[Ovirt-devel] Question about ovirt migration

[Chris Lalancette]

Atsushi SAKAI wrote:
> Hi, Chris
> 
> Thank you for your description.
> 
> Then I want to understand the software layer for *fake* managed node.
> But, Nested KVM does not support at this moment.(Ref)
> Would you point the code of guest boot on fake managed node?
> 
> Ref.
> [PATCH 0/9] [RFC] Add support for nested SVM (kernel)
> http://thread.gmane.org/gmane.comp.emulators.kvm.devel/21119

No, no.  We aren't using this at all (it's not in KVM at the moment, I don't
think, and even if it was, there is no reason to use it in our case).  What
happens is that when we install the WUI appliance, we also "install" 3 fake
managed nodes alongside the WUI appliance.  These are basically just libvirt XML
files, that will PXE boot when you start them up.  So, what happens is that you
boot up the WUI appliance; once that is up, you can boot up your fake managed
nodes.  Each of these fake managed nodes appears in the WUI as Hosts where you
can start guests.  And, indeed, you can start guests inside of them; however,
the guests inside of them are fully emulated only (not KVM accelerated), because
we aren't using the nested SVM type stuff.

In point of fact, the nested SVM stuff isn't really interesting to us; the whole
point of ovirt is to manage physical hardware.  So the fake managed nodes are
just a useful abstraction to let people who have limited hardware at least see
what ovirt can do.

Chris Lalancette