Btrfs and LBA errors

Btrfs and LBA errors

[constantine]

Hi everybody!

Btrfs filesystem could not be mounted because /dev/sdc1 had unreadable sectors.
It is/was  a single filesystem (not raid1 or raid0) over /dev/sda1 and
/dev/sdc1.
I wrote the unreadable sectors with hdparm but filesystem still cannot
be mounted (sectors were too early perhaps).

What should be my next step  for recovering the btrfs filesystem?


#### Analytically:

After
# smartctl -t long /dev/sdc
I got
"Extended offline   Completed: read failure   90%   19662   2176"

So LBA_of_first_error: 2176

With
# hdparm --read-sector
I found out that 2176-2183 LBA sectors were unreadable.

I wrote them with
# hdparm --write-sector
and now smartctl does not report any error.

However when I run
# btrfs fi show
I get:

Label: 'partition'  uuid: c1eb1aaf-665a-4337-9d04-3c3921aa67e0
        Total devices 2 FS bytes used 1.53TiB
        devid    2 size 1.37TiB used 195.00GiB path /dev/sda1
        *** Some devices missing


#### Complementary Information
# uname -a
Linux hostname 3.15.5-031505-generic #201407091543 SMP Wed Jul 9
19:44:36 UTC 2014 x86_64 x86_64 x86_64 GNU/Linux

#   btrfs --version
Btrfs v3.12

#   btrfs fi show
Label: 'partition'  uuid: c1eb1aaf-665a-4337-9d04-3c3921aa67e0
        Total devices 2 FS bytes used 1.53TiB
        devid    2 size 1.37TiB used 195.00GiB path /dev/sda1
        *** Some devices missing

Btrfs v3.12

# btrfs fi df /mnt/partition
ERROR: couldn't get space info - Inappropriate ioctl for device
ERROR: get_df failed Inappropriate ioctl for device

Re: Btrfs and LBA errors

[Russell Coker]

On Fri, 11 Jul 2014 13:42:40 constantine wrote:
> Btrfs filesystem could not be mounted because /dev/sdc1 had unreadable
> sectors. It is/was  a single filesystem (not raid1 or raid0) over /dev/sda1
> and /dev/sdc1.

What does "file -s /dev/sda1 /dev/sdc1" report?

-- 
My Main Blog         http://etbe.coker.com.au/
My Documents Blog    http://doc.coker.com.au/

Re: Btrfs and LBA errors

[constantine]

Thank you very much for your response:

# file -s /dev/sda1 /dev/sdc1
/dev/sda1: BTRFS Filesystem label "partition", sectorsize 4096,
nodesize 4096, leafsize 4096,
UUID=c1eb1aaf-665a-4337-9d04-3c3921aa67e0, 1683870334976/3010310701056
bytes used, 2 devices
/dev/sdc1: data

Re: Btrfs and LBA errors

[Russell Coker]

On Fri, 11 Jul 2014 21:29:07 constantine wrote:
> Thank you very much for your response:
> 
> # file -s /dev/sda1 /dev/sdc1
> /dev/sda1: BTRFS Filesystem label "partition", sectorsize 4096,
> nodesize 4096, leafsize 4096,
> UUID=c1eb1aaf-665a-4337-9d04-3c3921aa67e0, 1683870334976/3010310701056
> bytes used, 2 devices
> /dev/sdc1: data

Looks like the primary superblock on /dev/sdc1 is corrupt.

http://www.funtoo.org/BTRFS_Fun#btrfs-select-super_.2F_btrfs-zero-log

The above URL documents how to use the other superblocks.  It notes that you 
only get one chance - if you do this wrong you will dramatically reduce the 
probability of getting your data back.  Do not do this on the master copy of 
your data!

Buy 2 disks of equal or greater capacity, copy the block devices, and then 
working with the copy.  sdc needs to be replaced anyway no matter what you do 
and getting a replacement for sda would be a good strategy anyway.  Buy 2*4TB 
disks and you can make a new RAID-1 array that has more capacity than the non-
RAID filesystem you currently have.

-- 
My Main Blog         http://etbe.coker.com.au/
My Documents Blog    http://doc.coker.com.au/

Re: Btrfs and LBA errors

[constantine]

Thank you very much for your advice. It worked!

I verified that the superblocks 1 and 2 had similar information with
btrfs-show-super -i 1 /dev/sdc1 (and -i 2) and then with crossed
fingers:
btrfs-select-super  -s 2 /dev/sdc1
which restored my btrfs filesystem.

Then I runned scrub. For future reference, scrub was "stuck":
"
# btrfs scrub status partition/
scrub status for c1eb1aaf-665a-4337-9d04-3c3921aa67e0
scrub started at Thu Jul 10 21:30:41 2014, running for 2743 seconds
   total bytes scrubbed: 193.47GiB with 0 errors
"
and could not cancel it:
"
btrfs scrub cancel partition/
ERROR: scrub cancel failed on Downloads/: not running
"
After deleting /var/lib/btrfs/scrub.status.c1eb1aaf-665a-4337-9d04-3c3921aa67e0
it was completed successfully.

Many thanks!

Re: Btrfs and LBA errors

[Russell Coker]

On Tue, 15 Jul 2014 11:42:05 constantine wrote:
> Thank you very much for your advice. It worked!

Great!

> I verified that the superblocks 1 and 2 had similar information with
> btrfs-show-super -i 1 /dev/sdc1 (and -i 2) and then with crossed
> fingers:
> btrfs-select-super  -s 2 /dev/sdc1
> which restored my btrfs filesystem.
> 
> Then I runned scrub. For future reference, scrub was "stuck":
> "
> # btrfs scrub status partition/
> scrub status for c1eb1aaf-665a-4337-9d04-3c3921aa67e0
> scrub started at Thu Jul 10 21:30:41 2014, running for 2743 seconds
>    total bytes scrubbed: 193.47GiB with 0 errors
> "
> and could not cancel it:
> "
> btrfs scrub cancel partition/
> ERROR: scrub cancel failed on Downloads/: not running
> "
> After deleting
> /var/lib/btrfs/scrub.status.c1eb1aaf-665a-4337-9d04-3c3921aa67e0 it was
> completed successfully.

That's a known bug.

Now I hope you have made a good backup, the next problem you encounter may not 
be as easy to solve.

-- 
My Main Blog         http://etbe.coker.com.au/
My Documents Blog    http://doc.coker.com.au/