On Fri, 2007-05-18 at 11:25 +0100, Adam Bower wrote:
I have also known disks to fail when they've been moved from one data center to another, I believe this is to do with lubricants that are only needed to help spin the disk up going bad over time and if you let the disk get cold then the bearings shrink and the motor can't get the disk spinning again (or something)?
You get stiction in the platter motor bearings which combined with old lubricant, and worn cold bearings can be too much for the platter motor to overcome. There is a sort of twist/snapping motion that you can do which uses gyroscopic forces to overcome this in extreme circumstances...but essentially a drive doing this has "problems" This is the reason that one of the things SMART monitors is platter spinup time to try and preempt this.
Also the head-platter gap is maintained purely by aerodynamics and the head actually rests on the disk surface when it is stationary. There is a landing zone on the disk for this to happen away from data and the heads don't move from this point until the disk is up to speed. However repeated landings or landings on a contaminated or damaged disk surface can cause problems with the heads.
Desktop/server drives can in some cases have a lower tolerance to this than ones designed for mobile computing. Which is why you should exercise caution when applying drive spin down power management settings to desktop machines. I have seen some desktop drives with a design spin up cycle count of 10'000 events which if you had really aggressive drive spindown times could result in early failure of said drive