Mdadm 3.3 reshape issue

Mdadm 3.3 reshape issue

[Justin Maggard]

Today I was trying to reshape a RAID 5, going from 3 RAID disks to 4,
using the new mdadm 3.3 release.  I had already added the new
partition to the array, which was marked as spare.  Then I ran
`/sbin/mdadm -G -n4 /dev/md127` to start the reshape.  Unfortunately,
it would always return "Failed to initiate reshape".

After a little digging, I found that this was because I'm running a
SUSE SLE11-SP3 kernel, which has partially-implemented support for
changing data_offset.  In raid5.c:raid5_start_reshape(), I found this:

                /* Don't support changing data_offset yet */
                if (rdev->new_data_offset != rdev->data_offset)
                        return -EINVAL;

So, I understand where the "Failed to initiate reshape" error is
coming from.  But is there perhaps a better way to handle this
situation?  I also found that new_data_offset does not get restored
after hitting this error, so further attempts to reshape using a
backup-file continued to fail.  Would it make sense to at least set
new_data_offset back to what it was when we fail to initiate the
reshape?

Thanks,
-Justin

Re: Mdadm 3.3 reshape issue

[Roman Mamedov]

[-- Attachment #1: Type: text/plain, Size: 403 bytes --]

On Tue, 17 Sep 2013 11:56:06 -0700
Justin Maggard <jmaggard10@gmail.com> wrote:

> I'm running a SUSE SLE11-SP3 kernel

Or you could have mentioned which actual kernel version you are running...
Anyways to save people some googling and sifting through useless release
notes, this seems to be based on Linux 3.0.76.
Personally I'd suggest upgrading to 3.10 or 3.11.

-- 
With respect,
Roman

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Fwd: Mdadm 3.3 reshape issue

[Justin Maggard]

On Tue, Sep 17, 2013 at 12:12 PM, Roman Mamedov <rm@romanrm.net> wrote:
>
> On Tue, 17 Sep 2013 11:56:06 -0700
> Justin Maggard <jmaggard10@gmail.com> wrote:
>
> > I'm running a SUSE SLE11-SP3 kernel
>
> Or you could have mentioned which actual kernel version you are running...
> Anyways to save people some googling and sifting through useless release
> notes, this seems to be based on Linux 3.0.76.
> Personally I'd suggest upgrading to 3.10 or 3.11.

 The version I'm running is based on 3.0.82, with lots of patches from
SUSE.  The 3.0.82 source from kernel.org doesn't have any data-offset
changing capabilities at all, and thus does not have this problem.  I
am well aware that this issue is probably specific to the SUSE SLE11
kernel, and I am positive that 3.10 or 3.11 would work fine.

-Justin