Skip to main content

How to Use the OVHcloud Managed Kubernetes Load Balancer

Michael M.
Updated

In this tutorial, we are explaining how to deploy services on OVHcloud Managed Kubernetes using our LoadBalancer to get external traffic into your cluster. We will begin by listing the main methods to expose Kubernetes services outside the cluster, with their advantages and disadvantages. Then we will see a complete example of LoadBalancer service deployment.

Before you begin

This tutorial assumes that you already have a working OVHcloud Managed Kubernetes cluster, and some basic knowledge of how to operate it. If you want to know more on those topics, please look at the OVHcloud Managed Kubernetes Service Quickstart.

Warning: When a Load Balancer Service resource is created inside a Managed Kubernetes cluster, a Load Balancer for a Managed Kubernetes Service is automatically created. The Load Balancer for a Managed Kubernetes Service is hourly charged and will appear in your Public Cloud project. For more information, please refer to this page.

Some concepts: Cluster IP, Node Port, Ingress and Load Balancer

When you begin to use Kubernetes for real applications, one of the first questions is how to get external traffic into your cluster. The official doc gives you a good, but rather dry, explanation on the topic, but here we are trying to explain the concepts in a minimal, need-to-know way.

There are several ways to route the external traffic into your cluster:

  • Using Kubernetes proxy and ClusterIP: The default Kubernetes ServiceType is ClusterIP, which exposes the Service on a cluster-internal IP. To reach the ClusterIP from an external source, you can open a Kubernetes proxy between the external source and the cluster. It is usually only used for development.

  • Exposing services as NodePort: Declaring a Service of type NodePort exposes the service on each Node’s IP at a static port (the NodePort). You can then access the Service from the outside of the cluster by requesting <NodeIp>:<NodePort>. It can be used for production, with some limitations.

  • Exposing services as LoadBalancer: Declaring a Service of type LoadBalancer exposes it externally using a cloud provider’s load balancer. The cloud provider will provision a load balancer for the Service, and map it to its automatically assigned NodePort. It is the most widely used method in production environments.

Using Kubernetes proxy and Cluster IP

The default Kubernetes ServiceType is ClusterIP, that exposes the Service on a cluster-internal IP. To reach the ClusterIP from an external computer, you can open a Kubernetes proxy between the external computer and the cluster.

You can use kubectl to create such a proxy. When the proxy is up, you’re directly connected to the cluster, and you can use the Services internal IP (Cluster IP).

lb1.jpeg

This method isn’t suited for a production environment, but it’s interesting for development, debugging, or other quick-and-dirty operations.

Exposing services as Node Port

Declaring a service of type NodePort exposes the Service on each Node’s IP at a static port, the NodePort (a fixed port for that Service, in the default range of 30000-32767). You can then access the Service from the outside of the cluster by requesting <NodeIp>:<NodePort>. Every service you deploy as NodePort will be exposed in its own port, on every Node.

lb2.jpeg

It’s rather cumbersome to use NodePort Services in production. As you are using non-standard ports, you often need to set up an external load balancer that listens on standard ports and redirects the traffic to the <NodeIp>:<NodePort>.

Warning: In our OVHcloud Managed Kubernetes you have an easy way to access NodePort services. You need to get the nodes URL, a URL resolving via round-robin DNS to one random node of your cluster. Since NodePort services are exposed in the same port on every Node, you can use this nodes URL to access them.
In order to get the nodes URL, you get the control plane URL (the one given on kubectl cluster-info) and add the nodes element between the first and the second element of the URL
Example:
$ kubectl cluster-info Kubernetes control plane is running at https://xxxxxx.c1.gra9.k8s.ovh.net CoreDNS is running at https://xxxxxxx.c1.gra9.k8s.ovh.net/api/v1/namespaces/kube-system/services/kube-dns:dns/proxy Metrics-server is running at https://xxxxxx.c1.gra9.k8s.ovh.net/api/v1/namespaces/kube-system/services/https:metrics-server:/proxy
In this case, the nodes URL will be https://xxxxxx.nodes.c1.gra9.k8s.ovh.net and a service deployed on NodePort 30123 can be accessed on https://xxxxxx.nodes.c1.gra9.k8s.ovh.net:30123.

Exposing services as Load Balancer

Declaring a LoadBalancer service exposes it externally using a cloud provider’s load balancer. The cloud provider will provision a load balancer for the Service, and map it to its automatically assigned NodePort. How the traffic from that external load balancer is routed to the Service pods depends on the cluster provider.

lb3.jpeg

The LoadBalancer is the best option for a production environment, with two caveats:

  • Every Service of type LoadBalancer you deploy will get its own IP.
  • The LoadBalancer is usually billed by the number of exposed services and can, therefore, get expensive.

Note: There is a limit of 200 active LoadBalancer per OpenStack project (also named OpenStack tenant). This limit can be raised by contacting OVHcloud US support.

