disk damage test

disk damage test

[Ivan Ivanov]

I made a simple test to compare how reiserfs and ext3 survive disk damage.
The test was on loop device. To simulate disk bad blocks I wrote "0" in
random positions in the file that was used for loop.

Reiserfs is very sensitive to disk damage. Several bad blocks on disk can
make the entire file system unusable and all data to be lost:
'reiserfsck --fix-fixable' tells that I must use --rebuild-tree but it
doesn't help.

On the contrary ext3 survives major disk damages. After running fsck
two times much of the broken files and dirs are placed in lost+found
and almost all data is recovered.

Regards
Ivan

Re: disk damage test

[Oleg Drokin]

Hello!

On Sat, Apr 06, 2002 at 10:36:32AM +0300, Ivan Ivanov wrote:

> Reiserfs is very sensitive to disk damage. Several bad blocks on disk can
> make the entire file system unusable and all data to be lost:
> 'reiserfsck --fix-fixable' tells that I must use --rebuild-tree but it
> doesn't help.
Have you tried latest reiserfsprogs?
We are interested in such damaged filesystems, that cannot be repaired with
latest reiserfsprogs.

Bye,
    Oleg

Re: disk damage test

[Ivan Ivanov]

On Sun, 7 Apr 2002, Oleg Drokin wrote:

> Hello!
>
> On Sat, Apr 06, 2002 at 10:36:32AM +0300, Ivan Ivanov wrote:
>
> > Reiserfs is very sensitive to disk damage. Several bad blocks on disk can
> > make the entire file system unusable and all data to be lost:
> > 'reiserfsck --fix-fixable' tells that I must use --rebuild-tree but it
> > doesn't help.
> Have you tried latest reiserfsprogs?
> We are interested in such damaged filesystems, that cannot be repaired with
> latest reiserfsprogs.
>
> Bye,
>     Oleg
>

I am using reiserfsprogs-3.x.1b

----
Ivan

Re: disk damage test

[Oleg Drokin]

Hello!

On Sun, Apr 07, 2002 at 12:01:13PM +0300, Ivan Ivanov wrote:
> > > Reiserfs is very sensitive to disk damage. Several bad blocks on disk can
> > > make the entire file system unusable and all data to be lost:
> > > 'reiserfsck --fix-fixable' tells that I must use --rebuild-tree but it
> > > doesn't help.
> > Have you tried latest reiserfsprogs?
> > We are interested in such damaged filesystems, that cannot be repaired with
> > latest reiserfsprogs.
> I am using reiserfsprogs-3.x.1b

Ok, and reiserfsck - failed to repair the FS? Did it segfaulted or did
it just finished the work, but FS was still damaged, or something else?

Thank you.

Bye,
    Oleg

Re: disk damage test

[Ivan Ivanov]

On Sun, 7 Apr 2002, Oleg Drokin wrote:

> Hello!
>
> On Sun, Apr 07, 2002 at 12:01:13PM +0300, Ivan Ivanov wrote:
> > > > Reiserfs is very sensitive to disk damage. Several bad blocks on disk can
> > > > make the entire file system unusable and all data to be lost:
> > > > 'reiserfsck --fix-fixable' tells that I must use --rebuild-tree but it
> > > > doesn't help.
> > > Have you tried latest reiserfsprogs?
> > > We are interested in such damaged filesystems, that cannot be repaired with
> > > latest reiserfsprogs.
> > I am using reiserfsprogs-3.x.1b
>
> Ok, and reiserfsck - failed to repair the FS? Did it segfaulted or did
> it just finished the work, but FS was still damaged, or something else?
>
> Thank you.
>
> Bye,
>     Oleg
>
>
It finished work ok, but the filesystem is damaged.

----
Ivan

Re: disk damage test

[Oleg Drokin]

Hello!

On Sun, Apr 07, 2002 at 12:17:36PM +0300, Ivan Ivanov wrote:
> > Ok, and reiserfsck - failed to repair the FS? Did it segfaulted or did
> > it just finished the work, but FS was still damaged, or something else?
> It finished work ok, but the filesystem is damaged.

