<p dir="ltr">We aren't too far off from being able to support this now. The agent-based ML2 drivers already bind based on a combination of VNIC type and the host so multiple agents with different types can run on the same host. </p>
<p dir="ltr">I think the only part that is missing is each driver being able to inform Nova which bridge to plug into. </p>
<div class="gmail_quote">On Feb 28, 2016 7:48 PM, "Assaf Muller" <<a href="mailto:amuller@redhat.com">amuller@redhat.com</a>> wrote:<br type="attribution"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">On Sat, Feb 27, 2016 at 6:55 PM, Martinx - ジェームズ<br>
<<a href="mailto:thiagocmartinsc@gmail.com">thiagocmartinsc@gmail.com</a>> wrote:<br>
> Hey guys!<br>
><br>
> Next Ubuntu and Mitaka are promising something ultra mega cool!<br>
><br>
> Look at this!<br>
><br>
> ---<br>
> root@mitaka-1:~# apt install neutron-openvswitch-agent<br>
> Reading package lists... Done<br>
> Building dependency tree<br>
> Reading state information... Done<br>
> The following additional packages will be installed:<br>
> dpdk libdpdk0 openvswitch-common openvswitch-switch<br>
> ---<br>
><br>
> Xenial will brings DPDK-2.2 fully supported for 5 years!<br>
><br>
> However, I am curious about the following scenarios:<br>
><br>
> Will be possible to use, at the same time (same Network and Compute nodes /<br>
> Host Aggregate):<br>
><br>
> 1- Regular OVS bridges without DPDK for VXLAN Networks, with<br>
> OVS-Firewall-Driver and;<br>
><br>
> 2- OVS powered by DPDK for Provider Networks only ( without any firewall,<br>
> current case anyway, due to<br>
> <a href="https://bugs.launchpad.net/neutron/+bug/1531205" rel="noreferrer" target="_blank">https://bugs.launchpad.net/neutron/+bug/1531205</a>).<br>
<br>
Currently, a host may run a single OVS agent, configured for either<br>
regular OVS or OVS-DPDK. You cannot run both on a single host. You can<br>
mix and match between different hosts though. It is something we<br>
discussed a bit, but no concrete plans to change this at this time.<br>
<br>
We could support this by allowing an OVS agent to support two<br>
datapaths simultaneously by configuring two integration bridges, each<br>
with its own type. We would add a DPDK VNIC type so Nova would plug<br>
the VNIC to the correct bridge. Each integration bridge would have its<br>
own bridge mappings (The kernel datapath integration bridge would be<br>
connected to br-tun or to a VLAN bridge, and the DPDK datapath<br>
integration bridge would be connected to its own set of VLAN provider<br>
bridges. Another way to accomplish this use case is to start two OVS<br>
agents on the same host, each configured appropriately, but we'd need<br>
to make changes to ML2 to support this, perhaps differentiate between<br>
the two agents via an agent_type and bind ports appropriately. Again,<br>
we'd need a new VNIC type for DPDK ports.<br>
<br>
><br>
> ?<br>
><br>
> I have NFV Instances that are also, DPDK L2 Bridges running on KVM Guest /<br>
> VirtIO, that are physically wired using Provider Networks (flat and vlans).<br>
><br>
> So, for the Instance's vNICs (eth1 and eth2) that are used as a L2 bridge,<br>
> I don't want any kind of ovs-firewall (I'm not affected by LP #1531205 on<br>
> this case) and I want OVS+DPDK under it but, for SSH into the Instance to<br>
> manage it (via its eth0), it is still using regular VXLAN with Security<br>
> Groups - OVS-Firewall from now on (no need for DPDK under eth0 / VXLAN).<br>
><br>
> I'm curious about this specially because the OVS Ubuntu package, makes use<br>
> of Debian's Alternatives subsystem, and we need to choose one OVS (default),<br>
> or another (with DPDK), via "update-alternatives", so, will be possible to<br>
> select OVS with DPDK but, use regular bridges with it as well (for VXLAN<br>
> networks)?<br>
><br>
> If yes, how to create a VXLAN network with regular OVS and another<br>
> FLAT/VLAN network with OVS+DPDK ?<br>
><br>
> Thanks in advance!<br>
><br>
> Best,<br>
> Thiago<br>
><br>
> __________________________________________________________________________<br>
> OpenStack Development Mailing List (not for usage questions)<br>
> Unsubscribe: <a href="http://OpenStack-dev-request@lists.openstack.org?subject:unsubscribe" rel="noreferrer" target="_blank">OpenStack-dev-request@lists.openstack.org?subject:unsubscribe</a><br>
> <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" rel="noreferrer" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
><br>
<br>
__________________________________________________________________________<br>
OpenStack Development Mailing List (not for usage questions)<br>
Unsubscribe: <a href="http://OpenStack-dev-request@lists.openstack.org?subject:unsubscribe" rel="noreferrer" target="_blank">OpenStack-dev-request@lists.openstack.org?subject:unsubscribe</a><br>
<a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" rel="noreferrer" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
</blockquote></div>