charm-neutron-gateway/README.md

126 lines
4.7 KiB
Markdown

Overview
--------
Neutron provides flexible software defined networking (SDN) for OpenStack.
This charm is designed to be used in conjunction with the rest of the OpenStack
related charms in the charm store to virtualize the network that Nova Compute
instances plug into.
Neutron supports a rich plugin/extension framework for propriety networking
solutions and supports (in core) Nicira NVP, NEC, Cisco and others...
See the upstream [Neutron documentation](https://docs.openstack.org/neutron/latest/)
for more details.
Usage
-----
In order to use Neutron with OpenStack, you will need to deploy the
nova-compute and nova-cloud-controller charms with the network-manager
configuration set to 'Neutron':
nova-cloud-controller:
network-manager: Neutron
This decision must be made prior to deploying OpenStack with Juju as
Neutron is deployed baked into these charms from install onwards:
juju deploy nova-compute
juju deploy --config config.yaml nova-cloud-controller
juju deploy neutron-api
juju add-relation nova-compute nova-cloud-controller
juju add-relation neutron-api nova-cloud-controller
The Neutron Gateway can then be added to the deploying:
juju deploy neutron-gateway
juju add-relation neutron-gateway mysql
juju add-relation neutron-gateway rabbitmq-server
juju add-relation neutron-gateway nova-cloud-controller
juju add-relation neutron-gateway neutron-api
The gateway provides two key services; L3 network routing and DHCP services.
These are both required in a fully functional Neutron OpenStack deployment.
Configuration Options
---------------------
Port Configuration
==================
All network types (internal, external) are configured with bridge-mappings and
data-port and the flat-network-providers configuration option of the
neutron-api charm. Once deployed, you can configure the network specifics
using neutron net-create.
If the device name is not consistent between hosts, you can specify the same
bridge multiple times with MAC addresses instead of interface names. The charm
will loop through the list and configure the first matching interface.
Basic configuration of a single external network, typically used as floating IP
addresses combined with a GRE private network:
neutron-gateway:
bridge-mappings: physnet1:br-ex
data-port: br-ex:eth1
neutron-api:
flat-network-providers: physnet1
neutron net-create --provider:network_type flat \
--provider:physical_network physnet1 --router:external=true \
external
neutron router-gateway-set provider external
Alternative configuration with two networks, where the internal private
network is directly connected to the gateway with public IP addresses but a
floating IP address range is also offered.
neutron-gateway:
bridge-mappings: physnet1:br-data external:br-ex
data-port: br-data:eth1 br-ex:eth2
neutron-api:
flat-network-providers: physnet1 external
Alternative configuration with two external networks, one for public instance
addresses and one for floating IP addresses. Both networks are on the same
physical network connection (but they might be on different VLANs, that is
configured later using neutron net-create).
neutron-gateway:
bridge-mappings: physnet1:br-data
data-port: br-data:eth1
neutron-api:
flat-network-providers: physnet1
neutron net-create --provider:network_type vlan \
--provider:segmentation_id 400 \
--provider:physical_network physnet1 --shared external
neutron net-create --provider:network_type vlan \
--provider:segmentation_id 401 \
--provider:physical_network physnet1 --shared --router:external=true \
floating
neutron router-gateway-set provider floating
This replaces the previous system of using ext-port, which always created a bridge
called br-ex for external networks which was used implicitly by external router
interfaces.
Note: if the 'data-port' config item is set, then the 'ext-port' option is
ignored. This is to prevent misconfiguration of the charm. A warning is
logged and the unit is marked as blocked in order to indicate that the charm is
misconfigured.
Instance MTU
============
When using Open vSwitch plugin with GRE tunnels default MTU of 1500 can cause
packet fragmentation due to GRE overhead. One solution is to increase the MTU on
physical hosts and network equipment. When this is not possible or practical the
charm's instance-mtu option can be used to reduce instance MTU via DHCP.
juju set neutron-gateway instance-mtu=1400
Note that this option was added in Havana and will be ignored in older releases.