Re: [Hampshire] ext4 strangeness

On 14 January 2011 09:42, Chris Dennis <cgdennis@???> wrote:
> On 13/01/11 15:30, Hugo Mills wrote:
>>
>> On Thu, Jan 13, 2011 at 03:23:30PM +0000, Chris Dennis wrote:
>>>
>>> Hello folks
>>>
>>> I've come across a problem on a friend's computer.
>>>
>>> The computer is fairly old (see below), running Ubuntu 10.10, with /
>>> on a 40GB IDE drive, and /home on a 120GB Maxtor SATA drive.
>>>
>>> It's /home that's the problem.  It's on a single partition
>>> /dev/sda1, formatted as ext4.  I've run the Maxtor diagnostics on
>>> the drive, and no problems were reported.
>>>
>>> Ubuntu wants to check the filing system on every boot.  Running
>>> 'fsck -f' manually (from Knoppix 6.2) gives:
>>>
>>>   Block bitmap differences:<some number ranges, different each time>
>>>
>>> Sometimes it also offers to fix:
>>>
>>>   Free blocks count wrong for group #<numbers that change each time>
>>>
>>> I choose to fix those problems, but they're still there next time.
>>>
>>> I've also run 'fsck -fc /dev/sda1' -- it didn't report any bad
>>> blocks, and the other were still there.
>>
>>    This smacks to me of a disk that's constantly remapping duff
>> sectors into spares, and finding more duff sectors. Replace the
>> hardware ASAP.
>>
>> [...]
>>>
>>> ran fsck again, and the errors have returned!
>>>
>>> Why would copying data into a filing system cause errors like that?
>>
>>    Bad hardware.
>>
>>    Hugo.
>
> Thanks for your reply Hugo.
>
> I'd have thought the same if the initial disk scan hadn't been OK.
>
> So I did some more checks:
>
> * older Maxtor diagnostics -- no errors.
> * latest Seatools diagnostics -- no errors.
> * SMART (via HDAT2 CD) -- only one error in lifetime of disk,
>  no reallocated sectors, no pending or uncorrectable errors.
> * SMART extended offline test -- no errors.
>
> Then I ran
>
>  mkfs.ext3 -c -c /dev/sda1
>
> with full read-write checks for bad blocks -- it found none, and SMART still
> reports no reallocated sectors.
>
> fsck on the empty filing system was OK.  Copied all the data back as before
> (about 30GB), and fsck is still happy.
>
> Note that I've now made the partition ext3 instead of ext4.  Would that make
> a difference?  I did notice that when running fsck on the ext4 partition,
> the first 4 passes were almost instantaneous; only pass 5 'Checking group
> summary information' took a bit longer, and detected the errors.  With ext3,
> pass 1 takes much longer (i.e. 10 seconds instead of less than 1).
>
> Anyway, the data is all backed up, so I'll see how it goes for a while,
> unless anyone can think of any other tests I can run on the drive.
>
> cheers
>
> Chris

I had problems with a Maxtor HD recently.
smart reported no problems at all, but when trying to read all the
sectors on the HD with ddrescue it has quite a few non-readable
sectors.
I concluded that the SMART functionality on Maxtor 2.5 HDs is worthless.
At least with Seagate HDs, the SMART functionality actually reports problems.

I suggest you do a ddrescue or someother backup method on the failing
HD you have ASAP, and then do repairs to a copy of the HD image that
ddrescue creates.
Remember that every time the drive heads pass over the disk surface it
is likely to cause more damage, so every repair or read attempt you
make is likely to result in more sectors going bad.