openstack-ansible-os_neutron/doc/source/app-openvswitch.rst
Guoqiang Ding 93d83fb02d Update the doc link
Change-Id: I6f7e3e525ce43076526c6fe2fad3a2a0b1a67e22
2017-11-24 11:30:51 +08:00

5.1 KiB

Scenario - Using Open vSwitch

Overview

Operators can choose to utilize Open vSwitch instead of Linux Bridges for the neutron ML2 agent. This offers different capabilities and integration points with neutron. This document outlines how to set it up in your environment.

We recommend that you read the following documents before proceeding:

Prerequisites

All compute nodes must have bridges configured:

  • br-mgmt
  • br-vlan (optional - used for vlan networks)
  • br-vxlan (optional - used for vxlan tenant networks)
  • br-storage (optional - used for certain storage devices)

For more information see: http://docs.openstack.org/project-deploy-guide/openstack-ansible/newton/targethosts-networkconfig.html

These bridges may be configured as either a Linux Bridge (which would connect to the Open vSwitch controlled by neutron) or as an Open vSwitch.

Configuring bridges (Linux Bridge)

The following is an example of how to configure a bridge (example: br-mgmt) with a Linux Bridge on Ubuntu 16.04 LTS:

/etc/network/interfaces

auto lo
iface lo inet loopback

# Management network
auto eth0
iface eth0 inet manual

# VLAN network
auto eth1
iface eth1 inet manual

source /etc/network/interfaces.d/*.cfg

/etc/network/interfaces.d/br-mgmt.cfg

# OpenStack Management network bridge
auto br-mgmt
iface br-mgmt inet static
  bridge_stp off
  bridge_waitport 0
  bridge_fd 0
  bridge_ports eth0
  address MANAGEMENT_NETWORK_IP
  netmask 255.255.255.0

One br-<type>.cfg is required for each bridge. VLAN interfaces can be used to back the br-<type> bridges if there are limited physical adapters on the system.

Configuring bridges (Open vSwitch)

Another configuration method routes everything with Open vSwitch. The bridge (example: br-mgmt) can be an Open vSwitch itself.

The following is an example of how to configure a bridge (example: br-mgmt) with Open vSwitch on Ubuntu 16.04 LTS: *

/etc/network/interfaces

auto lo
iface lo inet loopback

source /etc/network/interfaces.d/*.cfg

# Management network
allow-br-mgmt eth0
iface eth0 inet manual
  ovs_bridge br-mgmt
  ovs_type OVSPort

# VLAN network
allow-br-vlan eth1
iface eth1 inet manual
  ovs_bridge br-vlan
  ovs_type OVSPort

/etc/network/interfaces.d/br-mgmt.cfg

# OpenStack Management network bridge
auto br-mgmt
allow-ovs br-mgmt
iface br-mgmt inet static
  address MANAGEMENT_NETWORK_IP
  netmask 255.255.255.0
  ovs_type OVSBridge
  ovs_ports eth0

One br-<type>.cfg is required for each bridge. VLAN interfaces can be used to back the br-<type> bridges if there are limited physical adapters on the system.

Warning: There is a bug in Ubuntu 16.04 LTS where the Open vSwitch service won't start properly when using systemd. The bug and workaround are discussed here: http://www.opencloudblog.com/?p=240

OpenStack-Ansible user variables

Set the following user variables in your /etc/openstack_deploy/user_variables.yml: *

# Ensure the openvswitch kernel module is loaded
openstack_host_specific_kernel_modules:
  - name: "openvswitch"
    pattern: "CONFIG_OPENVSWITCH="
    group: "network_hosts"

### neutron specific config
neutron_plugin_type: ml2.ovs

neutron_ml2_drivers_type: "flat,vlan"

# Typically this would be defined by the os-neutron-install
# playbook. The provider_networks library would parse the
# provider_networks list in openstack_user_config.yml and
# generate the values of network_types, network_vlan_ranges
# and network_mappings. network_mappings would have a
# different value for each host in the inventory based on
# whether or not the host was metal (typically a compute host)
# or a container (typically a neutron agent container)
#
# When using Open vSwitch, we override it to take into account
# the Open vSwitch bridge we are going to define outside of
# OpenStack-Ansible plays
neutron_provider_networks:
  network_flat_networks: "*"
  network_types: "vlan"
  network_vlan_ranges: "physnet1:102:199"
  network_mappings: "physnet1:br-provider"

Customization is needed to support additional network types such as vxlan, GRE or Geneve. Refer to the neutron agent configuration for more information on these attributes.