This tutorial requires Kubernetes 1.14 or later. To run the tutorial locally, we recommend using in version with Kubernetes 1.14+.

2. Create ConfigMap containing configuration for Envoy

The Envoy configuration below defines an external authorization filter envoy.ext_authz for a gRPC authorization server.

Save the configuration as envoy.yaml:

static_resources:
  listeners:
  - address:
      socket_address:
        address: 0.0.0.0
        port_value: 8000
    filter_chains:
    - filters:
      - name: envoy.http_connection_manager
        typed_config:
          "@type": type.googleapis.com/envoy.extensions.filters.network.http_connection_manager.v3.HttpConnectionManager
          codec_type: auto
          stat_prefix: ingress_http
          route_config:
            name: local_route
            virtual_hosts:
            - name: backend
              domains:
              - "*"
              routes:
              - match:
                  prefix: "/"
                route:
                  cluster: service
          http_filters:
          - name: envoy.ext_authz
            typed_config:
              "@type": type.googleapis.com/envoy.extensions.filters.http.ext_authz.v3.ExtAuthz
              transport_api_version: V3
              with_request_body:
                max_request_bytes: 8192
                allow_partial_message: true
              failure_mode_allow: false
              grpc_service:
                google_grpc:
                  target_uri: 127.0.0.1:9191
                  stat_prefix: ext_authz
                timeout: 0.5s
          - name: envoy.filters.http.router
  clusters:
  - name: service
    connect_timeout: 0.25s
    type: strict_dns
    lb_policy: round_robin
    load_assignment:
      cluster_name: service
      endpoints:
      - lb_endpoints:
        - endpoint:
            address:
              socket_address:
                address: 127.0.0.1
                port_value: 8080
admin:
  access_log_path: "/dev/null"
  address:
    socket_address:
      address: 0.0.0.0
      port_value: 8001
layered_runtime:
  layers:
    - name: static_layer_0
      static_layer:
        envoy:
          resource_limits:
            listener:
              example_listener_name:
                connection_limit: 10000
          global_downstream_max_connections: 50000

Create the ConfigMap:

kubectl create configmap proxy-config --from-file envoy.yaml

The following OPA policy restricts access to the /people endpoint exposed by our sample app:

• Alice is granted a guest role and can perform a GET request to /people.
• Bob is granted an admin role and can perform a GET and POST request to /people.

The policy also restricts an admin user, in this case bob from creating an employee with the same firstname as himself.

policy.rego

package envoy.authz
import input.attributes.request.http as http_request
default allow = false
allow {
    is_token_valid
    action_allowed
}
is_token_valid {
  token.valid
  now := time.now_ns() / 1000000000
  token.payload.nbf <= now
  now < token.payload.exp
}
action_allowed {
  http_request.method == "GET"
  token.payload.role == "guest"
  glob.match("/people", ["/"], http_request.path)
}
action_allowed {
  http_request.method == "GET"
  token.payload.role == "admin"
  glob.match("/people", ["/"], http_request.path)
}
action_allowed {
  http_request.method == "POST"
  token.payload.role == "admin"
  glob.match("/people", ["/"], http_request.path)
  lower(input.parsed_body.firstname) != base64url.decode(token.payload.sub)
}
token := {"valid": valid, "payload": payload} {
    [_, encoded] := split(http_request.headers.authorization, " ")
    [valid, _, payload] := io.jwt.decode_verify(encoded, {"secret": "secret"})
}

Store the policy in Kubernetes as a Secret.

kubectl create secret generic opa-policy --from-file policy.rego

In the next step, OPA is configured to query for the data.envoy.authz.allow decision. If the response is true the operation is allowed, otherwise the operation is denied. Sample input received by OPA is shown below:

data.envoy.authz.allow

false

An example of the complete input received by OPA can be seen .

4. Create App Deployment with OPA and Envoy sidecars

Our deployment contains a sample Go app which provides information about employees in a company. It exposes a /people endpoint to get and create employees. More information can on the app be found .

OPA is started with a configuration that sets the listening address of Envoy External Authorization gRPC server and specifies the name of the policy decision to query. More information on the configuration options can be found here.

