixie dust.
Not quite.
You're right about procedures, but you can't begin defining procedures until you have something concrete to aim at.
Tape is the One True Way for large scale backup, even today (ask Google), and I thought it might be useful to give an illustration of just how cheap it would be to use. Tape is a great simplifier, and eliminates a lot of the fanciness and feature-bloat associated with more sophisticated systems -- more sophisticated is not necessarily better.
Here's a straw man proposal for procedures:
I'd suggest backing up _everything_ -- cluster servers, local office IT servers, staff PCs, the lot -- for WMF internal archive and disaster recovery purposes. Something like monthly incremental backups, filing away media to the remote sites after verification, and yearly or six-monthly total backups to a complete new set of fresh media. For only a month's worth of work, replicated disk copies is fine: the tape archive is a back-stop, for when the replicated disks fail.
Dumps for external archives could also be made using the same drives, but to different media, and with a much more restrictive policy about what is saved.
-- Neil