Xen project Mailing List

Please correct me if i am wrong.From my understanding(Eventhough it seems wrong) All the process in domU polls its cpu which is actually virtual and process queue in this virtual cpu is processed by the actual CPU in which hypervisor has direct access to it.In that case if there are lot of process in the process queue of domU waiting to get cpu of domU,then wouldnt that affect actual CPU,since it is the actual processor that process the process queue of domU.(Even though It seems wrong).

Kev

On Fri, Nov 30, 2012 at 3:24 PM, Ian Campbell <Ian.Campbell@xxxxxxxxxx> wrote:

On Wed, 2012-11-28 at 16:17 +0000, kevin parker wrote:
> Hi all,
> I am trying to understand this.I am running XCP 1.6 on a
> machine with 4GB and i5 processor with 3 ubuntu 12.04 64 bit domUs
> each with one vcpu.
> On one of the domU i am having mongodb taking 100% CPU resulting in
> load average of 2.1, 2, 2.But still load average on dom0 is 0, 0, 0.
> If the vcpu is over utilized then why there is no impact on dom0,as
> its load average is showing 0.

Unless the domU is requiring services from dom0 (i.e. heavy i/o traffic
to a backend based in dom0) there's no reason that domU load average
should have any impact on dom0's (or any other domU) load average.

> on the xentop output i can see all other domU are idle but this
> showing 100% CPU usage.Can any one please explain why this is so?
> If there are 4 cores on dom0 and one of the vm takes 100% CPU then
> doesn't load average on dom0 should be 1.Can any one please clarify
> me.

vcpus under Xen separate entities isolated, domU vcpus do not somehow
consume dom0 time.

You are perhaps confusing this with the KVM model where guest vcpus are
actually threads in the host OS (which is equivalent to dom0).

Ian.

Re: [Xen-users] CPU load average above 2 on domU but load average 0 on dom0-XCP 1.6