Hi, Eric
On 12/08/2011 11:44 AM, Eric Blake wrote:
On 12/07/2011 08:40 PM, Lai Jiangshan wrote:
> "virt-top -1" can call virDomainGetPcpusUsage() periodically and get
> the CPU activities per CPU. (still require virt-top site patch).
>
> virsh is also added a pcpuinfo command which calls virDomainGetPcpusUsage(),
> it gets information about the physic CPUs, such as the usage of
> CPUs, the current attached vCPUs.
Meta-question - is this the time taken _by just the guest in question_
on the host's physical CPU, or is it the cumulative CPU time taken by
all process on the physical CPU, in which case this would be better
named as virNodeGetCpusUsage()?
It's per-domain usage.
And if it is per-domain, then what's
the difference between physical cpu usage and virtual cpu usage
the API tell us how many cpu utilization the whole guest occupies for
every physic CPUs. I don't know what is "virtual cpu usage" means.
(that
is, can we even tell the overhead of the hypervisor, and should we be
exposing that overhead to the user)?
it is not exposed to the normal user.
the administrator can use "virt-top -1" or "virsh pcpuinfo" to get
cpus
usage, to find out which physic cpu is busy, to change the cpu bound of (new) guests,
to find a proper physic cpu and offline it, or to do nothing, just to observe the stat of
the guests.
I/(I will) very appreciate for your reply.
Thanks,
Lai