On 16/01/19 03:35, Romulo Albuquerque wrote:
> Hi,
>
> I have a debian 8.2 (jessie) box, with a raid5 array running fine for
> more than 5 years.
> I tried to grew it from 6x2TB disks to 7x2TB, but the reshape got
> stuck due to failures on the new added disk.
> So, I bounced the system and tried to revert the reshape process.
> I stopped the array: mdadm --stop /dev/md127
>
> then try to revert the reshape process:
> mdadm --assemble /dev/md127 --run --force --update=revert-reshape
> /dev/sd[dcfgbe]1
>
> But it didn't work... I got a message asking for a backup-file that
> was lost after the reboot.
Okay, it's all looking good ...
I say that because the reshape position is 0, so it looks to me like the
reshape never actually started.
What version of mdadm are you running? ("mdadm --version"). What version
of kernel? ("uname -a").
This used to come up a lot - there are a few known bugs that would hang
a reshape like this - what's the betting you're on mdadm 3.3 or 3.4?
I strongly suspect that if you boot from an up-to-date recovery disk you
will be able to run your revert-reshape command no problem. There is
also a no-backup-file option that you might need - it won't do any
damage because the reshape never actually started.
Modern mdadm and kernel won't use a backup anyway, because they stash it
all in an area on the growing array.
Cheers,
Wol
