On Tue, 8 May 2007 03:14:26 +0100 "Daniel P. Berrange" wrote:
> Hmm, this is interesting. So what your example is showing is that the
> current & max-memory settings disagree between the hypervisor and XenD.
> Whether the HV is more correct, or whether XenD is more correct is a
> difficult question to decide upon !
Yeah, it is difficult to decide which of HV or XenD is correct.
> What has happened in this scenario I think is that the guest was booted
> with maxmem == 500, hence the Hypervisor shows '500000 kB'. Later the
> guest's max memory was reduced to 400 using XenD, however, since you
> can't truely change the max-memory maps of the guest, the HV still shows
> 500, while XenD now shows the reduced 400 MV limit. IMHO, XenD is wrong
> here because it is showing an artificially low limit which the HV has
> no intention of enforcing. xm list is merely showing the info it gets back
> from XenD, so if XenD disagrees with the HV, then so will xm list.
I see. Now it makes sense. Thank you for your explanations !
> Since we have the libvirt_proxy, even when running as non-root apps will
> (indirectly) be calling xenHypervisorGetDomainInfo, so it will be very
> rare that libvirt ever calls xenDaemonDomainGetInfo.
>
> So I think we just need to accept that virsh dominfo, will disagree with
> xm list in some scenarios & document this difference to explain to people
> why libvirt is more correct ;-)
Okay, I quite agree.
thanks a lot !
Saori