OpenKilda is an open-source OpenFlow controller initially designed for use in a global network with high control-plane latency and a heavy emphasis on latency-centric data path optimisation.
Branch: develop
Clone or download
nikitamarchenko Merge pull request #2046 from telstra/functional/isl_delete_test
Verify ability to rediscover deleted ISLs of different types
Latest commit a9dd401 Feb 21, 2019
Type Name Latest commit message Commit time
Failed to load latest commit information.
base Merge branch 'develop' into feat/virtual-lab Nov 21, 2018
confd Merge pull request #2044 from telstra/enhancement-dynamic-appname-in-… Feb 18, 2019
docs Design for synchronization switch rules using H&S approach Feb 7, 2019
kilda-bins Add md5 files for downloadable files Oct 8, 2017
services Merge pull request #2046 from telstra/functional/isl_delete_test Feb 21, 2019
tools/elk-dashboards Updated Kilda-CrudBolt-flow-filtered-table and Kilda-OFELinkBolt-view… Nov 19, 2018
.editorconfig Merge smoke and atdd src trees; refactor smoke and perf May 3, 2017
.gitignore Refactor packages. Add logger config Oct 19, 2018
.gitmodules Change kilda-bins download mechanism Oct 8, 2017
.travis.yml Fixed unit tests and sonar stage Dec 21, 2018 Add design documentation requirements in Sep 11, 2018
Jenkinsfile Updates to API Mar 2, 2017
Jenkinsfile.local Updates to API Mar 2, 2017
LICENSE.txt Add CLA and LICENSE Oct 6, 2017
Makefile Make 'make' command for func-tests to be self-sufficient, update readme Feb 12, 2019 Expand setup docs. Feb 6, 2019
Vagrantfile Finalized 17.04.02 of kilda/kildadev vagrant box Aug 23, 2017
What is OpenKilda.pdf Initial launch deck of OpenKilda Oct 11, 2017
bitbucket-pipelines.yml Add dev* branch target to pipeline builds Jul 27, 2017
bitbucket-python-pipeline-requirements.txt added mvn and docker-compose Jun 14, 2017
config.yml actualization of ansible templates with newly created kafka topics Dec 20, 2017
docker-compose.yml Eliminate mininet and TER Jan 17, 2019


Build StatusQuality Gate


Note that the build process will install additional packages. It is recommended that you build on a virtual machine.


The followings are required for building Kilda controller:

  • Maven 3.3.9+
  • JDK8
  • Python 2.7+
  • Python 3.5+
  • Docker Compose 1.20.0+
  • GNU Make 4.1+

On Ubuntu 16.04, you can install those dependancies like this:

apt-get install maven openjdk-8-jdk python python3 docker-compose virtualenv make

Note that your build user needs to be a member of the docker group for the build to work. Do that by adding the user to /etc/groups and logging out and back in again.

You also need to increase the maven RAM limit at least up to 1G.

export MAVEN_OPTS="-Xmx1g -XX:MaxPermSize=128m"

Some build steps require Python2. If Python3 is default in your system please use virtualenv. Ensure that you have tox installed:

virtualenv --python=python2 .venv
. .venv/bin/activate
pip install -U pip
pip install tox

Also, don't forget to install confd. This tool is used for creating config/properties files from templates. To install it execute the following command:

wget -O /usr/local/bin/confd
chmod +x /usr/local/bin/confd

How to build Kilda Controller

From the base directory run the following command:

make build-latest

Note that additional Ubuntu packages will be installed as part of the build process.

How to clean Kilda Controller

From the base directory run the following command:

make clean

How to run Kilda Controller

NB: To run Kilda, you should have built it already (ie the previous section). This is particularly important because docker-compose will expect some of the containers to already exist.

