Today, I feel even more like an idiot than my self-deprecating character normally allows (it's important to keep your feet on the ground in the heady, glamorous world of IT journalism).

After over ten years of dedicated service as a file and print server, and probably in sympathy with the recent demise of its manufacturer, my Elonex ProSentia server died and could not be resuscitated. 

It had a good innings, I reckon, plus the sporting grace to retire in a relatively quiet period of my working year. If only my own backup and disaster recovery strategy could have matched its understated but effective consistency. 

Because despite writing numerous articles over the last ten years on how to avoid losing your data in times of hardware failure, my own approach to the problem left a lot to be desired.

The last good full backup I had was three months old, and stored on a cartridge that required a DAT 72 drive and SCSI card to restore it. I found the drive, and I was sure that SCSI component was in the server last time I looked; but it wasn't there now and I didn't have a spare.

Over 40GB of data become instantly inaccessible. But the one thing that saved my bacon is that all my really important, regularly accessed files are kept on a USB stick cum MP3 player that travels with me wherever I go. This is not the most secure option, I know (well, it is at least password protected) and means I am occasionally stuffed when I accidentally leave it attached to the PC I last used.

To get around the latter problem, I make sure that incremental changes to the files on the USB stick are replicated onto the hard disk of my home, office and laptop PCs, and from there onto the server, at both the beginning and end of every day, using nothing more sophisticated than Windows' backup utility. Which means that I can always retrieve data that is no more than a couple of days old from four different sources according to convenience or location.

It's an amateurish, ad-hoc, some might say haphazard way of going about things which certainly raises some pertinent management and security questions, but it did actually work.

All that's left now is to find a new server and restore the other 40GB, though I suspect that sifting through all that information to decide what actually deserves or needs to be restored (and backed up in the future) will open another can of worms.

07 Aug 2006