On 09-10-31 12:24:00, Matthew Saltzman wrote:
It used to be fairly common for new disks to have a few bad
blocks--back in the dark days of early PCs when disk drive capacities
were measured in tens or low hundreds of megabytes. Then things
seemed to improve as manufacturing techniques improved. When disk
capacities were measured in tens or low hundreds of gigabytes, I
don't recall ever encountering a new drive with bad blocks. Now that
capacities have reached the terabyte range, it seems that a few bad
blocks on new drives are once again less rare.
"Modern" drives from the last decade or more use formatting that
skips over bad blocks. There should /never/ be a bad block on a new
drive. As blocks will go bad from time to time in otherwise properly
functioning drives, all "modern" drives have automatic recovery and
remapping of bad blocks (but only for blocks where the data could be
recovered or the block was written to). (That IBM drive fiasco was
mostly that their recovery process was very noisy and dreadfully slow.)
It would be nice if the monitor software could record the state of a
drive and issue reports when the number of bad blocks increases from
the starting state, rather than insisting that every bad block is a
sign of imminent failure. The persistent false alarm provokes the
user to ignore the monitor or turn it off entirely, thus risking
missing a warning of an imminent real failure.
Right. SMART already watches that bad blocks don't increase too much.
I just let smartd mail me when it thinks the drive is going bad. Also,
I turn on Automatic Offline Testing, which scans the disk every few
hours and may be able to catch a block going bad while it is still
automatically recoverable and remap it without causing any trouble.