openstack-ansible-os_neutron/doc/source/configure-network-services.rst

Configuring the Networking service (neutron) (optional)

The OpenStack Networking service (neutron) includes the following services:

Firewall as a Service (FWaaS)

Provides a software-based firewall that filters traffic from the router.

VPN as a Service (VPNaaS)

Provides a method for extending a private network across a public network.

BGP Dynamic Routing service

Provides a means for advertising self-service (private) network prefixes to physical network devices that support BGP.

SR-IOV Support

Provides the ability to provision virtual or physical functions to guest instances using SR-IOV and PCI passthrough. (Requires compatible NICs)

Firewall service (optional)

The following procedure describes how to modify the /etc/openstack_deploy/user_variables.yml file to enable FWaaS.

Deploying FWaaS v1

Note

The FWaaS v1 API is deprecated upstream. While FWaaS v1.0 is still maintained, new features will be implemented in FWaaS v2.0 API.

1. Override the default list of neutron plugins to include firewall:

   neutron_plugin_base:
     - firewall
     - ...

2. neutron_plugin_base is as follows:

   neutron_plugin_base:
      - router
      - firewall
      - vpnaas
      - metering
      - qos

3. Execute the neutron install playbook in order to update the configuration:

   # cd /opt/openstack-ansible/playbooks
   # openstack-ansible os-neutron-install.yml

4. Execute the horizon install playbook to show the FWaaS panels:

   # cd /opt/openstack-ansible/playbooks
   # openstack-ansible os-horizon-install.yml

The FWaaS default configuration options may be changed through the conf override mechanism using the neutron_neutron_conf_overrides dict.

Deploying FWaaS v2

FWaaS v2 is the next generation Neutron firewall service and will provide a rich set of APIs for securing OpenStack networks. It is still under active development.

Refer to the FWaaS 2.0 API specification for more information on these FWaaS v2 features.

FWaaS v2 requires the use of Open vSwitch. To deploy an environment using Open vSwitch for virtual networking, please refer to the following documentation:

• Scenario - Using Open vSwitch
• Scenario - Using Open vSwitch with DVR

Follow the steps below to deploy FWaaS v2:

Note

FWaaS v1 and v2 cannot be deployed simultaneously.

1. Add the FWaaS v2 plugin to the neutron_plugin_base variable in /etc/openstack_deploy/user_variables.yml:

   neutron_plugin_base:
     - router
     - metering
     - firewall_v2

   Ensure that neutron_plugin_base includes all of the plugins that you want to deploy with neutron in addition to the firewall_v2 plugin.

2. Run the neutron playbook to deploy the FWaaS v2 service plugin

   # cd /opt/openstack-ansible/playbooks
   # openstack-ansible os-neutron-install.yml

Virtual private network service (optional)

The following procedure describes how to modify the /etc/openstack_deploy/user_variables.yml file to enable VPNaaS.

1. Override the default list of neutron plugins to include vpnaas:

   neutron_plugin_base:
     - router
     - metering

2. neutron_plugin_base is as follows:

   neutron_plugin_base:
      - router
      - metering
      - vpnaas

3. Override the default list of specific kernel modules in order to include the necessary modules to run ipsec:

   openstack_host_specific_kernel_modules:
      - { name: "ebtables", pattern: "CONFIG_BRIDGE_NF_EBTABLES=", group: "network_hosts" }
      - { name: "af_key", pattern: "CONFIG_NET_KEY=", group: "network_hosts" }
      - { name: "ah4", pattern: "CONFIG_INET_AH=", group: "network_hosts" }
      - { name: "ipcomp", pattern: "CONFIG_INET_IPCOMP=", group: "network_hosts" }

4. Execute the openstack hosts setup in order to load the kernel modules at boot and runtime in the network hosts

   # openstack-ansible openstack-hosts-setup.yml --limit network_hosts\
   --tags "openstack_hosts-config"

5. Execute the neutron install playbook in order to update the configuration:

   # cd /opt/openstack-ansible/playbooks
   # openstack-ansible os-neutron-install.yml

6. Execute the horizon install playbook to show the VPNaaS panels:

   # cd /opt/openstack-ansible/playbooks
   # openstack-ansible os-horizon-install.yml

The VPNaaS default configuration options are changed through the conf override mechanism using the neutron_neutron_conf_overrides dict.

BGP Dynamic Routing service (optional)

The BGP Dynamic Routing plugin for neutron provides BGP speakers which can advertise OpenStack project network prefixes to external network devices, such as routers. This is especially useful when coupled with the subnet pools feature, which enables neutron to be configured in such a way as to allow users to create self-service segmented IPv6 subnets.

The following procedure describes how to modify the /etc/openstack_deploy/user_variables.yml file to enable the BGP Dynamic Routing plugin.

1. Add the BGP plugin to the neutron_plugin_base variable in /etc/openstack_deploy/user_variables.yml:

   neutron_plugin_base:
     - ...
     - neutron_dynamic_routing.services.bgp.bgp_plugin.BgpPlugin

   Ensure that neutron_plugin_base includes all of the plugins that you want to deploy with neutron in addition to the BGP plugin.

