I work in a datacenter with large numbers of un-raided servers. Generally when someone wants to fix a drive, they just want their data off. Corrupted Filesystems due to Physical Problems: Corrupted filesystems are frequently due to bad blocks in the filesystem metadata. The fs metadata tends to go first because its the most read part of the disk. I've had really good luck with ddrescue for this sort of error (at least for ext3). Have ddrescue skip error blocks and keep a log of bad blocks, otherwise it'll literally take a week to recover. (Instructions: http://www.forensicswiki.org/wiki/Ddrescue) Fried Drive Controllers: These will generally completely fail to turn on or read at all. They're usually not detected as disks. Replacing the PCB would probably work if I were any good at hacking type soldering. If you're tempted to try sticking a drive in the freezer, just let it sit for 1-4 months instead. I believe it's effectively the same fix but with far lower of a chance of borking the electronics due to mosture. Believe it or not a fair number of drives will come back to live after this period of time (~15-20% I would *guess*). Mainly you should just be aware of the warning signs. Disappearing files, folders that cause crashes, ext3 related stack traces, and filesystems being auto-remounted as read-only are all signs that its about time the evacuate to a new disk within a day, two at the max. Bad ball bearings generally don't kill hard drives. Disks making weird unlubricated drive bearing/shaft sounds can still work for a year or so. If this disk seems to shutter or obviously has problems starting to spin you should definetely copy your data to a new disk, anything less will mainly just injure people's hearing. The main problem with bad bearing is that it *really* increases the amount of heat in the computer (which in turn can kill hard drives).