Re: Fwd: Help with failed RAID-5 -> 6 migration

[Phil Turmel <philip@turmel.org>]

On 06/13/2013 09:58 AM, Keith Phillips wrote:
>> You won't be able to assemble and run your array without a backup file.
>>  You said you lost your original, so you will have to use a blank one
>> and tell mdadm to ignore the invalid file.
> 
> Ah, didn't realise this was an option. After a brief googling it seems
> my version of mdadm pre-dated the "--invalid-backup" option.
> 
> Cloned the git repo and built a newer version, and re-assembled with
> an empty "--backup-file" and the "--invalid-backup" option. Now it's
> chugging along happily again - at %20 and counting now, no errors in
> sight!

Good to hear. :-)

> Will do an ext4 fsck once it's finished the grow. Are there any tips
> for determining what data I trashed by losing the backup-file? Or is
> it just a case of trying to access stuff and seeing what's broken?

You have the reshape position where the process stopped in the original
mdadm -E reports.  Use that to query for inodes that contain those
sectors, then look up those inodes.

A quick google came up with:
http://smartmontools.sourceforge.net/badblockhowto.html

You'll have to reinterpret that to use the sector offsets in your array
rather than sector offset from smartctl.

> Thanks so much for the help, Phil :)

You're welcome.

Phil