Re: Converting thin stripe volume to linear

[Zdenek Kabelac <zdenek.kabelac@gmail.com>]

Dne 14. 03. 25 v 2:30 Patrick Hemmer napsal(a):
> On Thu, Mar 13, 2025, at 20:27, Zdenek Kabelac wrote:
>> Dne 14. 03. 25 v 1:17 Patrick Hemmer napsal(a):

> I'm describing exactly what I want. To convert a thin striped volume to linear.

Hi

Unfortunately you were using misleading terms but we are almost getting to the 
right point  -  as you were like talking  about volume named  'thin_tdata' 
which was striped LV and now is raid5 LV - but it's not 'thin' volume.

You want to preserve thin-pool functionality, you just want to relocate _tdata 
to the new drive and drop 'striped' functionality out of it.

ATM your thin_tdata volume is already raid5_n and conversion to 'raid1'  type 
is not supported ATM as lvm2 rejects this conversion for stacked LV.

So for quickest results I'd recommend these steps:
(Assuming you are 'brave enough to make them)

- Extend your 'ssd' VG with your new drive

- Create new LV (in the same VG) on new storage with the size matching _tdata 
volume size.
   # lvs -a --units b
   # lvcreate  -Lsamesizesize --name new_data_lv ssd  /dev/newpv

- Do an 'component' activation of _tdata - for this whole thin-pool and all 
it's thin volumes must be inactivate.

   # lvchange -ay ssd/thin_tdata
  This will activate volume in read-only mode.

- Then do a  'dd' copy of this _tdata volume to new destination.
   # dd   /dev/ssd/thin_tdata  ->  /dev/ssd/new_data_lv

Now 'tricky' part:

- Take a 'vgcfgbackup -f backup ssd' of current VG

- After all data have been copied make sure LVs are INACTIVE
   lvchange -an ssd/....  or simply  vgchange -an ssd

- Edit 'ascii' metadata format in your text editor - i.e.: 'vim  backup'
   look for section with LV name "thin_tdata"
   Rename this LV to some other unused name i.e. "thin_Xdata".

   Than look for the LV section with newly created LV and rename
   this LV to the name:  "thin_tdata"

   write lvm2 metadata and exit editor

- Now restore lvm2 metadata with 'vgcfgrestore --force -f backup ssd'

- check with  'lvs -ao+seg_pe_ranges'  that are all desired changes are in effect

- you now should be able to activate pool and use data from new location.

- lvremove now unused  thin_Xdata volume - evetually you should be able to 
vgremove unused  PVs from 'ssd' VG.

- You may apply similar workflow to _tmeta  - although there should work pvmove.

Regards

Zdenek

PS: Let me know if find some step unclear - it's better to ask instead of 
causing data lose.

PPS: - yeah we will need to figure out how to enable this missing workflow 
within the lvm2 - so the user does not need to do above step in such manual way.