[linux-lvm] lvconvert -m 0 /dev/varlin/distfiles /dev/md0 => zero filled logical volume

[Raimonds Cicans]

On 17.03.2016 09:32, Raimonds Cicans wrote:
> On 16.03.2016 23:36, Raimonds Cicans wrote:
>> I am trying to move LVM setup from old RAID md0 to new RAID md1
>>
>> I did "lvconvert -m 1 /dev/path/to/logical_volume /dev/md1" on bunch of
>> logical volumes.
>> Everything went without problems.
>>
>> Then I tried to remove md0 from one of logical volumes by running
>> command:
>> lvconvert -m 0 /dev/varlin/distfiles /dev/md0
>>
>> I tried to mount this logical volume, but failed.
>>
>> Then I checked first  megabyte of logical volume it was zero filled.
>>
>> Partial output of "lvs -o+devices":
>>    LV              VG     Attr       LSize   Pool Origin Data%  Meta%
>> Move Log Cpy%Sync Convert Devices
>>    backup          varlin -wi-ao---- 100,00g
>> /dev/md0(6304)
>>    backup          varlin -wi-ao---- 100,00g
>> /dev/md0(8544)
>>    data            varlin -wi-ao----  20,00g
>> /dev/md0(28384)
>>    distfiles       varlin -wi-a-----  20,00g
>> /dev/md1(2738)
>>    ftp             varlin rwi-aor---  10,00g         100,00
>> ftp_rimage_0(0),ftp_rimage_1(0)
>>    ftpdown         varlin rwi-aor---  10,00g         100,00
>> ftpdown_rimage_0(0),ftpdown_rimage_1(0)
>>
>>
>> lvm2: 2.02.116
>> kernel: 4.4.2
>>
>>
>
> I made following test:
> I unmounted /opt and dumped to files following volumes:
> /dev/mapper/varlin-opt
> /dev/mapper/varlin-opt_rimage_0
> /dev/mapper/varlin-opt_rimage_1
>
> and then diff-ed them.
>
> /dev/mapper/varlin-opt_rimage_0 was not equal with /dev/mapper/varlin-opt.
> Why??? I thought 100 in Cpy%Sync column of lvs output mean volume is
> fully synchronized.
>
> With "lvconvert -m0" I removed /dev/mapper/varlin-opt_rimage_0 from
> /dev/mapper/varlin-opt.
>
> I did some test of volume - everything was normal.
>
> My plan for now:
> for each mirrored volume:
> 1) do volume content backup
> 2) umount volume
> 3) run md5sum for volume itself and all mirror sub-volumes
> 4) with "lvconvert -m0" remove sub-volume which is not equal to volume
> 5) do some volume tests
> 6) if something went wrong - restore volume content from backup
> 7) if on step 4 was removed new RAID md1, then make pvmove to md1
>
>

Plan failed, because one of sata controllers failed.
I moved all disks to other computer and booted from SystemRescueCD.
All volumes was synced. I checked by md5sum-ing volumes itself an both 
their legs.
Then I without problems moved all remaining volumes to new RAID.

So, case is closed.

Best regards,
Raimonds Cicans