Is default (install) LVM2 thin pool metadata size appropriate?

[mailinglists]

Hi guys.

I'm just checking if the default (install) metadata size (68MB) for thin pool LVM is appropriately set. Below is what I have got during default install. I worry i might get metadata full and run into problems. However my worrying might be baseless, 'cause i'm still new to thin lvm.

# lvs -a
  LV              VG   Attr       LSize  Pool Origin Data%  Meta%  Move Log Cpy%Sync Convert
  data            pve  twi-aotz--  1.01t             1.45   1.13                            
  [data_tdata]    pve  Twi-ao----  1.01t                                                    
  [data_tmeta]    pve  ewi-ao---- 68.00m                                                    
  [lvol0_pmspare] pve  ewi------- 68.00m                                                    
  root            pve  -wi-ao---- 50.00g                                                    
  swap            pve  -wi-ao---- 16.00g

What do you guys have on your installs? Is it normal to have such small metadata lvs?

 

[seventh]

Hi,

No, it's not safe in my opinion, have a look at this thread regarding DRBD and LVM thin:
https://forum.proxmox.com/threads/drbd9-lvm-thin-provisioning.28584/#post-143856

You could increase the metadata by 1GB by running the following command:

lvextend --poolmetadatasize +1G pve/data

Also, don't forget to check the settings in /etc/lvm/lvm.conf and change the values for autoextend of thinpool.

 

[mailinglists]

Thanks for hints. Should I report a bug?
What do people at ProxMox have to say? Anyone?

 

[mailinglists]

Bug report submitted. https://bugzilla.proxmox.com/show_bug.cgi?id=1241

 

[greenedv]

Sorry to push this thread up but I just run into this really nasty problem that I had over 100G free data but 64m Metadata was full - WTF.
I highly suggest to "waste" Harddiskspace and set the metadata to max allowed size (according to manpage = 16G) right from the beginning.
Second use two Volumegroups. One for the "local" pve and a second for the new default thin pool "local-lvm". After increasing the size of the Metadata with the command "lvextend --poolmetadatasize +1G pve/data" you need to repair the LV to actually use the more space with "lvconvert --repair /pve/data". For that command you need to turn off the PVGroup in which the proxmox also resides so you have no way around to reboot the server because of some MB missing in F*ing Metadata.
It could potentially fry the whole lvm config because no data can be written at the "lvm-thin".

If anyone stumbles over this .. use swapoff pve to turn of the swap LVM and use lvreduce pve/swap to reduce the swap LV if you don't want to/can't plug in another harddrive.

Best Regards.

 

[mailinglists]

I guess it is up to Fabian and the guys to answer your hopes.
I care not anymore, because I mostly use ZFS by now.