From the base directory run these commands:

  1. docker-compose up


  1. make up-test-mode

How to debug Kilda Controller components

An important aspect of troubleshooting errors and problems in your code is to avoid them in the first place. It's not always easy enough so we should have a reliable mechanism. Adding any diagnostic code may be helpful, but there are more convenient ways. Just a few configuration changes and you'll be able to use a debug toolkit. As a basis, let's take the northbound component. This is a simple REST application providing the interface for interaction with the switch, link, flow, feature, health-check controllers. The first thing that we need to do is to add


to the CMD block in services/src/northbound/Dockerfile, where 50505 is the port we’ll use for debugging. It can be any port, it’s up to us. The final file will be the following:

FROM kilda/base-ubuntu
ADD target/northbound.jar /app/
CMD ["java","-agentlib:jdwp=transport=dt_socket,address=50505,suspend=n,server=y","-jar","northbound.jar"]

Since debugging is done over the network, that also means we need to expose that port in Docker. For that purpose we need to add "50505:50505" to the northbound ports block in docker-compose.yml as in example below.

  container_name: openkilda-northbound
    context: services/src/northbound/
    - '8088:8080'
    - '50505:50505'

After making those changes we need to configure remote debug in IntelliJ Idea: navigate to Edit Configurations -> Remote and set up the debug port as 50505. This completes the main part of the configuration.

Next, we just run docker-compose up. If everything above was done correctly you must see:

"java -agentlib:jdwp=transport=dt_socket,address=50505,suspend=n,server=y -jar northbound.jar"

in the command column for the open-kilda_northbound. The command docker ps -a --no-trunc | grep northbound could be helpful. Also check open-kilda_northbound logs, the log record

Listening for transport dt_socket at address: 50505

must be presented.

After all these steps you just need to run the debugger. Console log should contain the following message:

Connected to the target VM, address: 'localhost:50505', transport: 'socket'

To check how debugging works we need to:

  • set up a breakpoint;
  • make a call to execute some functionality;

In some cases, we must have an approach for debugging a deploy process for a couple (or more) components that interact with each other. Let's suppose both of them work under docker and some component doesn't belong to us and provided as a library. The typical case: WorkflowManager (further WFM) and Storm. The approach that is going to be used is almost the same as for northbound but there are nuances. First of all, we need to check which version of Storm is used in Open Kilda Controller. For that open services/storm/Dockerfile and find the version of Storm. In our case, the Storm version is 1.1.0. To be able to debug Storm we have to clone the sources from the GitHub repo and switch to the release 1.1.0. git checkout -b 1.1.0 e40d213. Information about releases can be found here

Then go to services/wfm/Dockerfile and add ENV STORM_JAR_JVM_OPTS "-agentlib:jdwp=transport=dt_socket,server=y,suspend=n,address=50506" The final file should be as in example below:

FROM kilda/storm:latest

ENV STORM_JAR_JVM_OPTS "-agentlib:jdwp=transport=dt_socket,server=y,suspend=n,address=50506"

And it only remains to add a port 50506 for the WFM contaner as in the example below:

  container_name: wfm
    context: services/wfm
    - "50506:50506"

Then we should configure remote debugging in IntelliJ Idea and the set up the debug port as 50506. After executing docker-compose up you should see the following log record Listening for transport dt_socket at address: 50506 in the WFM logs. As soon as you see it run the debugger - you'll able to debug both components: WFM and Storm.

In order to debug a topology, for example, OfEventWfmTopology, we should navigate to Maven Projects, Profiles and toggle local checkbox. Then open OfEventWfmTopology application debug configuration and add --local to Program arguments, execute docker-compose up and run in the debug mode OfEventWfmTopology.

How to run tests

Please refer to the Testing section on our Wiki.

How to run floodlight-modules locally

From the base directory run these commands:

  1. make build-floodlight
  2. make run-floodlight

How to build / test locally without containers

Start with the following

  1. make unit

From there, you can go to specific projects to build / develop / unit test. Just follow the make unit trail. Most projects have a maven target.

