Hi folks,
In my previous mail I said that having found really valuable
information, interesting tests on the Internet, and thanks to a lot of
attempts and tests I've done on my own with a dedicated machine, I would
have some time to add a lot of valuable information that isn't into the
official wiki about recovering from situation that could have been
considered as definitely lost.
It took me some more time, and a lot of validation and thorough tests
(as no error, bad advice, or unverified supposition can be allowed). I
also managed to understand (and show) how to very simply use overlays so
that write operation can't happen to any real RAID member.
The guide is about everything to know when recreating a new array over a
not working anymore or seriously screwed past existing one. After a
clear introduction about what it technically covers and implies.
It's also about recovery of accidentally overwritten member(s), and/or
understanding about rebuild process effects on the different members of
an array. In case of RAID6 for example, an erroneous full rewrite of
every member can surprisingly be recovered in some situations - and in
which case which rewritten disk has been rewritten with what - and is it
recoverable, and how.
https://raid.wiki.kernel.org/index.php/Irreversible_mdadm_failure_recovery
I did my best efforts for the English correctness; anyway feel free to
feedback or correct if you see some errors (be it English or technical
things, although it has been thoroughly tested).
Best regards,
Julien Robin
