Lingxian Kong, senior developer at Catalyst Cloud, offers this guide.

This guide explains how to deploy and configure the octavia-ingress-controller in Kubernetes cluster on top of an OpenStack cloud.

What’s an Ingress Controller, anyway?

In Kubernetes, Ingress allows external users and client applications access to HTTP services. Ingress consists of two components:

  • Ingress Resource is a collection of rules for the inbound traffic to reach Services. These are Layer 7 (L7) rules that allow hostnames (and optionally paths) to be directed to specific Services in Kubernetes.
  • Ingress Controller acts on the rules set by the Ingress Resource, typically via an HTTP or L7 load balancer.

It’s vital that both pieces are properly configured to route traffic from an outside client to a Kubernetes Service.

1. Benefits of using the octavia-ingress-controller

As an OpenStack-based public cloud provider, one of our goals at Catalyst Cloud is to continuously provide customers with the capability to innovate by delivering robust and comprehensive cloud services. After deploying Octavia and Magnum services in the public cloud, we’re now thinking about how to help customers to develop their applications running on the Kubernetes cluster and making their services accessible to the public in a high-performance way.

After creating a Kubernetes cluster in Magnum, the most common way to expose the application to the outside world is to use a LoadBalancer service. In the OpenStack cloud, Octavia (LBaaS v2) is the default implementation of LoadBalancer type service, as a result, for each LoadBalancer type service, there is a load balancer created in the cloud tenant account. There are a few drawbacks of this approach:

  • The cost of Kubernetes Service is a little bit high if it’s one-to-one mapping from the service to Octavia load balancer and the customers have to pay for a load balancer per exposed service, which can get expensive.
  • There is no filtering, no routing, etc. for the service. That means you can send almost any kind of traffic to it, like HTTP, TCP, UDP, WebSockets, gRPC, or whatever.
  • The traditional ingress controllers such as the NGINX ingress controller, HAProxy ingress controller, Træfik, etc. don’t make much sense in the cloud environment because the user still needs to expose the service for the ingress controller itself which may increase the network delay and decrease the performance of the application.

The octavia-ingress-controller can solve all of the above problems in an OpenStack environment by creating a single load balancer for multiple NodePort type services in an Ingress. In order to use the octavia-ingress-controller in Kubernetes cluster, use the value openstack for the annotation in the metadata section of the Ingress Resource as shown below:

annotations: openstack

2. How to deploy the octavia-ingress-controller

2.1 Prepare the kubeconfig file

The kubeconfig file is used to configure access to Kubernetes clusters. This is a generic way of referring to configuration files in Kubernetes. The following commands are performed in a Kubernetes cluster created using kubeadm.

  • Install cfssl tools, which are used for generating TLS certs
wget && chmod +x cfssl_linux-amd64 && mv cfssl_linux-amd64 /usr/local/bin/cfssl
wget && chmod +x cfssljson_linux-amd64 && mv cfssljson_linux-amd64 /usr/local/bin/cfssljson
wget && chmod +x cfssl-certinfo_linux-amd64 && mv cfssl-certinfo_linux-amd64 /usr/local/bin/cfssl-certinfo
  • Re-use the CA cert and key in the existing cluster
pushd /etc/kubernetes/pki
cat > ca-config.json <<EOF
  "signing": {
    "default": {
      "expiry": "87600h"
    "profiles": {
      "kubernetes": {
        "usages": [
            "key encipherment",
            "server auth",
            "client auth"
        "expiry": "87600h"
cat > ingress-openstack-csr.json <<EOF
  "CN": "octavia-ingress-controller",
  "hosts": [],
  "key": {
    "algo": "rsa",
    "size": 2048
  "names": [
      "C": "NZ",
      "ST": "Wellington",
      "L": "Wellington",
      "O": "Catalyst",
      "OU": "Lingxian"
cfssl gencert -ca=ca.crt -ca-key=ca.key -config=ca-config.json -profile=kubernetes ingress-openstack-csr.json | cfssljson -bare ingress-openstack
# You can take a look at the files generated by cfssl
ls -l | grep ingress-openstack
  • Create a kubeconfig file for the octavia-ingress-controller
ca_data=$(cat ca.crt | base64 | tr -d '\n')
client_cert_data=$(cat ingress-openstack.pem | base64 | tr -d '\n')
client_key_data=$(cat ingress-openstack-key.pem | base64 | tr -d '\n')
cat <<EOF > /etc/kubernetes/ingress-openstack.conf
apiVersion: v1
kind: Config
- cluster:
    certificate-authority-data: ${ca_data}
    server: https://${k8s_master_ip}:6443
  name: kubernetes
preferences: {}
- name: octavia-ingress-controller
    client-certificate-data: ${client_cert_data}
    client-key-data: ${client_key_data}
- context:
    cluster: kubernetes
    user: octavia-ingress-controller
  name: [email protected]
current-context: [email protected]

2.2 Configure role-based access control (RBAC) for the octavia-ingress-controller user

For testing purpose, grant the cluster-admin role for octavia-ingress-controller as user so that the user has full access to the Kubernetes cluster.

