[pve-devel] [PATCH storage] zfs: get the real mountpoint in path()

[Wolfgang Bumiller]

> On February 12, 2016 at 5:16 PM Dietmar Maurer <dietmar at proxmox.com> wrote:
> 
> 
> 
> 
> > On February 12, 2016 at 5:01 PM Wolfgang Bumiller <w.bumiller at proxmox.com>
> > wrote:
> > 
> > 
> > 
> > > On February 12, 2016 at 4:51 PM Dietmar Maurer <dietmar at proxmox.com> wrote:
> > > 
> > > 
> > > > > > Since mountpoints can be inherited and modified (and set to 'legacy'
> > > > > > and 'none') we need to deal with this. The current method of assuming
> > > > > > default paths doesn't always work.
> > > > > 
> > > > > Can't we make this a requirement instead (document and verify somehow)?
> > > > 
> > > > For example, we could verify that it is correctly mounted
> > > > in the storage activate method?
> > > 
> > > OK, I remember now that the current method does not work with zfs on /.
> > > To make that work we could add a 'mountpoint' property to the storage
> > > config? I just want to avoid that additional query inside path().
> > 
> > The mountpoint is a per-subvolume setting, so it doesn't matter where
> > exactly the pool's root is mounted.
> 
> But 'we' create the subvolumes, so we can set it to whatever we want...

Sure, and 'legacy' is probably the most sensible choice there IMHO.