How can you tell?
Did sybsequent run of reiserfsck --check finds any errors or did kernel
prints error messages, or any other criteria?

Thank you.

Bye,
    Oleg

Re: disk damage test

[Ivan Ivanov]

On Sun, 7 Apr 2002, Oleg Drokin wrote:

> Hello!
>
> On Sun, Apr 07, 2002 at 12:17:36PM +0300, Ivan Ivanov wrote:
> > > Ok, and reiserfsck - failed to repair the FS? Did it segfaulted or did
> > > it just finished the work, but FS was still damaged, or something else?
> > It finished work ok, but the filesystem is damaged.
>
> How can you tell?
> Did sybsequent run of reiserfsck --check finds any errors or did kernel
> prints error messages, or any other criteria?
>
> Thank you.
>
> Bye,
>     Oleg
>
>

So,  reiserfsck --check tels that I must run  reiserfsck --rebuild-tree.
I am running  reiserfsck --rebuild-tree two times and it tels everything
is OK.  reiserfsck --check tels OK too. But mount tels that directory not
found. What directory !! may be /. mount -t reiserfs tels the same.

I am trying to reproduce this again but I can't, I filled with "0" random
parts of the file and may be I can't hit the right blocks now.
When ReIsEr2 is zeroed reiserfsck --rebuild-sb and then reiserfsck
--rebuild-tree work fine - everything is recovered.

kernel is 2.4.18 with reiserfs quota patches.
reiserfsprogs 3.x.1b

-----
Regards
Ivan

Re: disk damage test

[Oleg Drokin]

Hello!

On Sun, Apr 07, 2002 at 02:21:10PM +0300, Ivan Ivanov wrote:

> So,  reiserfsck --check tels that I must run  reiserfsck --rebuild-tree.
> I am running  reiserfsck --rebuild-tree two times and it tels everything
> is OK.  reiserfsck --check tels OK too. But mount tels that directory not
> found. What directory !! may be /. mount -t reiserfs tels the same.

Hm. Have you saved a file from that particular experiment?
Anything was left in logs?

Bye,
    Oleg

Re: disk damage test

[Ivan Ivanov]

On Sun, 7 Apr 2002, Oleg Drokin wrote:

> Hello!
>
> On Sun, Apr 07, 2002 at 02:21:10PM +0300, Ivan Ivanov wrote:
>
> > So,  reiserfsck --check tels that I must run  reiserfsck --rebuild-tree.
> > I am running  reiserfsck --rebuild-tree two times and it tels everything
> > is OK.  reiserfsck --check tels OK too. But mount tels that directory not
> > found. What directory !! may be /. mount -t reiserfs tels the same.
>
> Hm. Have you saved a file from that particular experiment?
> Anything was left in logs?
>
> Bye,
>     Oleg
>
>

Sorry, I formated the same file ext3 for comparison - it was stupid.
Nothig unusual in logs:

Apr  6 09:31:22 idi kernel: VFS: Can't find ext3 filesystem on dev
loop(7,0).
Apr  6 09:31:22 idi kernel: VFS: Can't find ext3 filesystem on dev
loop(7,0).
Apr  6 09:31:22 idi kernel: reiserfs: checking transaction log (device
07:00) ...
Apr  6 09:31:22 idi kernel: Using r5 hash to sort names
Apr  6 09:31:22 idi kernel: ReiserFS version 3.6.25
Apr  6 09:33:57 idi kernel: VFS: Can't find ext3 filesystem on dev
loop(7,0).
Apr  6 09:33:57 idi kernel: VFS: Can't find ext3 filesystem on dev
loop(7,0).
Apr  6 09:33:57 idi kernel: reiserfs: checking transaction log (device
07:00) ...
Apr  6 09:33:57 idi kernel: Using r5 hash to sort names
Apr  6 09:33:57 idi kernel: ReiserFS version 3.6.25

------
Ivan