sector hard errors

[Chris Murphy]

On Thu, Mar 12, 2015 at 1:20 PM, jd1008 <jd1008 at gmail.com> wrote:

> [  117.660065] sd 4:0:0:0: [sdb]
> [  117.660703] Add. Sense: Unrecovered read error - auto reallocate failed

What is sdb used for? Single drive or is it in some kind of raid?

This is a URE with failed reallocation which means the data and
checksum for this sector mismatch, and the drive's ECC can't recover
it. Since it can't reconstruct it, the data can't be moved so it just
stays here. The problem isn't corrected until the sector is
overwritten.

What do you get for:
smarctl -l scterc /dev/sdb


> So, I am puzzled as to how quickly were the spare sectors consumed so that
> automatic sector forwarding ran out os spare sectors?

The data on the sector is just bad. The firmware apparently doesn't
think the signal encoding on that sector is weak or otherwise
indicating the sector itself may be bad, like a surface defect. I'm
guessing it's getting deterministically bad data and it can't fix it.
If it were a partial read, then that might suggest a bad sector, at
which point it'd be flagged as pending reallocation. This is
conditional. Upon write, the drive firmware determines whether it's a
transient error or persistent, if it's persistent then it'll use a
reserve sector (reassigns the same LBA to a different physical
sector).


> So quickly on a
> brand new drive??
> When someone buys a brand new drive, how many spare sectors is it guaranteed
> to have? ZERO? ONE??  How many?

Depends on the drive model, use case, and manufacturer. Any enterprise
SATA or SAS drive, it's a swap out for just one of these. For
consumer, it's stick a wet finger in the air. I think most of them
would accommodate, but arguably the drive is functioning normally - so
far. But you're right, it's brand new and to have a sector read error
on a brand new drive is unexpected.

-- 
Chris Murphy