[PVE-User] BTRFS...

[Wolfgang Bumiller]

On Tue, Feb 02, 2016 at 11:40:47AM -0200, Gilberto Nunes wrote:
> Well
> 
> I almost do it, 'cause one of feature I appreciate in btrfs is the hability
> to increase or decrease disk size.
> I know LVM can do it as well, but LVM are always on top whatever X
> Filesystem.

>From our perspective it would bring a bunch more useful features to the
table like snapshots, but for this particular task it's not actually any
more convenient than running resize2fs on an image or LVM, so we only
see the "not-yet mature enough" side of the story. At least with our
current kernel version.
And we wouldn't get rid of the additional layers since it would still
have to be used on top of our storage backend management.

As for convenience: eg. for the `pct resize` command if the container is
offline we just run our backend's resize function (which is still
required with btrfs to eg increase the image file or zvol or lvm
portion) and then run resize2fs on it. With btrfs we'd also have to
mount it for such operations because they only work on mountpoint-paths.
Although for a running container this isn't actually a problem as you
*can* pass a path like /proc/${ContainerPID}/root/ which I find almost
surprising since some other commands like 'mount' often do something
unexpectedly *bad* with paths like that.

> So do you have more layers... Btrfs is direct into the device. Or am I
> wrong?!
> 
> 
> 2016-02-02 11:30 GMT-02:00 Paul Gray <gray at cs.uni.edu>:
> 
> > On 02/02/2016 07:11 AM, Gilberto Nunes wrote:
> > > And more important: any one here already use or test BTRFS inside
> > > Proxmox? With qcow2 or raw images???
> >
> > I'm using btrfs for a trial vm image storage, but not yet at a point
> > where I could make a recommendation for or against it either way.
> >
> > -Paul