Save the deployment as deployment.yaml:

kind: Deployment
apiVersion: apps/v1
metadata:
  name: example-app
  labels:
    app: example-app
spec:
  replicas: 1
  selector:
    matchLabels:
      app: example-app
  template:
    metadata:
      labels:
        app: example-app
    spec:
      initContainers:
        - name: proxy-init
          image: openpolicyagent/proxy_init:v5
          # Configure the iptables bootstrap script to redirect traffic to the
          # Envoy proxy on port 8000, specify that Envoy will be running as user
          # 1111, and that we want to exclude port 8282 from the proxy for the
          # OPA health checks. These values must match up with the configuration
          # defined below for the "envoy" and "opa" containers.
          args: ["-p", "8000", "-u", "1111", "-w", "8282"]
          securityContext:
            capabilities:
              add:
            runAsNonRoot: false
            runAsUser: 0
      containers:
      - name: app
        image: openpolicyagent/demo-test-server:v1
        ports:
      - name: envoy
        image: envoyproxy/envoy:v1.17.0
        volumeMounts:
        - readOnly: true
          mountPath: /config
          name: proxy-config
        args:
        - "envoy"
        - "--config-path"
        - "/config/envoy.yaml"
        env:
        - name: ENVOY_UID
          value: "1111"
      - name: opa
        # Note: openpolicyagent/opa:latest-envoy is created by retagging
        # the latest released image of OPA-Envoy.
        image: openpolicyagent/opa:0.30.2-envoy
        volumeMounts:
        - readOnly: true
          mountPath: /policy
          name: opa-policy
        args:
        - "run"
        - "--server"
        - "--addr=localhost:8181"
        - "--diagnostic-addr=0.0.0.0:8282"
        - "--set=plugins.envoy_ext_authz_grpc.addr=:9191"
        - "--set=plugins.envoy_ext_authz_grpc.path=envoy/authz/allow"
        - "--set=decision_logs.console=true"
        - "--ignore=.*"
        - "/policy/policy.rego"
        livenessProbe:
          httpGet:
            path: /health?plugins
            scheme: HTTP
            port: 8282
          initialDelaySeconds: 5
          periodSeconds: 5
        readinessProbe:
          httpGet:
            path: /health?plugins
            scheme: HTTP
            port: 8282
          initialDelaySeconds: 5
          periodSeconds: 5
      volumes:
      - name: proxy-config
        configMap:
          name: proxy-config
      - name: opa-policy
        secret:
          secretName: opa-policy

kubectl apply -f deployment.yaml

Check that the Pod shows 3/3 containers READY the STATUS as Running:

kubectl get pod
NAME                           READY   STATUS    RESTARTS   AGE
example-app-67c644b9cb-bbqgh   3/3     Running   0          8s

The proxy-init container installs iptables rules to redirect all container traffic through the Envoy proxy sidecar. More information can be found .

In a second terminal, start a minikube tunnel to allow for use of the LoadBalancer service type.

minikube tunnel

In the first terminal, create a LoadBalancer service for the deployment.

kubectl get service example-app-service
NAME                  TYPE           CLUSTER-IP      EXTERNAL-IP     PORT(S)          AGE
example-app-service   LoadBalancer   10.109.64.199   10.109.64.199   8080:32170/TCP   5s

Set the SERVICE_URL environment variable to the service’s IP/port.

minikube:

export SERVICE_HOST=$(kubectl get service example-app-service -o jsonpath='{.status.loadBalancer.ingress[0].ip}')
export SERVICE_URL=$SERVICE_HOST:8080
echo $SERVICE_URL

minikube (example):

10.109.64.199:8080

6. Exercise the OPA policy

For convenience, we’ll want to store Alice’s and Bob’s tokens in environment variables.

export ALICE_TOKEN="eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJyb2xlIjoiZ3Vlc3QiLCJzdWIiOiJZV3hwWTJVPSIsIm5iZiI6MTUxNDg1MTEzOSwiZXhwIjoxNjQxMDgxNTM5fQ.K5DnnbbIOspRbpCr2IKXE9cPVatGOCBrBQobQmBmaeU"
export BOB_TOKEN="eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJyb2xlIjoiYWRtaW4iLCJzdWIiOiJZbTlpIiwibmJmIjoxNTE0ODUxMTM5LCJleHAiOjE2NDEwODE1Mzl9.WCxNAveAVAdRCmkpIObOTaSd0AJRECY2Ch2Qdic3kU8"

Check that Alice can get employees but cannot create one.

curl -i -H "Authorization: Bearer $ALICE_TOKEN" http://$SERVICE_URL/people

Check that Bob can get employees and also create one.

Check that Bob cannot create an employee with the same firstname as himself.

Congratulations for finishing the tutorial !

This tutorial showed how to use OPA as an External authorization service to enforce custom policies by leveraging Envoy’s External authorization filter.

Envoy’s external authorization filter allows optional response headers and body to be sent to the downstream client or upstream. An example of a rule that returns an object that not only indicates if a request is allowed or not but also provides optional response headers, body and HTTP status that can be sent to the downstream client or upstream can be seen here.