Testing MoM

[Doron Fediuck]

----- Original Message -----
> From: "Adam Litke" <agl at us.ibm.com>
> To: "Doron Fediuck" <dfediuck at redhat.com>
> Cc: "infra" <infra at ovirt.org>
> Sent: Tuesday, November 27, 2012 4:07:19 PM
> Subject: Re: Testing MoM
> 
> Another thing we want to do is expand the unit tests.  A lot of
> functionality
> can be independently verified through these types of tests and it
> won't even
> require virtualization.  I am working to expand the test cases and
> we'll make
> sure that the existing Jenkins job runs those automatically.
> 

I'm sure the above is much simpler than getting a server :)
So regardless of unitests, we still need the ability to test ballooning
and memory sharing which requires real iron.

> On Tue, Nov 27, 2012 at 06:55:03AM -0500, Doron Fediuck wrote:
> > Hi Guys,
> > since mom is now required by vdsm, we'd like to have Jenkins
> > running a few tests
> > of vdsm+mom to make sure we have no regressions, etc.
> > 
> > The thing with mom is, that we need a real host to see memory
> > sharing (KSM)
> > kicks in correctly, and if the host starts swapping mom tunes the
> > VMs memory balloon
> > to fix it.
> > 
> > So basically the question is, if we have a real jenkins slave,
> > which we can use
> > as a hypervisor for mom tests. Obviously as a slave it may run
> > other tests while
> > vdsm+mom tests are not running.
> > 
> > I thought about nested kvm, but IIUC advanced features such as
> > memory balloon
> > may have issues there.
> > 
> > Appreciate your feedback on this.
> > Doron
> > 
> 
> --
> Adam Litke <agl at us.ibm.com>
> IBM Linux Technology Center
> 
>