On 4/12/18 3:16 am, Phil Turmel wrote:
On 12/3/18 12:49 PM, Niklas Hambüchen wrote:
On 2018-12-03 18:35, Phil Turmel wrote:
Your drives appear to be fine. I suspect you have a problem with other
hardware in this box.
When I repeat `smartctl -t short` on these disks, it fails at exactly the same sector:
Disk 1:
Num Test_Description Status Remaining LifeTime(hours) LBA_of_first_error
# 1 Short offline Completed: read failure 40% 16424 7501728
# 2 Short offline Completed: read failure 40% 16398 7501728
Disk 2:
Num Test_Description Status Remaining LifeTime(hours) LBA_of_first_error
# 1 Short offline Completed: read failure 50% 16424 1758544
# 2 Short offline Completed: read failure 50% 16398 1758544
Doesn't this suggest that this is not unfortunate behaviour of a power supply, but permanent damage to the disks (even if originally caused by a power or power supply problem)?
Those failures should increment your Current_Pending_Sector attribute in
those drives. But you say those remain zero. So I'm stumped.
Yeah, that's weird. A SMART test will abort on the first error and it
always bumps the Current_Pending_Sector counts (well, it has on all my
drives anyway).
Try running a read on the disk with :
dd if=/dev/sdX of=/dev/null bs=1M conv=noerror
That will read every sector (or block of 8 on a 4K drive) and will keep
barging through after read failures. At the end of that you'll likely
have a spray of Current_Pending_Sector(s) which will give you an
indication of just how bad things are.
I'm with Phil. It sounds like a power issue. If there is a power hiccup
while the drive is writing, it'll occasionally write out a corrupt
sector. That turns into a URE as the ECC doesn't match, so the drive
never gets a clean read. That also means it'll never auto-reallocate on
read.
When you attempt to write it, it'll write cleanly and therefore leave no
trace of reallocation in the SMART data (because it hasn't),
In the case of a genuine bad sector, the drive will try and write it,
that will fail, so it'll try and write it elsewhere (reallocate). If the
drive is truly dying it'll run out of spots to reallocate to and
complain loudly. That's pretty rare however.
I have an old Toshiba laptop drive here (in a drawer) with a
reallocation count in the high 4 digits, and it still has plenty to
spare. The impact of that on system performance was massive as it was
always seeking all over the platter to get the out of sequence
reallocated sectors, but it was still working and reliable.
Regards,
Brad
