[Openstack] [openstack-dev][neutron]

Andreas Scheuring scheuran at linux.vnet.ibm.com
Wed Feb 22 07:44:33 UTC 2017


Hi Roua, 
AFAIK this is not possible with OpenStack. Neutron only has the concept
of attaching ports to a network. A concept where you do not attach a
port to a network - but create a direct link between instances does not
exist yet (except that you can create dedicated a network for each
direct link - but this would still use a bridge).

I wonder what the problem is that you try to solve with having that
mesh. And why using a bridge is not an option.


Before thinking of any of the implementation specifics, I would ask you
to come up with a design on how you create the mesh. I'm not sure how
you wanna achieve this using a tap or a veth without having a bridge
involved...



-- 
-----
Andreas 
IRC: andreas_s



On Di, 2017-02-21 at 16:03 +0100, Roua Touihri wrote:
> Hello everybody,
> 
> 
> How can we interconnect two VMs without using a bridge or a switch as
> an intermediate. That is, only via a virtual link (e.g. veth or tap).
> In fact, I see that when we create an aditional subnet and two ports
> of the given subnet. Then when I attach each port to a running VM,
> neutron use a bridge as an intermediate element. I want to create a
> mesh topology between several VMs of the same tenant in addition to or
> without the default network created by neutron.
> 
> 
> If such a configuration were not possible, I must then create new APIs
> for doing that. However, I do not know how to get started with this
> task. I guess that I should create new classes similarly to the
> "neutron.create_port" one and maybe to overload the veth/tap
> constructor.
> 
> 
> Thanks in advance
> 
> 
> 
> -- 
> 
> ​Kindly,
> 
> 
> 
> 
> R.T
> 
> 
> _______________________________________________
> Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
> Post to     : openstack at lists.openstack.org
> Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack





More information about the Openstack mailing list