[kwlug-disc] Checking a USB disk for errors

[unsolicited]

I thought that was the problem with smartctl, usb or no - it's merely 
reading the smart, and smart is problematic.

I thought the only real way to test was to exercise (preferably write) 
the entire drive. Multiple times even, so that bad sectors get marked 
out beforehand rather than on the fly. (Nasty when the last spare sector 
is used up and you hit the next bad sector.)

It also seems to me that this list's common wisdom was that when you 
start seeing bad sectors (outside of smart), the spare pool is used up, 
the drive is going (eventually), and you should no longer trust it 
absolutely as a live drive. (Backups OK, to the extent the drive 
survives, on the theory that you can re-write a backup after replacing a 
bad backup drive. Even better if it's your secondary backup.)

The problem being there's no way to know except after the fact, and 
after the fact it's a PITA!

IIRC, my final arbiter on whether a disk is healthy or not is the 
manufacturer's self-booting utilities. Granted, the problem is the drive 
being on USB, and I don't recall last time if their utility was USB 
aware or not.

In any case, for the LaCie/Samsung drive, it will be the Samsung utility 
that matters, not LaCie. USB enclosed drives are just drives like any 
other - I frequently pop drives in and out of them. If you can pop the 
drive out and get it more directly connected you may get where you're 
trying to go faster. (Even eSata is just Sata, using it if you have it 
you'll be fine. It's the USB/Sata transport change that's problematic.)


On 14-05-16 04:25 PM, William Park wrote:
> On Fri, May 16, 2014 at 10:08:33AM -0400, Khalid Baheyeldin wrote:
>> # badblocks -b 4096 -c 4096 -sv /dev/sdb 23000000 22135600
> ...
>> Pass completed, *2 bad blocks found*. (2/0/0 errors)
>>
>> But now after repeating the destructive test (-w) a few times, the error
>> disappeared?
>>
>> # badblocks -b 4096 -c 4096 -svw /dev/sdb 23000000 22135600
> ...
>> Pass completed, *0 bad blocks found*. (0/0/0 errors)
>>
>> Did the drive reallocate the sector and hence it is no longer reporting
>> that one as defective?
>>
>> Is it safe to use the drive, or it would be suspect? There is no click of
>> death or grinding sounds.
>
> Try scanning the whole disk.  If that passes, it should be okay.
> You can try using 'smartctl', but, at least for me, it never finds
> anything wrong on disks that I know are dying.
>