Kapua can be run in a few different ways. See the following chapters for more information about how to start and run Kapua.

Most of the following descriptions focus on a developer centric setup and not on a full blown production setup.

Docker containers

Running Kapua on local docker containers is as easy as:

  1. Install docker
  2. Run mvn -f assembly -Pdocker once
  3. Run the docker images you want to start, see assembly/


OpenShift is a PaaS (Platform As A Service) platform based on Kubernetes. Kapua support deployments into OpenShift Origin, which is an open source community project. Origin sources, can be found here. We support Kapua on OpenShift Origin 1.4.1.

Currently we support running OpenShift only on Linux OS. If you would like to run Kapua on Mac OS or on Windows for development purposes, please install Linux (for example Fedora 25) into virtual machine (for example VirtualBox) and install OpenShift there.

Starting OpenShift cluster

For running Kapua on an OpenShift you need to have OpenShift cluster installed and started in a first place. You can install it by yourself or rely on the script we provides:

sudo kapua/dev-tools/src/main/openshift/

If you are running your OpenShift cluster for a first time, execute the following initialized script as well:


Initialization script is responsible for logging you into a cluster and creating new OpenShift project for Kapua.

If for some reasons, you cannot start your cluster, try to execute the startup script with option DOCKERIZED=FALSE:

sudo DOCKERIZED=FALSE kapua/dev-tools/src/main/openshift/

Option DOCKERIZED=FALSE tells startup script to use standard binary installation of OpenShift Origin instead of Docker-based oc cluster up command.

Deploying Kapua into OpenShift cluster

Now when you have OpenShift cluster up, running and initialized, execute the following script:

cd kapua/dev-tools/src/main/openshift

Now open the following URL in your web browser - http://localhost:8080. And log-in into Kapua UI using default credentials:


Adding metrics

If you have enabled metrics support in OpenShift (e.g. with oc cluster up --metrics) then you can also install Grafana for Hawkular to visualize your data:

oc new-app -f

External access

In order to enable devices to access Kapua we need to allow external access to the broker's MQTT connector. In the default deployment there are two ways to achieve this.

First, the broker exposes MQTT over WebSocket transport. As WebSocket is based on HTTP we can define a router inside the Openshift to get those device connections to the broker. For example, if your Openshift deployment is running at the address, you can connect the Kura Simulator like this

java -jar target/kapua-simulator-kura-*-SNAPSHOT-app.jar --broker ws://

Not all MQTT clients have WebSocket support, so we need to enable direct MQTT over TCP access to the broker as well. By default, Kapua comes with the NodePort service that routes all traffic from port 31883 to the broker. So you can connect your MQTT clients directly to this service. For the simulator example similar to the above, that would look something like

java -jar target/kapua-simulator-kura-0.2.0-SNAPSHOT-app.jar --broker tcp://kapua-broker:kapua-password@

This is suitable only for the local deployments. In the cloud or production environments, you should deploy a proper LoadBalancer Openshift service to enable external traffic flow to the broker.

Ensuring enough entropy

It may happen that firing up docker containers and starting up application which use Java's SecureRandom (which happens in the next step a few times) run dry the Linux Kernel's entropy pool. The result is that some application will block during startup (even longer than 30 seconds) which will trigger OpenShift to kill the pods since they are considered unresponsive (which they actually are).

You can check the amount of entropy the kernel has available with the following command:

 cat /proc/sys/kernel/random/entropy_avail

If this number drops to zero, then the kernel has run out of entropy and application will block.

One solution (there are a few others) is to install haveged a user-space daemon which provides entropy to the kernel.

On CentOS 7 it can be installed with the following commands (all as root):

 yum install epel-release # only if you 
 yum install haveged
 systemctl enable --now haveged

As the package comes from the EPEL repositories. If you haven't yet enabled those repositories, then you need to do this before trying to install haveged:

 yum install epel-release

For more information about haveged see

For more information about the "EPEL repositories" see


Kapua can also be run with Vagrant.

Installing Vagrant

Before Vargant can be used to run Kapua it needs to be installed. This is different on each distribution.

Installing Vagrant on Fedora 25

Do not use Vagrant from Fedora 25. Kapua requires to use the VirtualBox provide from Vargant and cannot run with the libvirt provider which Fedora uses. So it is necessary to install Virtualbox and Vagrant from different locations.

Run the following commands in order to install Vagrant (all as root):

dnf install kernel-devel

rpm --import oracle_vbox.asc

dnf install
dnf install

There may by more up to date versions of the binaries. You should check and install more recent versions, if possible.

By manually installing RPMs you won't receive updates for those packages. You will need to manually check for security updates and bug fixes.

Running Kapua

After Vagrant is installed you can run Kapua by running:

cd dev-tools/src/main/vagrant
sudo ./ base-box

results matching ""

    No results matching ""