On Wed, Oct 22, 2014 at 11:00:09AM +0100, Mark Rogers wrote:
On 21 October 2014 17:33, Chris Green cl@isbd.net wrote:
A file that doesn't change has hard links from multiple snapshots to the same data on the disk so there's only actually one copy of it.
Actually I think it is important to think of rsnapshot as a bit of a hybrid, which takes incremental backups but presents them as multiple snapshots.
One of the limitations of taking "snapshots" but using hard links is that you only have one copy of each unique file version (that is the reason to do it, of-course). That means that if you lose a few disk sectors on the backup drive, any file that gets borked is borked in all "copies" of the backup where that file was unchanged. In a "real" backup with multiple snapshots you could take the file from a different snapshot.
This is a "feature" that it shares with traditional incremental backups, where failure of a small part of the backup media can affect multiple backups. This can be mitigated against with RAID etc.
Any backup (even multiple actual copies of a file) that goes to a single drive is basically vulnerable to a failure of that drive.
I do frequent (as in hourly/daily) hard-linked snapshot backups to another drive on my desktop machine and less frequent (as in daily, monthly, yearly) backups to my NAS in the garage (which is a longish way away from the house). So, hopefully, I've got most things covered. Of course there are particular cases that aren't covered but you can't do *everything*. For critical stuff (like company files, accounts, photo archive, etc.) I add other backups but they are just "as it is now" copies with no history.
That said, as far as the topic of this thread is concerned: Backups *are* pretty simple, by their nature, but any solution will evolve to take into account the fact that real life is not. Ideally you'd just take a complete duplicate of everything onto 100% reliable media every day and never worry about it again, but disk capacities, transfer speeds, media reliability, etc all result in trade offs. I'm pretty sure that rsnapshot will have started out as a simple script (along the lines of what Chris has now) and evolved to deal with the real world complications. Some won't effect Chris, some might in the future even if they haven't yet. Trusting a backup means trusting the backup mechanism, and if Chris has more trust in hist own solution than one that's been tweaked and tested by a larger group of people over some considerable time then that's fine, it's only his data that's at risk.
The advantage is that it's simple because it only does what I want it to do rather than having the complexity needed to provide flexibility for multiple needs. It also means I can change it if/when I need to.
Indeed, having a simpler system that he fully understands may well increase the trustworthiness. For my part I would prefer an established solution over a roll-my-own. But the bottom line is that far too many people simply don't have backups at all, and any of the options discussed here are likely better than that.
Agreed! The backup system needs to 'just work' too, with no effort or interaction from the user.