2. Execute the neutron install playbook in order to update the configuration:

   # cd /opt/openstack-ansible/playbooks
   # openstack-ansible os-neutron-install.yml

SR-IOV Support (optional)

The following procedure describes how to modify the OpenStack-Ansible configuration to enable Neutron SR-IOV support.

1. Define SR-IOV capable physical host interface for a provider network

   As part of every Openstack-Ansible installation, all provider networks known to Neutron need to be configured inside the /etc/openstack_deploy/openstack_user_config.yml file. For each supported network type (e.g. vlan), the attribute sriov_host_interfaces can be defined to map ML2 network names (net_name attribute) to one or many physical interfaces. Additionally, the network will need to be assigned to the neutron_sriov_nic_agent container group.

   Example configuration:

   provider_networks
     - network:
       container_bridge: "br-vlan"
       container_type: "veth"
       container_interface: "eth11"
       type: "vlan"
       range: "1000:2000"
       net_name: "physnet1"
       sriov_host_interfaces: "p1p1,p4p1"
       group_binds:
         - neutron_linuxbridge_agent
         - neutron_sriov_nic_agent

2. Configure Nova

   With SR-IOV, Nova uses PCI passthrough to allocate VFs and PFs to guest instances. Virtual Functions (VFs) represent a slice of a physical NIC, and are passed as virtual NICs to guest instances. Physical Functions (PFs), on the other hand, represent an entire physical interface and are passed through to a single guest.

   To use PCI passthrough in Nova, the PciPassthroughFilter filter needs to be added to the conf override nova_scheduler_default_filters. Finally, PCI devices available for passthrough need to be allow via the conf override nova_pci_passthrough_whitelist.

   Possible options which can be configured:

   # Single device configuration
   nova_pci_passthrough_whitelist: '{ "physical_network":"physnet1", "devname":"p1p1" }'
   
   # Multi device configuration
   nova_pci_passthrough_whitelist: '[{"physical_network":"physnet1", "devname":"p1p1"}, {"physical_network":"physnet1", "devname":"p4p1"}]'
   
   # Whitelisting by PCI Device Location
   # The example pattern for the bus location '0000:04:*.*' is very wide. Make sure that
   # no other, unintended devices, are whitelisted (see lspci -nn)
   nova_pci_passthrough_whitelist: '{"address":"0000:04:*.*", "physical_network":"physnet1"}'
   
   # Whitelisting by PCI Device Vendor
   # The example pattern limits matches to PCI cards with vendor id 8086 (Intel) and
   # product id 10ed (82599 Virtual Function)
   nova_pci_passthrough_whitelist: '{"vendor_id":"8086", "product_id":"10ed", "physical_network":"physnet1"}'
   
   # Additionally, devices can be matched by their type, VF or PF, using the dev_type parameter
   # and type-VF or type-PF options
   nova_pci_passthrough_whitelist: '{"vendor_id":"8086", "product_id":"10ed", "dev_type":"type-VF", physical_network":"physnet1"}'

   It is recommended to use whitelisting by either the Linux device name (devname attribute) or by the PCI vendor and product id combination (vendor_id and product_id attributes)

3. Enable the SR-IOV ML2 plugin

   The conf override neutron_plugin_type variable defines the core ML2 plugin, and only one plugin can be defined at any given time. The conf override neutron_plugin_types variable can contain a list of additional ML2 plugins to load. Make sure that only compatible ML2 plugins are loaded at all times. The SR-IOV ML2 plugin is known to work with the linuxbridge (ml2.lxb) and openvswitch (ml2.ovs) ML2 plugins. ml2.lxb is the standard activated core ML2 plugin.

   neutron_plugin_types:
     - ml2.sriov

4. Execute the Neutron install playbook in order to update the configuration:

   # cd /opt/openstack-ansible/playbooks
   # openstack-ansible os-neutron-install.yml
   # openstack-ansible os-nova-install.yml

5. Check Neutron SR-IOV agent state

   After the playbooks have finished configuring Neutron and Nova, the new Neutron Agent state can be verified with:

   # neutron agent-list --agent_type 'NIC Switch agent'
   +--------------------------------------+------------------+-----------+-------+----------------+-------------------------+
   | id                                   | agent_type       | host      | alive | admin_state_up | binary                  |
   +--------------------------------------+------------------+-----------+-------+----------------+-------------------------+
   | 3012ff0e-de35-447b-aff6-fdb55b04c518 | NIC Switch agent | compute01 | :-)   | True           | neutron-sriov-nic-agent |
   | bb0c0385-394d-4e72-8bfe-26fd020df639 | NIC Switch agent | compute02 | :-)   | True           | neutron-sriov-nic-agent |
   +--------------------------------------+------------------+-----------+-------+----------------+-------------------------+

Deployers can make changes to the SR-IOV nic agent default configuration options via the neutron_sriov_nic_agent_ini_overrides dict. Review the documentation on the conf override mechanism for more details.