Re: [PATCH] mdadm: fix reshape from RAID5 to RAID6 with backup file

[Nigel Croxon <ncroxon@redhat.com>]

Jes,

Can we have a revert until the issue has been resolved for all?

Thanks Oleskandr for the diagnose.

-Nigel

> On Mar 22, 2021, at 12:21 PM, Oleksandr Shchirskyi <oleksandr.shchirskyi@linux.intel.com> wrote:
> 
> Hello Nigel,
> 
> I have collected more info regarding this issue.
> I can confirm what Mariusz said, it's a regression caused by patch 4ae96c802203
> The reason for failure during the reshape, is that in this patch sync_max value is set to max, but the function wait_for_reshape_imsm, used in some reshape scenarios, relies on this parameter, and doesn't expect, that value can be max. This leads to reshaping fail.
> Here's an example of a debug log from this method, when the issue is hit:
> 
> mdadm: wait_for_reshape_imsm: wrong next position to set 4096 (2048)
> mdadm: imsm_manage_reshape: wait_for_reshape_imsm returned error!
> 
> With this patch reverted, the issue is not observed. See my logs below:
> 
> # mdadm -CR imsm0 -e imsm -n4 /dev/nvme[0-3]n1 && mdadm -CR volume -l0 --chunk 64 --size=10G --raid-devices=1 /dev/nvme0n1 --force
> # mdadm -D /dev/md/volume                                                               /dev/md/volume:
>         Container : /dev/md/imsm0, member 0
>        Raid Level : raid0
>        Array Size : 10485760 (10.00 GiB 10.74 GB)
>      Raid Devices : 1
>     Total Devices : 1
>             State : clean
> ...
> # mdadm -G /dev/md/imsm0 -n2
> # mdadm -D /dev/md/volume
> /dev/md/volume:
>         Container : /dev/md/imsm0, member 0
>        Raid Level : raid4
>        Array Size : 10485760 (10.00 GiB 10.74 GB)
>     Used Dev Size : 10485760 (10.00 GiB 10.74 GB)
>      Raid Devices : 3
>     Total Devices : 2
>             State : clean, degraded
> ...
> # git revert 4ae96c802203ec3cfbb089240c56d61f7f4661b3
> Auto-merging Grow.c
> [master 1166854] Revert "mdadm: fix reshape from RAID5 to RAID6 with backup file"
> 1 file changed, 2 insertions(+), 5 deletions(-)
> # mdadm -Ss; wipefs -a /dev/nvme[0-3]n1
> # make clean; make; make install-systemd; make install
> # mdadm -CR imsm0 -e imsm -n4 /dev/nvme[0-3]n1 && mdadm -CR volume -l0 --chunk 64 --size=10G --raid-devices=1 /dev/nvme0n1 --force
> # mdadm -G /dev/md/imsm0 -n2
> # mdadm -D /dev/md/volume
> /dev/md/volume:
>         Container : /dev/md/imsm0, member 0
>        Raid Level : raid0
>        Array Size : 20971520 (20.00 GiB 21.47 GB)
>      Raid Devices : 2
>     Total Devices : 2
> 
>             State : clean
> ...
> #
> 
> On 3/16/2021 4:59 PM, Nigel Croxon wrote:
>> ----- Original Message -----
>> From: "Mariusz Tkaczyk" <mariusz.tkaczyk@linux.intel.com>
>> To: "Jes Sorensen" <jes@trained-monkey.org>, "Nigel Croxon" <ncroxon@redhat=
>> .com>, linux-raid@vger.kernel.org, xni@redhat.com
>> Sent: Tuesday, March 16, 2021 10:54:22 AM
>> Subject: Re: [PATCH] mdadm: fix reshape from RAID5 to RAID6 with backup fil=
>> e
>> Hello Nigel,
>> Blame told us, that yours patch introduce regression in following
>> scenario:
>> #mdadm -CR imsm0 -e imsm -n4 /dev/nvme[0125]n1
>> #mdadm -CR volume -l0 --chunk 64 --raid-devices=3D1 /dev/nvme0n1 --force
>> #mdadm -G /dev/md/imsm0 -n2
>> At the end of reshape, level doesn't back to RAID0.
>> Could you look into it?
>> Let me know, if you need support.
>> Thanks,
>> Mariusz
>> I=E2=80=99m trying your situation without my patch (its reverted) and I=E2=
>> =80=99m not seeing success.
>> See the dmesg log.
>> [root@fedora33 mdadmupstream]# mdadm -CR volume -l0 --chunk 64 --raid-devic=
>> es=3D1 /dev/nvme0n1 --force
>> mdadm: /dev/nvme0n1 appears to be part of a raid array:
>>       level=3Dcontainer devices=3D0 ctime=3DWed Dec 31 19:00:00 1969
>> mdadm: Creating array inside imsm container md127
>> mdadm: array /dev/md/volume started.
>> [root@fedora33 mdadmupstream]# cat /proc/mdstat=20
>> Personalities : [raid6] [raid5] [raid4] [raid0]=20
>> md126 : active raid0 nvme0n1[0]
>>      500102144 blocks super external:/md127/0 64k chunks
>> md127 : inactive nvme3n1[3](S) nvme2n1[2](S) nvme1n1[1](S) nvme0n1[0](S)
>>      4420 blocks super external:imsm
>> unused devices: <none>
>> [root@fedora33 mdadmupstream]# mdadm -G /dev/md/imsm0 -n2
>> [root@fedora33 mdadmupstream]# cat /proc/mdstat=20
>> Personalities : [raid6] [raid5] [raid4] [raid0]=20
>> md126 : active raid4 nvme3n1[2] nvme0n1[0]
>>      500102144 blocks super external:-md127/0 level 4, 64k chunk, algorithm=
>>  5 [2/1] [U_]
>> md127 : inactive nvme3n1[3](S) nvme2n1[2](S) nvme1n1[1](S) nvme0n1[0](S)
>>      4420 blocks super external:imsm
>> unused devices: <none>
>> dmesg says:
>> [Mar16 11:46] md/raid:md126: device nvme0n1 operational as raid disk 0
>> [  +0.011147] md/raid:md126: raid level 4 active with 1 out of 2 devices, a=
>> lgorithm 5
>> [  +0.044605] md/raid0:md126: raid5 must have missing parity disk!
>> [  +0.000002] md: md126: raid0 would not accept array
>> -Nigel
> 
> -- 
> Regards,
> Oleksandr Shchirskyi
>