[libvirt] [PATCH 0/9] Add basic driver for Parallels Virtuozzo Server

[Daniel Veillard]

On Tue, Apr 17, 2012 at 10:01:50PM +0400, Dmitry Guryanov wrote:
> On 04/16/2012 06:13 AM, Daniel Veillard wrote:
> >On Fri, Apr 13, 2012 at 10:26:14PM +0400, Dmitry Guryanov wrote:
> >>Parallels Virtuozzo Server is a cloud-ready virtualization
> >>solution that allows users to simultaneously run multiple virtual
> >>machines and containers on the same physical server.
> >>
> >>Current name of this product is Parallels Server Bare Metal and
> >>more information about it can be found here -
> >>http://www.parallels.com/products/server/baremetal/sp/.
> >>
> >>This driver will work with PVS version 6.0 , beta version
> >>scheduled at 2012 Q2.
> >   Okay, I started to review this version 2 of the driver. IIRC the
> >first version was relying on an API which wasn't LGPL compatible so that
> >was a no go.
> >   What I understand from that second version is that now the driver
> >talks to the prlctl command to get information back (using JSON). That
> >looks acceptable, but before making further review can you fix the
> >following things I found out in reviewing quickly patch 1-3
> >
> >   - the driver open doesn't seems to make any check about availability
> >     of prlctl command, it should fail to open if this is not found at
> >     Open() time.
> >   - all the entry points in the driver structure are marked as 0.9.11,
> >     since 0.9.11 is out already this would need to be bumped to 0.9.12
> >     assuming that will be the next version and the patches make it in
> >     time for that release (scheduled at the end of the month)
> Ok, I'll correct these things.
> 
> >   - the configure check blindly assumes that if compiled on linux the
> >     pvs driver should be activated. Since we rely on a command line tool
> >     to provide the interface that's a relatively safe assumption, but
> >     if I understand correctly Parrallels requires a modified kernel
> >     version, which is not upstream, right ?
> PVS is a full distribution, based on cloud linux and it
> has modified kernel.
> >To that extend you're at the
> >     same level as the OpenVZ driver (you rely on it underneath, right ?)
> >     Do you support all linux archs ? If not please improve the configure
> >     time check to the architecture you actually support.
> Ok, I'll add additional checks.

  Okay, should be simple enough to fix those 3 bits

> >In general I wonder why make it a new driver instead of ramping up the
> >existing OpenVZ one, are the 2 really incompatible, or is the existing
> >OpenVZ not proper for current versions ? Basically I wonder if we really
> >need 2 drivers assuming the implementation of the hypervisor is based on
> >the same core for both,
> First, PVS and OpenVZ are different products, PVS includes
> full virtualization support and OS level one, OpenVz supports
> only OS level virtualization. In PVS both types of virtual
> environments can be managed using single prlctl utility, while
> vzctl can handle only containers. This version of pvs driver
> supports only virtual machines, but support of containers
> planned too.
> 
> Second, prlctl and vzctl+vzlist have completely different output,
> command-line parameters also differ, especially in devices and
> network parameters. I think that drivers, which supports two
> types of virtualization and works with two different utilities
> will be too complicated and it's better to have two separate
> drivers.
> Also OpenVZ driver works directly with containers config
> files in some places, which is not possible while using prlctl.

  Okay, thanks for the explanations :-) The simple fact that PVS
can do full virt support is IMHO a sufficient differentiator !

Daniel

-- 
Daniel Veillard      | libxml Gnome XML XSLT toolkit  http://xmlsoft.org/
daniel at veillard.com  | Rpmfind RPM search engine http://rpmfind.net/
http://veillard.com/ | virtualization library  http://libvirt.org/