bad HD sectors
dihnat at dminet.com
Sun Jun 20 13:29:29 UTC 2010
On Sun, Jun 20, 2010 at 12:00:59AM -0700, JD wrote:
> So, it is not clear that fsck will detect new
> bad blocks.
Man 'badblocks'. Note, from the manpage:
"...it is strongly recommended that users not run badblocks directly,
but rather use the -c option of the e2fsck and mke2fs programs."
> I read about the old days when the drives did not do auto bad block
> forwarding (re-mapping).
I lived through "the old days". (Funny....they don't seem that far
back...) I still have a full-sized 5.25" drive in the basement...acting
as a doorstop.
> There was a low level formatter which actually detected the bad blocks
> and remapped them. Unfortunately for today, once the drive runs out of
> spare good blocks to forward bad blocks to, throw it away. I really
> wish there was such a low level formatter for today's drives. It would
> extend the life of a drive a little longer.
No, you don't. Having blocks spontaneously go bad, and having to scan
the system for them, was a Bad Thing(Tm). It took a lot of time--and
drives were far smaller than today. If you didn't catch 'em--and scans
were skipped often, because of the time required--you could suffer
unrecognized damage for a long time.
At least, with internal drive intelligence, and S.M.A.R.T. (SMART)
reporting, you can run quitely and with a fair degree of comfort in the
probability that everything is being properly stored, with bad blocks
being detected and remapped without constant scanning on your part.
The fact that so many people, with operating systems of all kinds,
*don't* turn on SMART monitoring and can run for years without failure
says it is successful.
The concomitant flip side is that when you finally *do* start seeing
SMART errors, you know that so many sectors have failed that it's filled
the reserve. This is cumulative--by the time this happens, there' no
real advantage in trying to 'low-level format'; it may never complete,
but even if it does, you've no confidence that it's going to remain in
service. SOMETHING was failing--either the controller electronics, or
the drive's recording surface. It's time to replace it.
The bad news is that customers have rewarded drive vendors not for "fast,
big, and reliable", but "fast, big, and cheap". Drives generally fail
more often than you'd want. The good news is that "cheap" part--drives
that are stupidly big can be found for pennies on the gigabyte. 1TB
SATA for $64 (shopper.cnet.com).
Hell, I remember in the early '80s--I was on contract at Bell
Telephone Laboratories in Naperville, IL. They had a **huge**
celebration--commemorative coffee mugs and other chachkes, cake and
coffee, at a big outdoor ceremony. What was the occasion?
The data storage for the entire data center had just crossed the magic 1
dihnat at dminet.com
More information about the users