ContainerImage.Pinniped/deploy/supervisor

Deploying the Pinniped Supervisor

What is the Pinniped Supervisor?

The Pinniped Supervisor app is a component of the Pinniped OIDC and Cluster Federation solutions. It can be deployed when those features are needed.

Installing the Latest Version with Default Options

kubectl apply -f https://github.com/vmware-tanzu/pinniped/releases/latest/download/install-supervisor.yaml

Installing an Older Version with Default Options

Choose your preferred release version number and use it to replace the version number in the URL below.

# Replace v0.3.0 with your preferred version in the URL below
kubectl apply -f https://github.com/vmware-tanzu/pinniped/releases/download/v0.3.0/install-supervisor.yaml

Installing with Custom Options

Creating your own deployment YAML file requires ytt from Carvel to template the YAML files in the deploy/supervisor directory. Either install ytt or use the container image from Dockerhub.

1. git clone this repo and git checkout the release version tag of the release that you would like to deploy.
2. The configuration options are in deploy/supervisor/values.yml. Fill in the values in that file, or override those values using additional ytt command-line options in the command below. Use the release version tag as the image_tag value.
3. In a terminal, cd to this deploy/supervisor directory
4. To generate the final YAML files, run ytt --file .
5. Deploy the generated YAML using your preferred deployment tool, such as kubectl or kapp. For example: ytt --file . | kapp deploy --yes --app pinniped-supervisor --diff-changes --file -

Configuring After Installing

Exposing the Supervisor App as a Service

The Supervisor app's endpoints should be exposed as HTTPS endpoints with proper TLS certificates signed by a Certificate Authority which will be trusted by your user's web browsers. Because there are many ways to expose TLS services from a Kubernetes cluster, the Supervisor app leaves this up to the user. The most common ways are:

1. Define an Ingress resource with TLS certificates. In this case, the ingress will terminate TLS. Typically, the ingress will then talk plain HTTP to its backend, which would be a NodePort or LoadBalancer Service in front of the HTTP port 80 of the Supervisor pods.

   The required configuration of the Ingress is specific to your cluster's Ingress Controller, so please refer to the documentation from your Kubernetes provider. If you are using a cluster from a cloud provider, then you'll probably want to start with that provider's documentation. For example, if your cluster is a Google GKE cluster, refer to the GKE documentation for Ingress. Otherwise, the Kubernetes documentation provides a list of popular Ingress Controllers, including Contour and many others.

2. Or, define a TCP LoadBalancer Service which is a layer 4 load balancer and does not terminate TLS. In this case, the Supervisor app will need to be configured with TLS certificates and will terminate the TLS connection itself (see the section about OIDCProviderConfig below). The LoadBalancer Service should be configured to use the HTTPS port 443 of the Supervisor pods as its targetPort.

   Warning: Do not expose the Supervisor's port 80 to the public. It would not be secure for the OIDC protocol to use HTTP, because the user's secret OIDC tokens would be transmitted across the network without encryption.

3. Or, expose the Supervisor app using a Kubernetes service mesh technology, e.g. Istio. Please see the documentation for your service mesh. Generally, the setup would be similar to the description above for defining an ingress, expect the service mesh would probably provide both the ingress with TLS termination and the service.

For either of the first two options mentioned above, if you installed using ytt then you can use the related service_* options from deploy/supervisor/values.yml to create a Service. If you installed using install-supervisor.yaml then you can create the Service separately after installing the Supervisor app. There is no Ingress included in the ytt templates, so if you choose to use an Ingress then you'll need to create that separately after installing the Supervisor app.

Example: Using a LoadBalancer Service

This is an example of creating a LoadBalancer Service to expose port 443 of the Supervisor app outside the cluster.

apiVersion: v1
kind: Service
metadata:
  name: pinniped-supervisor-loadbalancer
  # Assuming that this is the namespace where the supervisor was installed. This is the default in install-supervisor.yaml.
  namespace: pinniped-supervisor
spec:
  type: LoadBalancer
  selector:
    # Assuming that this is how the supervisor pods are labeled. This is the default in install-supervisor.yaml.
    app: pinniped-supervisor
  ports:
  - protocol: TCP
    port: 443
    targetPort: 443

Example: Using a NodePort Service

A NodePort Service exposes the app as a port on the nodes of the cluster.

This is convenient for use with kind clusters, because kind can expose node ports as localhost ports on the host machine without requiring an Ingress, although kind also supports several Ingress Controllers.

A NodePort Service could also be used behind an Ingress which is terminating TLS.

For example:

apiVersion: v1
kind: Service
metadata:
  name: pinniped-supervisor-nodeport
  # Assuming that this is the namespace where the supervisor was installed. This is the default in install-supervisor.yaml.
  namespace: pinniped-supervisor
spec:
  type: NodePort
  selector:
    # Assuming that this is how the supervisor pods are labeled. This is the default in install-supervisor.yaml.
    app: pinniped-supervisor
  ports:
  - protocol: TCP
    port: 80
    targetPort: 80
    nodePort: 31234 # This is the port that you would forward to the kind host. Or omit this key for a random port.

Configuring the Supervisor to Act as an OIDC Provider

The Supervisor can be configured as an OIDC provider by creating OIDCProviderConfig resources in the same namespace where the Supervisor app was installed. For example:

apiVersion: config.pinniped.dev/v1alpha1
kind: OIDCProviderConfig
metadata:
  name: my-provider
  # Assuming that this is the namespace where the supervisor was installed. This is the default in install-supervisor.yaml.
  namespace: pinniped-supervisor
spec:
  # The hostname would typically match the DNS name of the public ingress or load balancer for the cluster.
  # Any path can be specified, which allows a single hostname to have multiple different issuers. The path is optional.
  issuer: https://my-issuer.example.com/any/path

If you are using a LoadBalancer Service to expose the Supervisor app outside your cluster, then you will also need to configure the OIDCProviderConfig with TLS certificates, so the app can terminate TLS. You can create the certificates however you like, for example you could use cert-manager. Keep in mind that your users will load some of these endpoints in their web browsers, so the TLS certificates should be signed by a Certificate Authority that will be trusted by those browsers.

If you have terminated TLS outside the app, for example using an Ingress with TLS certificates, then you do not need to configure TLS certificates on the OIDCProviderConfig.