Reallocated sector count is the only S.M.A.R.T stat I've ever seen any use from, and that count doesn't go up until your drive is quite into the danger zone. Also, for most drives I've had failing, that count has been 0.
-f0dder
Thanks for sharing your experience. I guess you learned about your drive failures by other means -- presumably through observing odd / errant behavior, is that right?-ewemoa
Yes.
If you're at the same physical location, the change in sounds from the drive when doing typical access patterns is often a dead giveaway - at least for 3.5" drives, 2.5" are do damn silent you might not notice.
I recently had the main drive (120GB 2.5" WD something-something) in my linux server crash. I started noticing that I'd sometimes get freezes in Minecraft, long SSH login times, and then stutters when copying files. Minecraft is Minecraft

, but I had a good idea what was happening when the SSH login times ramped up. When the file copying started stuttering is around the time when DMA errors showed up in 'dmesg' output. I let it run a bit too long before getting a replacement drive, but PANTS-ON-HEAD-RETARDED-luckily lost no data (and apparently no corruption either - some data had .torrent files to check against, others .md5sums). Took around 40 hours to get all data off the drive, though.
I really should set up some dmesg logging that scans for DMA errors and the like... 40hz or Stoic (or anybody else) got any ideas?