AW: AW: ZFS: Corrupted pool metadata after adding vdev to a pool - no opportunity to rescue data from healthy vdevs? Remove a vdev? Rewrite metadata?

[Volodymyr Kostyrko]

14.09.2012 08:29, Thomas Göllner (Newsletter) wrote:
> Thanks for the code. It helped a little bit to understand some details.
>
> I tried the "-e" Option of "zdb" and I listed the uberblocks of the discs with "zdb -l -u <dev>".
> As I can see, the there is a "create_txg: 56482". With the uberblocks I was able to find the corresponding date/time which is the time I added the new vdev.
>
> But every command and option combo I'm trying gets me just two results:
> "Input/output error" or
> "Device not configured"
> So I can't read out data via "-bbb" or so...
>
> Perhaps there was more destroyed during the process of creating and adding the new vdev than I thought first. It seems to me, that the NAS4free GUI mixed up the discs during formatting or something like that. After adding the new vdev I hat 16TB of space which was way too much and in the info I could see the discs of the old vdev twice... But the NAS was still playing out my data. After the restart the pool faulted.
>
> So my last idea is the "-C" option of the "zdb" command. Perhaps I can give a new config which it has to use. But where do I find the config file or what is it supposed to look like?

Try reading Max Bruning's blog at http://mbruning.blogspot.com/, there's 
a good data walk on raidz and raidz on-disk format.

> I also think there is no way to write new or edit the lables of the discs?

This idea is called Block Pointer Rewrite and is not implemented yet. I 
have found no code to do that.

-- 
Sphinx of black quartz judge my vow.