On Thu, Mar 08, 2007 at 12:11:48AM +0530, Shuveb Hussain wrote:
Hi,
Any reason why actual requests are serviced by daemons? Is it just to
increase the stability of the applications that link to libvirt or are
there other reasons? I have started implementing OpenVZ support.
Should I follow the same path and implement a daemon as well?
The answer is rather complicated, and depends on the backend being used.
With Xen...
- If possible we do direct hypercalls - only certain operations can
be done in this way though, and only if you are root.
- If you are non-root, we use a small setuid process to let non-root
make direct hypercalls, bypassing the XenD daemon.
- For some operations, we have no choice but to talk to XenD daemon
With QEMU...
- There is no way to enumerate running QEMU instances, or to get the
config of existing instances. Thus the only way to track this info
is to have a daemon which is used to manage QEMU instances
For remote access, we're working on a general purpose daemon so that
any driver can be used remotely, even if the driver itself is not network
aware.
So in summary, for OpenVZ if you can write the driver without needing a
daemon that is just fine - and indeed preferred - since having a daemon
adds a significant amount of complexity.
Regards,
Dan.
--
|=- Red Hat, Engineering, Emerging Technologies, Boston. +1 978 392 2496 -=|
|=- Perl modules:
http://search.cpan.org/~danberr/ -=|
|=- Projects:
http://freshmeat.net/~danielpb/ -=|
|=- GnuPG: 7D3B9505 F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 -=|