Sunday, February 01, 2009


The death of my Seagate 1T drive found me with my pants down: I was testing my new PC, all was working acceptable and the drive had worked ok for almost 2 months, so I decided to move all my data on this drive, for convenience.

Trusting Seagate was a stupid move. Just after days of moving all my personal data on that drive, much with no backups, the drive stopped working because of the Seagate firmware bugs.

Now I have removed all the Seagate drives from my PCs. I have cleaned the 1.5T drive and will try to return and get back my money in the following days. The dead 1T drive is still laying around and I'm searching for a way to recover my data from it. After that, I'll return it, too.

The new PC is running now with 2 Samsung drives (in RAID1) I was already having. They're slower than the Seagate, smaller (400GB vs 1TB) and make more noise. On the other hand, they work while Seagate drives don't.

The problem is that I've become so paranoid that even RAID1 is not enough. I was wondering what to do and then I've realised:
  • my really-really important data is not that big
  • I have a 60GB USB drive that I haven't used lately and is just gathering dust on my desk

So why not use it as a backup storage medium?

After some skimming over the available backup solutions (the solution had to be in Ubuntu ready for use), I've decided to use dirvish.

It's not the simplest or the most complete solution out there, but, for me, it has a good balance between features and configuration complexity. Here's some of the features I like about it:
  • the backups produced by dirvish are transparent - they are exact copies of the files you back-up and you don't need to decompress them.
  • new backups are incremental and only changes are copied
  • if a file hasn't been changed, it creates a hardlink to the old one; this way space is saved and if your files change rarely, making backups frequently will increase the space needed by the backups just marginally.
  • because it uses hardlinks, you can delete older backups without worrying you will loose some important file from (when using deltas with archives this can be a problem - you have to keep your full chain of full and intermediate backups or you will not be able to reconstruct the latest backup - there's no such problem with dirvish and your latest incremental backup is as ready to use as the first full backup you've made)