Hi guys! Im trying to design a sound and robust long-term backup strategy for my personal data.

One of my defense layers will be Dropbox. I have a free account which should work fine when it comes to retrieving recent versions of my files. But a free Dropbox account is not a good solution for long-term backups. And I do not wish to store all my files here, e.g. documents containing financial information. Also, because I have limited space I dont intend to store heavy files here, such as my photo collection or my music collection.

Another layer will be hosted backup. I will most certainly go with Colin Percivals Tarsnap here, a service I have great confidence in. Tarsnap encrypts the data client-side before sending it to the server, so I dont need to worry about anyone else getting access to my data. But I dont feel comfortable about depending solely on hosted backup. Even if I use a reliable service there are still ways to screw it up, e.g. by forgetting to pay for the hosting or by displacing ones private key, which will render the Tarsnap backups totally useless.

So I still want hard-drives as another layer. And I dont think one single hard drive is enough. As we all know, hard drives will fail sooner or later. And if I have backup on multiple hard-drives I can add another layer of geographic redundancy by placing these hard-drives in different locations.

However, making backups to multiple hard-drives seems to be quite tricky to say the least. Ive looked through the documentation of various open source backup solutions such as bup, zbackup, obnam, attic and BackupPC, but the documents Ive found doesnt really address this issue.

Ideally the backup system would have some kind of distributed index to keep track of where (on what disk) some version of a file is stored. If I lose file X, it would be very convenient if I can search the index to know what drive X is stored on. (Recently created files may not have been backed up to more than one hard-drive.)

I did find a Bacula add-on called vchanger, but as far as I can tell it seems to be quite complex to setup. And AFAIK Bacula also requires a home server, something that I would like to avoid if possible. So I hope to find a simpler solution.

So now I would like to have some feedback from you guys about my ideas. Is it possible to achieve what I want? It might be that I'm overthinking the issue and that there is a simpler yet adequate solution.

tl;dr: I want to write my backups to multiple hard-drives, that I can store in different locations. I would like to avoid having a home server if possible.

I just backup the whole drive using .tar and then gzip after booting using a live CD.

tar -cvf Backup.tar /media/xxx
gzip Backup.tar


Extract

gunzip Backup.tar.gz
tar -xvf Backup.tar

Final size ~5-8Gb depending on what you have installed.

Check out Duplicity.
I've never used it but it has "versioned backups" whatever that means.

Hope that helps.

1 members found this post helpful.

Thank you for the suggestions guys!

The tools I mentioned above would let me do something similar. Im not that concerned about disk usage since hard-drives are quite cheap these days. What Im concerned about is keeping the system somewhat consistent. It would be great if I could do snapshots and then be able to see where version n of file X is stored.

I'm a little weary of relying solely on encrypted backups. As I said earlier, if you lose the key you lose the backup.

An acquaintance of mine suggested Microsoft Onedrive. An open source/non-Microsoft alternative could be rsync.net. Going that route requires a lot of trust in the hosting provider, which might still be a reasonable trade-off.

If I go with hosted backups I could still complement with e.g. bup for offline backups. In order to mitigate the risk of losing the external hard-drive I could regularly clone it to another drive with dd. Of course cloning would not be entirely perfect since any file-level issue/corruption on disk A would propagate to disk B. And this solution would not get me a consistent index either. When I need to restore a file I might need to look in three different places in order to find it Dropbox, rsync.net and the external hard-drive. But that is feasible, if not maximally convenient.

I will also consider the ZFS file system. It seems like FreeNAS could support offline disk rotation, though setting up such a solution would probably be quite time-consuming and expensive.

So now I have some paths to further research. But feel free to add additional suggestions or feedback. The more suggestions, the better!