On 02/23/2013 05:19 AM, Stone wrote:

ok befor i doing something worg i will ask you more questions :)
what do you mean with "take a backup of the array" and how it works?

First priority is to recover your data in the encrypted volume.  You
can't fix the partition misalignment on sdb and sde without destroying
their content.  So *after* we get your data back, you need to save it
somewhere else when you repartition.

sorry i dont know what you mean
after this i create on all four devices the partiontable new with parted
and the starting sector must be 2048.

Not yet.  We have to save your data first.  Start sector 34 is bad for
performance.  But that is where your data is, so you have to use it
until you get you data back, and can put the data on some other storage
system.

should i make a backup copy of all devices partiontables? if yes how?

for x in /dev/sd[bce] ; do parted $x unit s print ; done

The partition structure on /dev/sdc is causing the array to be too short
for the filesystem.  There are two possibilities:

1) The partition doesn't go far enough to the end of the disk,

2) The partition starts too far into the disk (move start sector to 34
like sdb and sde).

We can see that the partition on sdc does start further into the disk
than sdb, so that is suspicious.  But you don't remember repartitioning
sdc, so changing it might misalign your existing data.

I don't know if you can fix #1--I need to see the parted report with
"unit s".  If there's room at the end, you try that first and see the
results of "fsck -n".  (The size of /dev/sdc1 needs be at least
3907025920 sectors.)

If that still has many errors, you try fixing #2.

Phil

ps.  I hope this odyssey has emphasized to all lurkers how terrible it
can be to use "mdadm --create" without careful, thorough preparation.