Open Stack

Hello,

I have a question about the l2gw. I did a deployment, I described the
steps here:
https://review.openstack.org/#/c/453209/

The unicast traffic works fine, but I dont understand what is the idea
behind the handling of the broadcast traffic.

Looking at openvswitch:

I obtain the uuid with `vtep-ctl list-ls`

vtep-ctl list-remote-macs <uuid>

In this output I get an entry for each VM that has an interface in the
L2 network I am bridging:

----
# vtep-ctl list-remote-macs <uuid>
ucast-mac-remote
  fa:16:3e:c2:7b:da -> vxlan_over_ipv4/10.1.1.167

mcast-mac-remote
-----

The ucast-mac-remote entry is created by Openstack when I start a VM.
(Also it is never removed when I delete the instance, is this a bug ? )
Note that 10.1.1.167 is the IP address of the hypervisor where the VM is
running.

But mcast-mac-remote is empty. So this means that ARP learning for
example works only in 1 way. The VM in openstack does not receive any
broadcast traffic, unless I do manually:

vtep-ctl add-mcast-remote ee87db33-1b3a-42e9-bc09-02747f8a0ad5
unknown-dst  10.1.1.167

This creates an entry in the table mcast-mac-remote and everything works
correctly.


Now I read here http://networkop.co.uk/blog/2016/05/21/neutron-l2gw/
about sending add-mcast-remote to the network nodes and then doing some
magic I dont really understand. But I am confused because in my setup
the tenant does not have a L3 router, so there is not a qrouter
namespace for this network, I was planning to keep the network node out
of the game.

Is anyone running this in production and can shed some light ?

thanks

Saverio











-- 
SWITCH
Saverio Proto, Peta Solutions
Werdstrasse 2, P.O. Box, 8021 Zurich, Switzerland
phone +41 44 268 15 15, direct +41 44 268 1573
saverio.proto at switch.ch, http://www.switch.ch

http://www.switch.ch/stories