I think I'm coming to the conclusion that RAID can be more of a problem than the ones it tries to cure. I've long been a fan of RAID, having used it in the mid-90's on corporate servers, when it was veryy, very expensive, but invaluable in saving our butts when discs failed. For general machines, if possible, I'll try and mirror the discs, giving me redudancy, but on my main server (domain controller, exchange, etc), I bought an 8 channel SATA RAID card, 8 discs and two drive caddies (not all at once). The first caddy was RAID 5 using 4 * 160Gb discs (3 + a hot spare), while the second was JBOD (5 * 500Gb, for archives and the ripped DVDs).
Recently I noticed the caddy beeping and it showed a disc failure; the spare hadn't kicked in, so I just rebooted and it rebuilt the array. When a new disc arrived I just plugged it in and configured that as the spare. I did get some disc corruption though, requiring me to fix my exchange store, but everything seemed OK. Then, last week, at 8am on a Sunday morning, as I was getting into the car to head to the airport, I heard the beeping again; not having time to do anything about it, I just prayed that it would last the week until I was back. No such luck; on Wednesday afternoon Web mail died, indicating the server was dead. This may explain why any mail to me bounced for two or three days, until I had a chance to setup Mail Enable as a temporary stop gap; this is an excellent SMTP product and took less than 5 minutes to get up and running; I'd be happing using their pro version, but I already have Exchance licenses and it seems wasteful to pay for something I don't need.
On arrival back home I found the power supply for the drives (it's a separate power supply than the machine itself) rather hot; I switched off and the next morning nothing. An old power supply, driving 8 drives -too much for it. I tried a new supply and booted up, but the controller couldn't rebuild the array. Bugger. Of course, I know that RAID 5 doesn't protect against multiple disc failures, but that doesn't usually happen - disc reliability is excellent these days; what I hadn't counted on was a complete failure in writing to all of the discs, thus corrupting the lot of them. Luckily I have backups, done to another disc. What I don't have backed up though, is my music, which is going to mean a long job of re-ripping. Sigh.
I've spent the day re-installing the server, but this time have used a 40Gb plain IDE drive as the main OS drive. I'm now going to setup Acronis True Image to image that disc every night, so if the IDE drive fails I can simply put the image back - a matter of minutes rather than an entire day of reinstall. The RAID array will be re-configured for day to day storage as before; the second array will hopefully be OK, since nothing writes to that.
I'm going to setup True Image agents on my other machines to automatically image overnight, as well as the multitude of other backups I'm using; the day-to-day dev box has an xcopy script (which copies to the RAID array), OneCare which backups to another drive and Mozy as well. I've not lost anything apart from time, and quite frankly, I can't be bothered to lose any more time if something goes wrong again. As it is I'll probably have to reconfigure my other machines, since a new install meansa new SID on the domain, so logins will probably change. Luckily it's been raining all day at the Grand Prix was on, so I had something to do while waiting.
Incidently, why is the Exchange install process so annoying? You first do domain prep, then forest prep, then install exchange, all three of which require you to type in the product key. Very annoying.
Now to image the server, then try and restore the backups, then image again if the restore works.