Academic exercise: trying to recover a corrupted pool

[Borja Marcos]

> On 16 Jan 2018, at 14:57, Mateusz Kwiatkowski <kwiat at panic.pl> wrote:
> 
> On 01/08/2018 03:25 PM, Borja Marcos wrote:
>> Hi,
>> ONLY AS AN ACADEMIC EXERCISE, WARNING :)
>> I have a broken ZFS pool and I’m wondering wether it should be readable. The pool was made with four
>> apparently troublesome OCZ SSD drives pulled from other systems. They are connected to a LSI2008 adapter.
>> The pool was created as a raidz2, so it’s supposed to survive the loss of two drives. It has lost two of them
>> and I am unable to import it.
>> I have lost no useful data, I was using it just for testing. Now it has become an interesting study subject though :)
>> Any ideas? I have tried to recover even doing the “radical thing” (zdb -Z -AAA -e -p /dev poolname). No success.
>> Again, I am just curious.
>> Thanks!
> 
> Few years ago I made similar exercise with pool broken by some ZFS on Linux bug. After many unsuccessful attempts to import the pool I came up with an idea to read data without importing the pool.
> Here's the script: https://gitlab.com/kwiat/zfs-recovery
> 
> The code is in no way effective nor beautiful, but I learned a lot about ZFS when I was working on it. :-)

Sorry for the very belated answer. It didn’t work either. I guess that a SSD failure can be especially catastrophic
even though I tried to read the drivers with dd and I got no errors.




Borja.