On 04/23/2017 05:42 PM, Andreas Klauer wrote:

On Sun, Apr 23, 2017 at 05:11:08PM +0200, Patrik Dahlström wrote:

quoted

* I try to find 0x53EF (ext4 magic) within the first 128M of
/dev/sd[abcde]. Not after? This will be an indication of my "New
Offset". I need to adjust the offset a bit since the ext4 magic is
located at 0x438 offset.

Yes, that should be about it.
 

quoted

* I create a 5 and a 6 drive raid set and try to find an offset where
they both carry the same raw data. With some overlays, I should be able
to create both these raids at the same time, correct?

Yes, two sets of overlays.

So overlay A is your 5 disk raid5, overlay B is your 6 disk raid5, 
and then you'll just have to take a stab at it with hexdump.

So kind of like,

hexdump --skip $((17*1024*1024*1024)) --length 4096 /dev/md42
hexdump --skip $((17*1024*1024*1024)) --length 4096 /dev/md43

If that produces the same random-looking data then your reshape 
might have progressed 17G-ish and you could try to use that as 
a starting point for a linear device mapping that uses the 
first 17G of the 6 disk raid and everything else from the 5disk.

The further the grow processed the larger a zone of overlap there 
should be since more data ends up on the additional drive so 
the original representation of the same data isn't yet written 
into on the old drives.

Does that make sense?

5 disks: a b c d e   : f g h i j   : k l m n o   : p q r s t : ...
6 disks: a b c d e f : g h i j k l : m n o p q r : s t       : ...

If it stopped at "t", both have "r s t"... and that "r s t" 
would be what you have to find. You have to be wary of false 
matches though (such as zeroes or other common patterns of data 
you might find anywhere).

And there's one more thing, you mentioned a disk had a bad sata cable.
If that disk got kicked and reshape went on for a while afterwards, 
you should take that disk out of consideration. (Specify as "missing" 
when creating the arrays.) It will have outdated/unreshaped data on 
it that would be hard to incorporate into your recovery attempt...

The bad SATA cable has been there for a while and reported spurious I/O
errors on the old 5 disk raid. It really started to produce I/O errors
when I placed the new drive in that slot.

Is it OK to specify the new drive as missing? No special information
that should be located on that drive?

I wrote a small program to help me locate the correct offset:
https://gist.github.com/Risca/3eda5e7aba3dc6b72d61e79eaf7cc147

I am on the right track with that program?

Regards
Andreas Klauer