On Mon, Jan 22, 2018 at 10:31 AM, Dan Kenigsberg <danken@redhat.com> wrote:

On Mon, Jan 22, 2018 at 12:01 AM, Matteo Capuano <kapu.net@gmail.com> wrote:

...

Hi folks,

i'm trying to install oVirt hyperconverged on Google Cloud Engine (GCE) to create a lab environment. Nested virtualization on GCE is in public beta, it seems to work fine but there's a catch: i cannot enable mac-spoofing on the GCE instances' nics. So here's my question: is there a way to install hosted-engine on a network there you can't enabling mac-spoofing?

I don't quite understand. Can you start multiple nested VM in a single GCE node, each communicating with its own mac? I believe that you'd need mac-spoofing for that simple task, too.

I can start three nested VMs in a single GCE instance and have them communicating with each others but this is going to add another virtualization level on my lab. What I'm trying to do is to launch three different GCE instances with Centos 7.4 and install on each of them the hyperconverged nodes. GCE doesn't allow to enable mac-spoofing on instance's nics so I'm unable to communicate with the engine from outside the host where it's installed. To make a nested VMs to be accessible from outside the host, GCE suggest to use a bridge network with iptables forwarding rules ( https://cloud.google.com/compute/docs/instances/enable-nested-virtualization... ) but this is not going to work with high-availability.

Self-hosted engine has to be highly-available. It has to run equally well on all hosted-engine hosts, thus I believe it needs all hosts to allow mac-spoofing.

So I'm trying to do something impossible. @Sandro : this is a test environment. I'm learning to use oVirt in my off-work time so I'm trying to see if I can move my learning platform from a dedicated server to a cloud service. Thank you both Matteo