Re: [libvirt] [Qemu-devel] pvpanic plans?
On Tue, Aug 27, 2013 at 8:20 AM, Richard W.M. Jones <rjones(a)redhat.com> wrote:
On Tue, Aug 27, 2013 at 04:08:12PM +0300, Ronen Hod wrote:
> So the right solution is to send a heart-beat to a management
> application (using qemu-ga or whatever), and let it decide how to
> handle it.
This is host-centric solution and assumes that a management tool is
making all of the decisions. This doesn't work in an IaaS environment
where these sort of policy decisions need to be driven from the guest.
Furthermore, you really want the watchdog daemon to run with real time
priority which implies a heightened privilege level. This rules out
using qemu-ga for that purpose.
Agreed. The qemu watchdog lets you do this already. You can (using
the qemu monitor, or libvirt) capture watchdog events and put them
into your management application. Watchdog firing does *not*
necessarily mean a guest reboot.
Ack, but the current watchdog does not work for Windows guests and is
not aware of guest time.
That's why I think having a virtio-ilo makes sense. This is not a
solved problem today.
Regards,
Anthony Liguori
[Note what I say applies to the qemu watchdog device. The Linux
watchdog daemon may independently initiate a guest reboot, but you can
configure it to perform other actions instead.]
Rich.
--
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Fedora Windows cross-compiler. Compile Windows programs, test, and
build Windows installers. Over 100 libraries supported.
http://fedoraproject.org/wiki/MinGW