NB: Several projects have a dependency on the maven parent; look at make unit

How to build / test key use cases

Look in the base/hacks/usecase directory and you'll find several makefiles that will assist with the development and testing of that use case.

As an example, you can execute the following command for more information on the network discovery use case:

make -f base/hacks/usecase/network.disco.make help

# or

cd base/hacks/usecase
make -f network.disco.make help

How to use a VM to do development

VirtualBox and Vagrant are popular options for creating VMs. A VM may be your best bet for doing development with Kilda. There are a set of files in the source tree that will facilitate this.

  • NB1: Ensure you have VirtualBox and Vagrant installed and on the path
  • NB2: You'll re-clone kilda so that there aren't any write permission issues between the guest VM and the host.


  1. From the root directory, look at the Vagrantfile; feel free to change its parameters.
  2. vagrant up - create the VM; it'll be running after this step.
  3. vagrant ssh - this will log you into the vm.
  4. ssh-keygen -t rsa -C "" - you'll use this for GitHub. Press for each question; three in total.
  5. Add the ~/.ssh/ key to your GitHub account so that you can clone kilda
cat ~/.ssh/
  1. Clone and Build
# NB: Instead of putting it in vm-dev, you can use /vagrant/vm-dev
#     This has the added benefit that the code will appear outside the VM
#     i.e. /vagrant is shared with the same directory as the Vagrantfile
git clone<your_github_account>/open-kilda.git vm-dev
cd vm-dev
git checkout mvp1rc
make build-base
docker-compose build
make unit
make up-test-mode

How to use confd for config/properties templating

Pre-requirements: you need confd version v0.14.0+ for processing yams/json as backend. You can download it from official confd site

We have confd for managing config/properties files from templates. Confd configs, templates and variable file stored in confd/ folder. confd/conf.d/*.toml - files with desctiption how to process templates (src. path, dst.path.... etc) confd/templates/*/*.tmpl - templates in go-template format confd/vars/main.yaml - file with all variables substituted in templates

How should I add new template

  1. create and place template file to confd/templates/ folder
  2. create and place template description in confd/conf.d/ filder
  3. change (if needed) vars in confd/main.yaml
  4. run: make update-props-dryrun for checking that templates can be processed
  5. run: make update-props for applying templates

Common use cases

An exmaplte, you already have neo4j server, and want to use it instead of dockerized neo4j. You can add neo4j endpoints to confd/vars/main.yaml and create properties template for services which use neo4j:


src = "topology-engine/topology_engine.ini.tmpl"
dest = "services/topology-engine/queue-engine/topology_engine.ini"
keys = [ "/" ]
mode = "0644"


kilda_neo4j_host: "neo4j"
kilda_neo4j_user: "neo4j"
kilda_neo4j_password: "temppass"


[kafka]{{ getv "/kilda_kafka_te_consumer_group" }}
flow.topic={{ getv "/kilda_kafka_topic_flow" }}
speaker.topic={{ getv "/kilda_kafka_topic_speaker" }}
topo.eng.topic={{ getv "/kilda_kafka_topic_topo_eng" }}
northbound.topic={{ getv "/kilda_kafka_topic_northbound" }}
bootstrap.servers={{ getv "/kilda_kafka_hosts" }}
environment.naming.prefix = {{ getv "/kilda_environment_naming_prefix" }}

worker.pool.size={{ getv "/kilda_worker_pool_size" }}

hosts={{ getv "/kilda_zookeeper_hosts" }}

host={{ getv "/kilda_neo4j_host" }}
user={{ getv "/kilda_neo4j_user" }}
pass={{ getv "/kilda_neo4j_password" }}

In this example we will generate file services/topology-engine/queue-engine/ from template confd/templates/topology-engine/topology_engine.ini.tmpl

How enable travis CI

Someone with admin rights should log in using github account to and on the page activate telstra/open-kilda repository. All configurations for travis are located in .travis.yml. For adding new scripts you should create new line under script parameter.

- make run-test
- new command