virt functional tests and testing improvements

[Antoni Segura Puimedon]

----- Original Message -----
> From: "Dan Kenigsberg" <danken at redhat.com>
> To: "Antoni Segura Puimedon" <asegurap at redhat.com>
> Cc: "infra" <infra at ovirt.org>
> Sent: Sunday, October 6, 2013 11:29:44 AM
> Subject: Re: virt functional tests and testing improvements
> 
> On Sat, Oct 05, 2013 at 08:55:42AM -0400, Antoni Segura Puimedon wrote:
> > Hi Everybody,
> > 
> > We are readying the way to run the virt functional tests as a jenkins job.
> > In
> > order to achieve that, we'd like to be able to reboot one of the hosts
> > holding
> > the jenkins slave VMs so that the setting "option kvm_intel nested=1" (This
> > is required because virt functional tests spawn a few VMs). It would also
> > be
> > necessary to have a vdsm hook for the VM creation that puts the vmx
> > capability
> > on the guests XML.
> 
> I vote in favor of this suggestion since we urgently need to know if and
> when vm startup is broken. Thanks for this initiative.
> 
> However, we should all remember that nested=1 is not yet of enterprise
> quality. If we set it, and then see instability introduced to our host,
> we should reset it and walk the less aggressive path of using
> faqemu-like hook on the host, running our tested guest in emulation.

Do we have some host with AMD cpu? I heard that the nested virt works much
better with their CPUs.

> 
> > 
> > A related improvement would be to be able to have some guide for adding
> > one's
> > hosts to jenkins as slaves for extra cycles and maybe to have some
> > dedicated
> > slaves for complex testing setups.
>