Neutron DVR OpenStack Liberty

Distributed Virtual Routers aka DVR were created to avoid single point of failure on neutron nodes.

When using standard routers, all the traffic is passing out through Neutron servers. Inside network servers, router namespaces are created routing all traffic and NAT forwarding between instances and public networks. When a network node falls down, instance traffic will no longer be available until a new namespace is created and executed in another network node.

Distributed routers is a way to avoid the SPOF neutron nodes were. When using DVR, router namespaces, are directly created inside compute nodes where all instance and l3 traffic are routed.

If you want to know more about DVR check this awesome links:

• http://blog.gampel.net/2014/12/openstack-neutron-distributed-virtual.html

• http://blog.gampel.net/2014/12/openstack-dvr2-floating-ips.html

• http://blog.gampel.net/2015/01/openstack-DVR-SNAT.html

A previous OpenStack Liberty installation is required, mine was done with RDO packstack.

Configure all Neutron Servers

Edit ml2 configuration file with the following:

# vi /etc/neutron/plugins/ml2/ml2_conf.ini

mechanism_drivers = openvswitch,l2population
type_drivers = flat,vlan,vxlan
tenant_network_types = vxlan
vni_ranges = 10:100
vxlan_group = 224.1.1.1
enable_security_group = True

Edit neutron configuration file, enable DVR and uncomment dvr_base_mac option

# vi /etc/neutron/neutron.conf

router_distributed = True
dvr_base_mac = fa:16:3f:00:00:00

Configure l3 agent to use dvr_snat

# vi /etc/neutron/l3_agent.ini

agent_mode = dvr_snat

Restart neutron server

systemctl restart neutron-server

Configure all Compute Nodes

Install ml2 package

yum install openstack-neutron-ml2

Edit openvswitch agent file as below:

# vi /etc/neutron/plugins/ml2/openvswitch_agent.ini 

l2_population = True
arp_responder = True
enable_distributed_routing = True

Enable DVR and select an interface driver to be used by l3 agent

# vi /etc/neutron/l3_agent.ini

interface_driver = neutron.agent.linux.interface.OVSInterfaceDriver
agent_mode = dvr

Edit ml2 configuration file as below:

# vi /etc/neutron/plugins/ml2/ml2_conf.ini

type_drivers = flat,vlan,vxlan
tenant_network_types = vxlan
mechanism_drivers = openvswitch,l2population
vni_ranges = 10:100
vxlan_group = 224.1.1.1
enable_security_group = True

Start and enable metadata agent in compute nodes

systemctl start neutron-l3-agent neutron-metadata-agent
systemctl enable neutron-l3-agent neutron-metadata-agent

Create an external bridge with an external IP associated on it

# vi /etc/sysconfig/network-scripts/ifcfg-br-ex

DEVICE=br-ex
DEVICETYPE=ovs
TYPE=OVSBridge
BOOTPROTO=static
IPADDR=192.168.100.4                                                          
NETMASK=255.255.255.0
GATEWAY=192.168.100.1
ONBOOT=yes

Modify an unused interface connected with the same network as the IP configured with br-ex, edit the interface to be used as OVS port by br-ex

# vi /etc/sysconfig/network-scripts/ifcfg-eth1
DEVICE=eth1
TYPE=OVSPort
DEVICETYPE=ovs
OVS_BRIDGE=br-ex
ONBOOT=yes

Restart network service to apply changes on the interfaces and openvswith-agent

systemctl restart network
systemctl restart neutron-openvswitch-agent

Create an external network and a subnet on it

neutron net-create external_network --provider:network_type flat --provider:physical_network extnet  --router:external --shared
neutron subnet-create --name public_subnet --enable_dhcp=False --allocation-pool=start=192.168.100.100,end=192.168.100.150 --gateway=192.168.100.1 external_network 192.168.100.0/24

Create a router and associate external network as router gateway

neutron router-create router1
neutron router-gateway-set router1 external_network

Create an internal network, a subnet and associate an interface to the router

neutron net-create private_network
neutron subnet-create --name private_subnet private_network 10.0.1.0/24
neutron router-interface-add router1 private_subnet

Boot 2 instances

nova boot --flavor m1.tiny --image cirros --nic net-id=154da7a8-fa49-415e-9d35-c840b144a8df test1
nova boot --flavor m1.tiny --image cirros --nic net-id=154da7a8-fa49-415e-9d35-c840b144a8df test2

Create 2 floating ips and associate it to instances

neutron floatingip-create external_network
neutron floatingip-create external_network
nova floating-ip-associate test1 192.168.100.101
nova floating-ip-associate test2 192.168.100.102

Test if all works as expected pinging floating ips

# ping 192.168.100.101
# ping 192.168.100.102

As you can see, in network nodes, a snat namespace is created

# sudo ip netns
qdhcp-154da7a8-fa49-415e-9d35-c840b144a8df
snat-77fef58a-6d0c-4e96-b4b6-5d8e81ebead3

In compute nodes, a fip namespace per instance with floating ip associated running on the compute node are created and a qrouter namespace are created.

# sudo ip netns
fip-4dfdabb0-d2d6-4d4a-8c00-84df834eec8b
qrouter-77fef58a-6d0c-4e96-b4b6-5d8e81ebead3

Best regards, Eduardo Gonzalez