You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

9.6 KiB


Kuryr mascot

Docker for OpenStack Neutron

Kuryr-libnetwork is Kuryr's Docker libnetwork driver that uses Neutron to provide networking services. It provides containerised images for the common Neutron plugins.

This repo provides libnetwork specific functionalities such as handler methods for libnetwork apis. All the logic/utilities that can be shared among different container networking frameworks such as Docker's libnetwork, K8s's CNI and so on, is maintained in separate Kuryr repo as a common library.


  • Docker libnetwork remote driver

  • Docker libnetwork IPAM driver

  • Support for Linux Bridge, Open vSwitch, Midonet, and IOvisor port bindings

  • Support for using existing Neutron networks:

    docker network create -d kuryr --ipam-driver=kuryr --subnet= --gateway= \
       -o mynet
    docker network create -d kuryr --ipam-driver=kuryr --subnet= --gateway= \
       -o mynet
  • Support for using existing Neutron ports:

    docker run -it --net=kuryr_net --ip= ubuntu
    if a port in the corresponding subnet with the requested ip address
    already exists and it is unbound, that port is used for the
  • Support for the Docker "expose" option:

    docker run --net=my_kuryr_net --expose=1234-1238/udp -it ubuntu
    This feature is implemented by using Neutron security groups.


  • Docker 1.12 with SwarmKit (the new Swarm) does not support remote drivers. Therefore, it cannot be used with Kuryr. This limitation is to be removed in Docker 1.13.

Getting it running with a service container


The necessary components for an operating environment to run Kuryr are:

  • Keystone (preferably configured with Keystone v3),
  • Neutron (preferably mitaka or newer),
  • DB management system suh as MySQL or Mariadb (for Neutron and Keystone),
  • Neutron agents for the vendor you choose,
  • Rabbitmq if the Neutron agents for your vendor require it,
  • Docker 1.9+

Building the container

The Dockerfile in the root of this repository can be used to generate a wsgi Kuryr Libnetwork server container with docker build:

docker build -t your_docker_username/libnetwork:latest .

Additionally, you can pull the upstream container:

docker pull kuryr/libnetwork:latest

Note that you can also specify the tag of a stable release for the above command instead of latest.

How to run the container

First we prepare Docker to find the driver:

sudo mkdir -p /usr/lib/docker/plugins/kuryr
sudo curl -o /usr/lib/docker/plugins/kuryr/kuryr.spec \
sudo service docker restart

Then we start the container:

docker run --name kuryr-libnetwork \
  --net=host \
  --cap-add=NET_ADMIN \
  -e SERVICE_USER=admin \
  -e USER_DOMAIN_NAME=Default \
  -v /var/log/kuryr:/var/log/kuryr \
  -v /var/run/openvswitch:/var/run/openvswitch \

Where: * SERVICE_USER, SERVICE_PROJECT_NAME, SERVICE_PASSWORD, SERVICE_DOMAIN_NAME, USER_DOMAIN_NAME are OpenStack credentials * IDENTITY_URL is the url to the OpenStack Keystone v3 endpoint * A volume is created so that the logs are available on the host * NET_ADMIN capabilities are given in order to perform network operations on the host namespace like ovs-vsctl

Other options you can set as '-e' parameters in Docker run: * CAPABILITY_SCOPE can be "local" or "global", the latter being for when there is a cluster store plugged into the docker engine. * LOG_LEVEL for defining, for example, "DEBUG" logging messages. * PROCESSES for defining how many kuryr processes to use to handle the libnetwork requests. * THREADS for defining how many threads per process to use to handle the libnetwork requests.

Note that you will probably have to change the IDENTITY_URL address for the address where your Keystone is running. In this case it is because the example assumes running the container with --net=host on an all in one deployment where Keystone is also binding locally.

Alternatively, if you have an existing kuryr.conf, you can use it for the container:

docker run --name kuryr-libnetwork \
  --net host \
  --cap-add NET_ADMIN \
  -v /etc/kuryr:/etc/kuryr:ro \
  -v /var/log/kuryr:/var/log/kuryr:rw \
  -v /var/run/openvswitch:/var/run/openvswitch:rw \

Getting it from source

$ git clone
$ cd kuryr-libnetwork


$ sudo pip install -r requirements.txt

Installing Kuryr's libnetwork driver

Running the following will grab the requirements and install kuryr:

$ sudo pip install .

Configuring Kuryr

Generate sample config, etc/kuryr.conf.sample, running the following:

$ ./tools/

Rename and copy config file at required path:

$ cp etc/kuryr.conf.sample /etc/kuryr/kuryr.conf

For using Keystone v3, edit the Neutron section in /etc/kuryr/kuryr.conf, replace ADMIN_PASSWORD:

auth_url =
username = admin
user_domain_name = Default
project_name = service
project_domain_name = Default
auth_type = password

Alternatively, for using Keystone v2, edit the Neutron section in /etc/kuryr/kuryr.conf, replace ADMIN_PASSWORD:

auth_url =
username = admin
project_name = service
auth_type = password

In the same file uncomment the bindir parameter with the path for the Kuryr vif binding executables. For example, if you installed it on Debian or Ubuntu:

bindir = /usr/local/libexec/kuryr

Running Kuryr

Currently, Kuryr utilizes a bash script to start the service. Make sure that you have installed tox before the execution of the command below:

$ sudo ./scripts/

After Kuryr starts, please restart your Docker service, e.g.:

$ sudo service docker restart

The bash script creates the following file if it is missing:

  • /usr/lib/docker/plugins/kuryr/kuryr.json: Json spec file for libnetwork.

Note the root privilege is required for creating and deleting the veth pairs with pyroute2 to run.

Testing Kuryr

For a quick check that Kuryr is working, create a network:

$ docker network create --driver kuryr --ipam-driver kuryr \
--subnet test_net
$ docker network ls
NETWORK ID          NAME                DRIVER
785f8c1b5ae4        test_net            kuryr

To test it with tox:

$ tox

You can also run specific test cases using the -e flag, e.g., to only run the fullstack test case:

$ tox -e fullstack

Generating Documentation

We use Sphinx to maintain the documentation. You can install Sphinx using pip:

$ pip install -U Sphinx

In addition to Sphinx you will also need the following requirements (not covered by requirements.txt):

$ pip install oslosphinx reno 'reno[sphinx]'

The source code of the documentation are under doc, you can generate the html files using the following command. If the generation succeeds,a build/html dir will be created under doc:

$ cd doc
$ make html

Now you can serve the documentation at http://localhost:8080 as a simple website:

$ cd build/html
$ python -m SimpleHTTPServer 8080


To create Docker networks with subnets having same/overlapping cidr, it is expected to pass unique pool name for each such network creation Docker command. Docker cli options -o and --ipam-opt should be used to pass pool names as shown below:

$ sudo docker network create --driver=kuryr --ipam-driver=kuryr \
  --subnet --ip-range \
  -o \ \

Now Docker user creates another network with same cidr as the previous one, i.e, but with different pool name, neutron_pool2:

$ sudo docker network create --driver=kuryr --ipam-driver=kuryr \
  --subnet --ip-range \
  -o \ \

External Resources

The latest and most in-depth documentation is available at: