Neutron/TrunkPort

Overview

Neutron extension to access lots of neutron networks over a single vNIC as tagged/encapsulated traffic.

API

FIXME attribute types

neutron trunk port

attribute name	attribute type
admin_state_up
binding:host_id	str
binding:vif_details
binding:vif_type
device_id	uuid
device_owner
id	uuid
mac_address
name	str
status
tenant_id	uuid

wishlist attributes:

subport_count

neutron subport

Usage of pre-existing neutron port attributes:

attribute name	attribute value
device_id	UUID OF PARENT TRUNK PORT
device_owner	trunk-port

Additional neutron port attributes:

attribute name	attribute type
trunk-port:segmentation-id	uint
trunk-port:segmentation-type	str

Ignored and problematic neutron port attributes:

attribute name	reason
mac_address	no mechanism to tell the guest os/app what mac to use
binding:*	FIXME

nova

FIXME nova boot --nic trunk-port-id

CLI usage example

# Trunk ports are created independently of networks.
neutron trunk-port-create --name trunk-port0

# Networks for later subports.
neutron net-create net0
neutron net-create net1

# Optional subnets.
neutron subnet-create net0 10.0.0.0/24
neutron subnet-create net1 10.0.1.0/24

# Ports having --device-owner 'network:trunk-port' are subports of the trunk port given by uuid.
# The subport without --trunk-port:* options is the default subport.
# The default subport's traffic will be seen as untagged inside the instance.
# You likely want network connectivity during boot, so you should create at least the default subport before booting.
neutron port-create net0 --name port0 --device-owner network:trunk-port --device-id TRUNK-PORT0-UUID

# Other subports can be created at any time, including before boot.
# The traffic of further subports has to be differentiated inside the instance by encapsulation, so you need to provide a segmentation type and id.
neutron port-create net1 --name port1 --device-owner network:trunk-port --device-id TRUNK-PORT0-UUID --trunk-port:segmentation-type vlan --trunk-port:segmentation-id 101

# The only vNIC in your instance corresponds to the trunk port, so boot your instance with the trunk port given. Do not add subports as NICs to 'nova boot'.
# Use an image with support for vlan interfaces. CirrOS will not cut it. eg: sudo ip link add ... type vlan ...
nova boot ... --image VLAN-CAPABLE-IMAGE --nic trunk-port-id=TRUNK-PORT0-UUID --poll vm0

# The typical cloud image will auto-configure eth0 only and not the vlan interfaces (eth0.VLAN-ID).
ssh VM0-ADDRESS sudo ip link add link eth0 name eth0.101 type vlan id 101

# Other subports can be created at any time, including after boot.
neutron net-create net2
neutron subnet-create net2 10.0.2.0/24
neutron port-create net2 --name port2 --device-owner network:trunk-port --device-id TRUNK-PORT0-UUID --trunk-port:segmentation-type vlan --trunk-port:segmentation-id 102

# Again you need to bring your subport vlan interfaces up.
ssh VM0-ADDRESS sudo ip link add link eth0 name eth0.102 type vlan id 102

# Subports can be deleted at runtime too.
ssh VM0-ADDRESS sudo ip link delete dev eth0.102
neutron port-delete port1

# When you're all done, deleting the trunk port deletes all the subports too.
nova delete vm0
neutron trunk-port-delete trunk-port0
neutron net-delete net2
neutron net-delete net1
neutron net-delete net0

openvswitch vlan model

FIXME draw picture of wiring of tap interfaces, trunk and integration bridges

# create trunk bridge
ovs-vsctl --may-exist add-br TRUNK-BRIDGE
ovs-ofctl del-flows TRUNK-BRIDGE

# patch trunk bridge to integration bridge
ovs-vsctl add-port TRUNK-BRIDGE PATCH-TRUNK-END-PORT -- set Interface PATCH-TRUNK-END-PORT type=patch options:peer=PATCH-INT-END-PORT
ovs-vsctl add-port br-int PATCH-INT-END-PORT -- set Interface PATCH-INT-END-PORT type=patch options:peer=PATCH-TRUNK-END-PORT

# install default drop flows on patch ports
ovs-ofctl add-flow TRUNK-BRIDGE hard_timeout=0,idle_timeout=0,priority=2,in_port=PATCH-TRUNK-END-OFPORT,actions=drop
ovs-ofctl add-flow br-int hard_timeout=0,idle_timeout=0,priority=2,in_port=PATCH-INT-END-OFPORT,actions=drop

# make the tap port pass all tags
ovs-vsctl set Port TAP-PORT vlan_mode=trunk

##

# install remap flows for untagged traffic (default subport)
ovs-ofctl add-flow TRUNK-BRIDGE hard_timeout=0,idle_timeout=0,priority=3,dl_vlan=COMPUTE-INTERNAL-VLAN-ID,in_port=PATCH-TRUNK-END-OFPORT,actions=strip_vlan,normal
ovs-ofctl add-flow br-int hard_timeout=0,idle_timeout=0,priority=3,dl_vlan=0xffff,in_port=PATCH-INT-END-OFPORT,actions=mod_vlan_vid:COMPUTE-INTERNAL-VLAN-ID,normal

# install remap flows for tagged traffic (other subports)
ovs-ofctl add-flow TRUNK-BRIDGE hard_timeout=0,idle_timeout=0,priority=3,dl_vlan=COMPUTE-INTERNAL-VLAN-ID,in_port=PATCH-TRUNK-END-OFPORT,actions=mod_vlan_vid:INNER-VLAN-ID,normal
ovs-ofctl add-flow br-int hard_timeout=0,idle_timeout=0,priority=3,dl_vlan=INNER-VLAN-ID,in_port=PATCH-INT-END-OFPORT,actions=mod_vlan_vid:COMPUTE-INTERNAL-VLAN-ID,normal