cat <<EOF | kc create -f -
kind: ClusterRoleBinding
  name: octavia-ingress-controller
  - kind: User
    name: octavia-ingress-controller
  kind: ClusterRole
  name: cluster-admin

2.3 Prepare the octavia-ingress-controller service configuration

We need credentials of admin user and a normal user (e.g. demo) in OpenStack.

source openrc_admin
project_id=$(openstack project show demo -c id -f value)
auth_url=$(export | grep OS_AUTH_URL | awk -F'"' '{print $2}')
subnet_id=$(openstack subnet show private-subnet -c id -f value)
public_net_id=$(openstack network show public -c id -f value)
cat <<EOF > /etc/kubernetes/ingress-openstack.yaml
    kubeconfig: /etc/kubernetes/ingress-openstack.conf
    username: demo
    password: password
    project_id: ${project_id}
    auth_url: ${auth_url}/v3
    region: RegionOne
    subnet_id: ${subnet_id}
    fip_network: ${public_net_id}

2.4 Setting up the octavia-ingress-controller service

cat <<EOF > /etc/kubernetes/manifests/octavia-ingress-controller.yaml
apiVersion: v1
kind: Pod
  annotations: ""
    component: octavia-ingress-controller
    tier: control-plane
  name: octavia-ingress-controller
  namespace: kube-system
    - name: octavia-ingress-controller
      image: lingxiankong/openstack-ingress-controller:0.0.2
      imagePullPolicy: Always
        - /bin/octavia-ingress-controller
        - --config=/etc/kubernetes/ingress-openstack.yaml
      - mountPath: /etc/kubernetes/ingress-openstack.yaml
        name: ingressconfig
        readOnly: true
      - mountPath: /etc/kubernetes/ingress-openstack.conf
        name: kubeconfig
        readOnly: true
          cpu: 200m
  hostNetwork: true
  - hostPath:
      path: /etc/kubernetes/ingress-openstack.yaml
      type: FileOrCreate
    name: ingressconfig
  - hostPath:
      path: /etc/kubernetes/ingress-openstack.conf
      type: FileOrCreate
    name: kubeconfig
status: {}

Wait until the octavia-ingress-controller static pod is up and running.

$ kubectl get pod --all-namespaces | grep octavia-ingress-controller
kube-system   octavia-ingress-controller-lingxian-k8s-master   1/1       Running   0          1m

3. Setting up HTTP load balancing with Ingress

3.1 Create a backend service

Create a simple service (echo hostname) that listens on a HTTP server on port 8080.

$ cat <<EOF | kc create -f -
apiVersion: apps/v1
kind: Deployment
  name: hostname-echo-deployment
  replicas: 1
      app: hostname-echo
        app: hostname-echo
        - image: "lingxiankong/alpine-test"
          imagePullPolicy: Always
          name: hostname-echo-container
            - containerPort: 8080
apiVersion: v1
kind: Service
  name: hostname-echo-svc
     -  port: 8080
        protocol: TCP
        targetPort: 8080
    app: hostname-echo
  type: NodePort
$ kubectl get svc
NAME                TYPE        CLUSTER-IP     EXTERNAL-IP   PORT(S)          AGE
hostname-echo-svc   NodePort   <none>        8080:32066/TCP   33s

When you create a Service of type NodePort, Kubernetes makes your Service available on a randomly selected high port number (e.g. 32066) on all the nodes in your cluster. Generally, the Kubernetes nodes are not externally accessible by default, so creating this Service does not make your application accessible from the Internet. However, we can verify the service using its CLUSTER-IP on Kubernetes master node:

$ curl

To make your HTTP web server application publicly accessible, you need to create an Ingress resource.

3.2 Create an Ingress resource

The following command defines an Ingress resource that directs traffic that requests to the hostname-echo Service:

cat <<EOF | kc create -f -
apiVersion: extensions/v1beta1
kind: Ingress
  name: test-octavia-ingress
  annotations: "openstack"
  - host:
      - path: /hostname
          serviceName: hostname-echo-svc
          servicePort: 8080

Kubernetes creates an Ingress resource on your cluster. The octavia-ingress-controller service running in your cluster is responsible for creating/maintaining the corresponding resources in Octavia to route all external HTTP traffic (on port 80) to the hostname-echo NodePort Service you exposed.

Verify that the Ingress Resource has been created. Please note that the IP address for the Ingress Resource will not be defined right away (wait a few moments for the ADDRESS field to get populated):

$ kubectl get ing
NAME                   HOSTS            ADDRESS   PORTS     AGE
test-octavia-ingress             80        12s
$ # Wait until the ingress gets an IP address
$ kubectl get ing
NAME                   HOSTS            ADDRESS      PORTS     AGE
test-octavia-ingress   80        9m

For testing purposes, log in to a host that has a network connection with the OpenStack cloud. To do that, you need to update /etc/hosts file in the host to resolve to the Ingress IP address, then you should be able to access the backend service by sending HTTP request to the domain name specified in the Ingress Resource:

$ echo "" >> /etc/hosts
$ curl

Live demo

You can also see a live demo here or below.

Author Lingxian Kong will be heading up four sessions at the Vancouver Summit, check them out here.