Yes, when using the Open vSwitch plugin with VLAN provider networks, you must configure the physical switchports to work with host based tagging. Only the switchports that connect the NICs defined in bridge_mappings from ovs_quantum_plugin.ini need to be configured this way. These switchports need to allow the range of VLAN ids defined in network_vlan_ranges. The terms and procedures to do this vary between switch vendors, so you will need to follow your switch doc.
Which dhcp server do you mean? A quantum dhcp server or an external ('real') dhcp server? How many servers do you have? How many nics do they have?
I had this problem as well and my solution was to trunk the ports to which openstack was connected and create the vlans (range) on the switch itself.
interface Eth1/1 - 30
switchport mode trunk
switch port trunk allowed vlan all
vlan 100-200
This trunked the ports on the switch (physical ports 1 through 30), allowed all vlan traffic, and then let the switch know about vlans in the range 100 through 200. After that, openstack (the virtual interfaces neutron's agents create) will tag the outbound traffic for vlans 100 through 200 (allowed vlans setting in the neutron configuration options) and the switch will accept and route that traffic.
Asked: 2013-09-04 03:06:59 -0600
Seen: 1,418 times
Last updated: Nov 12 '13