Supported annotations

There are several annotations available to customize your load balancer:

  • service.beta.kubernetes.io/ovh-loadbalancer-proxy-protocol: Used on the service to enable the proxy protocol on all backends. Supported values: v1, v2, v2_ssl, v2_ssl_cn.

OVHcloud Load Balancer services handle four Proxy Protocol modes:

lb4.png

  • service.beta.kubernetes.io/ovh-loadbalancer-allowed-sources: Used on the service to specify allowed client IP source ranges. Value: comma separated list of CIDRs. For example: 10.0.0.0/24,172.10.0.1. Deprecated, please use loadBalancerSourceRanges spec instead, see Restrict Access For LoadBalancer Service.

  • service.beta.kubernetes.io/ovh-loadbalancer-balance: Used on the service to set the algorithm to use for load balancing. Supported values: first, leastconn, roundrobin, source. Default: roundrobin.

What about Ingress

According to the official documentation, an Ingress is an API object that manages external access to the services in a cluster, typically HTTP. What is the difference between the LoadBalancer or NodePort?

Ingress isn’t a type of Service, but an object that acts as a reverse proxy, and single entry point to your cluster that routes the request to the different services. The most basic Ingress is the NGINX Ingress Controller, where the NGINX take the role of reverse proxy, but also functions as SSL.

An Ingress is exposed to the outside of the cluster either via ClusterIP and Kubernetes proxy, NodePort, or LoadBalancer, and it routes incoming traffic according to configured rules.

The main advantage of using an Ingress behind a LoadBalancer is the cost: you can have lots of services behind a single LoadBalancer.

Deploying Load Balancer Services on OVHcloud Managed Kubernetes Clusters

In our OVHcloud Managed Kubernetes we propose a load balancing service enabling you to use LoadBalancer ServiceType.

Deploying a Hello World Load Balancer service

Create a hello.yml file for our ovhplatform/hello Docker image, defining the service type as LoadBalancer:

apiVersion: v1
kind: Service
metadata:
  name: hello-world
  labels:
    app: hello-world
spec:
  type: LoadBalancer
  ports:
  - port: 80
    targetPort: 80
    protocol: TCP
    name: http
  selector:
    app: hello-world
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: hello-world-deployment
  labels:
    app: hello-world
spec:
  replicas: 1
  selector:
    matchLabels:
      app: hello-world
  template:
    metadata:
      labels:
        app: hello-world
    spec:
      containers:
      - name: hello-world
        image: ovhplatform/hello:1.1
        ports:
        - containerPort: 80

Apply the file:

kubectl apply -f hello.yml

After applying the YAML file, a new hello-world service and the corresponding hello-world-deployment deployment are created:

$ kubectl apply -f hello.yml

service/hello-world unchanged
deployment.apps/hello-world-deployment configured

Note: The application you have just deployed is a simple Nginx server with a single static Hello World page. Basically, it deploys the Docker image ovhplatform/hello

List the services

Now you are going to use kubectl to see your service:

kubectl get service hello-world -w

You should see your newly created service:

$ kubectl get services
NAME          TYPE           CLUSTER-IP    EXTERNAL-IP     PORT(S)            AGE
hello-world   LoadBalancer   10.3.81.234          80:31699/TCP       9s

Since the LoadBalancer creation is asynchronous, and the provisioning of the load balancer can take several minutes, you will get a <pending> EXTERNAL-IP.

If you try again in a few minutes, you should get an EXTERNAL-IP:

$ kubectl get service hello-world
NAME          TYPE           CLUSTER-IP    EXTERNAL-IP       PORT(S)        AGE
hello-world   LoadBalancer   10.3.81.234   xxx.xxx.xxx.xxx   80:31699/TCP   4m

For each service you deploy with LoadBalancer type, you will get a new IPv4 with the xxx.xxx.xxx.xxx format to access the service.

Testing your service

If you point your web browser to the EXTERNAL-IP value, the hello-world service will answer you:

lb5.png

Cleaning up

At the end, you can proceed to clean up by deleting the service and the deployment.

Let’s begin by deleting the service:

kubectl delete service hello-world

If you list the services, you will see that hello-world doesn’t exist anymore:

$ kubectl delete service hello-world
service "hello-world" deleted

$ kubectl get services -l app=hello-world
No resources found.

Then you can delete the deployment:

kubectl delete deploy hello-world-deployment

Now, if you list your deployment, you will find no resources:

$ kubectl delete deploy hello-world-deployment
deployment.apps "hello-world-deployment" deleted

$ kubectl get deployments -l app=hello-world
No resources found.

Now, list the pods:

kubectl get pod -n default -l app=hello-world

You will see that the pod created for hello-world has been deleted too:

$ kubectl get pod -l app=hello-world
No resources found

Related articles