[one-users] ping between VMs on different hosts ?

[Valerio Schiavoni]

Hello Jamie,

On Tue, Apr 24, 2012 at 4:19 PM, Jaime Melis <jmelis at opennebula.org> wrote:

> If you're using virbr0, it means you're using the bridge created by libvirt,
> which isn't the best approach.

Yes, I'm using that bridge. I thought that one was enough!


> Imagine your hosts are all connected through the eth0 interface (they all
> have eth0 and they're all plugged in to the same switch).

They are.

> You have to create
> a bridge (for instance br0) and attach eth0 to br0.

Honestly, this is nowhere described on the documentation. It just says
"should have two bridges:" but not that the bridges have to be
created, or that in the case of existing bridges made by libvirt it is
not a good idea :-) Anyway, good to know, thanks !

> Then, you need to use
> br0 as your bridge in your network definition.
>
> If eth0 used to have an ip, now you should set:
> $ ifconfig eth0 0.0.0.0 up
> $ ifconfig br0 <yourip>/<yournetmask> up
>
> Of course you need to do this on your hosts, not in the frontend (unless the
> frontend is also running VMs).

Sure.

>
> Also, you will need to contextualize your VMs so they get the IP OpenNebula
> has assigned to them. I guess you already have this working, since your
> email suggests your network configuration does work for VMs deployed on the
> same host.

I did get contextualization working using the vmcontex.sh set to run
at runlevel2, yes.
My problem with contextualization is that it seems to be a unix/bsd
-only solution.
How do you use it for Windows images?

best,
valerio