I recently discovered the cache writes problems. As it seems to be present since almost 5 or 6 monts i don't expect a resolution in the next few days so i'm looking for a workaround. I don't want to feed my server with new ssd every few months waiting a resolution. 

For now this problem killed my 2 ssd cache in a year (probably less as i don't know when the problem started).

It seems a cache with a single ssd formatted with xfs could be a workaround.

I have only 3 shares on the cache: domains, appdata and system

How could i do to move from a btrfs raid1 cache pool to a single xfs (i think xfs pool is not possible) ?

Can i copy the 3 shares to a drive (unassigned or array) then destroy the pool and recreate the cache then copy back the share ?

if so do i have to delete libvirt.img and docker.img and let the system rebuild it ?

 

edit : path for docker.im is /mnt/user/system/docker/docker.img 

and appdata is /mnt/user/appdata

should i replace "user" with "cache" ?

Edited May 29, 20206 yr by caplam

The wiki has a section on replacing the cache drive, in a nutshell, you disable the docker and vm service so the tabs are no longer available in the GUI, set all your shares to cache: yes and run the mover. In theory, that should completely empty the cache pool. Then stop the array, unassign the cache drives, and set the number of visible cache slots while the array is stopped to 1 (ONE). That will allow you to assign a single cache drive and select XFS as the desired format. Start the array, the cache drive should say unmountable and offer an option to format. Make sure ONLY the cache drive is unmountable, and format it. Set the domains, appdata and system shares to cache prefer, and run the mover. After mover completes, reenable the VM and Docker services.

thanks you for your answer.
right now i have downgraded to 6.7.2 but it doesn’t solve all my problems.