Re: [smartmontools-support] strange error

[Milos R. <mil...@dc...>]

Thanks this looks logical
Since the disk is working in raid 1.

Even the sector is readable I am thinking of two possible solutions:

a) dd if=/dev/sda of=/dev/sdb count=1 bs=512 skip=502093169
b) remove from raid and then return to raid for synchronization

Somehow I am close to believe that only somekind of writing to that 
sector will make changes in the smart log.

Regards,

Milos

On 06/27/2013 03:52 PM, Gregory Sloop wrote:
> My comments are probably useless, as this is out at the edge of my
> knowledge - but I seem to recall that a failed read will continue to
> have problems until a write is made to the same problem block.
>
> Once a write is attempted, it will either succeeed, in which case you
> should be able to read it [transient error - not likely though],
>
> or [much more likely] it will fail to write the block, and
> will mark it as "bad" and remap it to one of the reserved spare
> blocks and future reads will then succeed.
>
> But that re-mapping doesn't occur on reads, only writes. So, until
> something tries to write to the problem block that it can't read now,
> the "read" failure will continue.
>
> So, provided that's right - an operation to force a write on that
> problem block should right things. [The main question before doing
> this is, do you have real data on need in the block currently. If no,
> then just attempt to write it. If yes, then recover the data using
> ddrescue or similar before writing to it.]
>
> HTH - but take with a grain of salt - I'm not completely sure I'm
> right - but that's the way I understand it.
>
> -Greg
>
>
>
> MR> Nothing changed the report is still the same.
> MR> As far as I understand the system pending sector will be replaced only
> MR> during write
> MR> Also tried another short test and same error report occurred.
>
> MR> Generally I have never seen smart error during test and empty Error log
> MR> report where system places read errors or other problems.
>
> MR> ddrescue is very powerful tool when you want to save the data from disk
> MR> because it will reread wrong sector in different way several times. But
> MR> according to my experience any other tool that (dd, badblocks, fsck, or
> MR> hdparm) have problem with reading the sector will definitely  make error
> MR> report log to the system and it would be recorded into smart data with
> MR> details.
>
> MR> Thanks,
>
> MR> Milos
>
> MR> On 06/27/2013 06:55 AM,ro...@sp...  wrote:
>>> This is has been a very educational thread for me so far. A couple of
>>> [probably newbie] inline questions:
>>>
>>>> The newer extended logs support 48-bit LBA. Please try:
>>>>
>>>>      smartctl -l xerror -l xselftest ...
>>>>
>>>> or:
>>>>
>>>>      smartctl -x ...
>>> Does this mean I can't trust the SMART Attributes to give a complete
>>> picture of the hard drive health?
>>>
>>>> It is not uncommon that self-tests are aborted immediately by the
>>>> firmware if a pending sector still exists.
>>>>
>>>> Please check the full disk with some tool. I prefer GNU ddrescue, which
>>>> was also suggested by Alex.
>>> I've used ddrescue once, but not for doing a pending sector read. Am I
>>> correct in guessing the command would something like:
>>>
>>> ddrescue /dev/sdb /dev/null
>>>
>>> If the purpose is to just force the disk to re-read the pending sector,
>>> wouldn't the other two methods work equally as well, i.e.
>>> sudo hdparm --read-sector 502093169 /dev/sdb
>>> or what Milos did
>>> badblocks /dev/sdb (perhaps with a -b and/or -c option to speed things up)
>>>
>>> BTW Milos, after you did the badblocks, did the pending sector go away?
>>>
>>>
>>>
>>>
>>>
>>>
>>> ------------------------------------------------------------------------------
>>> This SF.net email is sponsored by Windows:
>>>
>>> Build for Windows Store.
>>>
>>> http://p.sf.net/sfu/windows-dev2dev
>>> _______________________________________________
>>> Smartmontools-support mailing list
>>> Sma...@li...
>>> https://lists.sourceforge.net/lists/listinfo/smartmontools-support
> MR> ------------------------------------------------------------------------------
> MR> This SF.net email is sponsored by Windows:
>
> MR> Build for Windows Store.
>
> MR>http://p.sf.net/sfu/windows-dev2dev
> MR> _______________________________________________
> MR> Smartmontools-support mailing list
> MR>Sma...@li...
> MR>https://lists.sourceforge.net/lists/listinfo/smartmontools-support
>