Saturday, September 03, 2005

I really can't win

You might remember a while ago[^] that half of my RAID array died and when I replaced the drive I'd bought one just a tad too small so I had to do a reinstall. And you might also remember that a little while later[^] the second drive in the original array also died.

That was about half a year ago and I thought I was good to go for a couple of years at least.

No such luck. A fortnight ago, during that incredibly brief stay I had at home between France and The Philippines one of the new drives died. I had better things to do with my time then than to worry about it; that was a problem that could wait until I was back from Baguio. So I shut the machine down, took yet another flight and forgot about it.

Got home last night and fired the machine up. The RAID BIOS warned me that there was a critical error so I followed the standard steps to establish which drive was at fault. Like all good RAID BIOSes this one let me break the mirror and boot using the good drive. Which I did. And immediately proceeded to burn all the stuff I wanted to keep to a DVD (including my blog as text files).

Boy am I glad I did. Maybe 15 minutes after the DVD burn and verify had finished the machine blue screened, accompanied by the sound of the 'click of death' from the surviving hard disk drive. I spent most of today rebuilding the machine with a new 120 Gig drive. And, given my experience of RAID and cheap hard disk drives, I decided not to buy two drives. Yep, I'm back to plain vanilla single drive operation.

Maybe your experience of RAID has been better than mine; I've been running on RAID for about 6 years and mostly it's been ok but whenever it's failed I've been in the position where replacing a drive hasn't helped. Yep, that time when I bought a drive a trifle too small was probably my fault but the most recent experience, of two drives failing within a fortnight of each other has just about killed the idea for me. I know there were extraordinary circumstances surrounding it; my travel interrupting the repair process for a couple of weeks. But the surviving drive had been idle for those two weeks. If a consumer drive is so susceptible to stiction and to prolonged operation then I'm no longer going to rely on redundancy in the hardware; I'll just do more frequent saves to more permanent media and have done with it.

No comments: