Re: [PATCH] btrfs-progs: drop unique uuid test for btrfstune -M

[Anand Jain <anand.jain@oracle.com>]

David,

  We let the fsid to change but restrict if the user decides to undo.
  This bug is like a one-way trap.

  Any resolution on the issue below?

Thanks, Anand

 >> On 10/18/19 12:32 AM, David Sterba wrote:

>> I can't say I have a clear picture yet, can you please describe it in
>> some more desriptive way, like
>>
>> host1: create image1-uuid1
>>
>> host2: copy image1-uuid1 to image1-uuid2
>> host2: use image1-uuid2
>> host2: change image1-uuid2 back to uuid1     <-- I want this to work
>   From the bug as I received.
>      create btrfs root-image for the vm use.
>      copy root-image to root-image1
>      copy root-image to root-image2
>      start vm1 using root-image1
>      (when root-image1 has issues; shutdown vm1)
>      start vm2 using root-image2 with root-image1 accessible.
>      login to vm2
>        (change fsid so that root-image1 can be mounted)
>        btrsftune -m remote/root-image1
>        mount -o loop remote/root-image1 /mnt
>          analyze, collect logs, fix remote/root-image1
>        umount /mnt
>        (Revert the changed fsid so that vm2 can boot) <<<< Usecase wants 
> this to work
>        btrfstune -M $(btrfs in dump-super remote/root-image2 | \
>                       grep metadata_uuid | awk '{print $2}') \
>                       remote/root-image2
>      logout from vm2
>      start vm1 using root-image1