Kubernetes Gateway API

7 minute read page test

In addition to its own traffic management API, Istio includes beta support for the Kubernetes Gateway API and intends to make it the default API for traffic management in the future. This document describes the differences between the Istio and Kubernetes APIs and provides a simple example that shows you how to configure Istio to expose a service outside the service mesh cluster using the Gateway API. Note that these APIs are an actively developed evolution of the Kubernetes Service and Ingress APIs.

Setup

The Gateway APIs do not come installed by default on most Kubernetes clusters. Install the Gateway API CRDs if they are not present:

$ kubectl get crd gateways.gateway.networking.k8s.io &> /dev/null || \
  { kubectl kustomize "github.com/kubernetes-sigs/gateway-api/config/crd?ref=v0.6.1" | kubectl apply -f -; }

Install Istio using the minimal profile:

$ istioctl install --set profile=minimal -y

Differences from Istio APIs

The Gateway APIs share a lot of similarities to the Istio APIs such as Gateway and VirtualService. The main resource shares the same name, Gateway, and the resources serve similar goals.

The new Gateway APIs aim to take the learnings from various Kubernetes ingress implementations, including Istio, to build a standardized vendor neutral API. These APIs generally serve the same purposes as Istio Gateway and VirtualService, with a few key differences:

In Istio APIs, a Gateway configures an existing gateway Deployment/Service that has been deployed. In the Gateway APIs, the Gateway resource both configures and deploys a gateway. See Deployment Methods for more information.
In the Istio VirtualService, all protocols are configured within a single resource. In the Gateway APIs, each protocol type has its own resource, such as HTTPRoute and TCPRoute.
While the Gateway APIs offer a lot of rich routing functionality, it does not yet cover 100% of Istio’s feature set. Work is ongoing to extend the API to cover these use cases, as well as utilizing the APIs extensibility to better expose Istio functionality.

Configuring a Gateway

See the Gateway API documentation for information about the APIs.

In this example, we will deploy a simple application and expose it externally using a Gateway.

First, deploy the httpbin test application:

$ kubectl apply -f @samples/httpbin/httpbin.yaml@

Deploy the Gateway API configuration including a single exposed route (i.e., /get):

$ kubectl create namespace istio-ingress
$ kubectl apply -f - <<EOF
apiVersion: gateway.networking.k8s.io/v1beta1
kind: Gateway
metadata:
  name: gateway
  namespace: istio-ingress
spec:
  gatewayClassName: istio
  listeners:
  - name: default
    hostname: "*.example.com"
    port: 80
    protocol: HTTP
    allowedRoutes:
      namespaces:
        from: All
---
apiVersion: gateway.networking.k8s.io/v1beta1
kind: HTTPRoute
metadata:
  name: http
  namespace: default
spec:
  parentRefs:
  - name: gateway
    namespace: istio-ingress
  hostnames: ["httpbin.example.com"]
  rules:
  - matches:
    - path:
        type: PathPrefix
        value: /get
    backendRefs:
    - name: httpbin
      port: 8000
EOF

Set the Ingress Host environment variable:

$ kubectl wait -n istio-ingress --for=condition=ready gateways.gateway.networking.k8s.io gateway
$ export INGRESS_HOST=$(kubectl get gateways.gateway.networking.k8s.io gateway -n istio-ingress -ojsonpath='{.status.addresses[*].value}')

Access the httpbin service using curl:
```
$ curl -s -I -HHost:httpbin.example.com "http://$INGRESS_HOST/get"
HTTP/1.1 200 OK
server: istio-envoy
...
```
Note the use of the -H flag to set the Host HTTP header to “httpbin.example.com”. This is needed because the HTTPRoute is configured to handle “httpbin.example.com”, but in your test environment you have no DNS binding for that host and are simply sending your request to the ingress IP.

Access any other URL that has not been explicitly exposed. You should see an HTTP 404 error:

$ curl -s -I -HHost:httpbin.example.com "http://$INGRESS_HOST/headers"
HTTP/1.1 404 Not Found
...

Update the route rule to also expose /headers and to add a header to the request:

$ kubectl apply -f - <<EOF
apiVersion: gateway.networking.k8s.io/v1beta1
kind: HTTPRoute
metadata:
  name: http
  namespace: default
spec:
  parentRefs:
  - name: gateway
    namespace: istio-ingress
  hostnames: ["httpbin.example.com"]
  rules:
  - matches:
    - path:
        type: PathPrefix
        value: /get
    - path:
        type: PathPrefix
        value: /headers
    filters:
    - type: RequestHeaderModifier
      requestHeaderModifier:
        add:
        - name: my-added-header
          value: added-value
    backendRefs:
    - name: httpbin
      port: 8000
EOF

Access /headers again and notice header My-Added-Header has been added to the request:

$ curl -s -HHost:httpbin.example.com "http://$INGRESS_HOST/headers"
{
  "headers": {
    "Accept": "*/*",
    "Host": "httpbin.example.com",
    "My-Added-Header": "added-value",
...

Deployment methods

In the example above, you did not need to install an ingress gateway Deployment prior to configuring a Gateway. In the default configuration, a gateway Deployment and Service is automatically provisioned based on the Gateway configuration. For advanced use cases, manual deployment is still allowed.

Automated Deployment

By default, each Gateway will automatically provision a Service and Deployment of the same name. These configurations will be updated automatically if the Gateway changes (for example, if a new port is added).

These resources can be customized in a few ways:

Annotations and labels on the Gateway will be copied to the Service and Deployment. This allows configuring things such as Internal load balancers that read from these fields.
Istio offers an additional annotation to configure the generated resources:
Annotation Purpose
networking.istio.io/service-type Controls the Service.spec.type field. For example, set to ClusterIP to not expose the service externally. The default is LoadBalancer.

Annotation	Purpose
`networking.istio.io/service-type`	Controls the `Service.spec.type` field. For example, set to `ClusterIP` to not expose the service externally. The default is `LoadBalancer`.

The Service.spec.loadBalancerIP field can be explicit set by configuring the addresses field:

apiVersion: gateway.networking.k8s.io/v1beta1
kind: Gateway
metadata:
  name: gateway
spec:
  addresses:
  - value: 192.0.2.0
    type: IPAddress
...

Note: only one address may be specified.

(Advanced) The generated Pod configuration can be configured by Custom Injection Templates.

Resource Attachment and Scaling

Resources can be attached to a Gateway to customize it. However, most Kubernetes resources do not currently support attaching directly to a Gateway, but they can be attached to the corresponding generated Deployment and Service instead. This is easily done because both of these resources are generated with name <gateway name>-<gateway class name> and with a label istio.io/gateway-name: <gateway name>.

For example, to deploy a Gateway with a HorizontalPodAutoscaler and PodDisruptionBudget:

apiVersion: gateway.networking.k8s.io/v1beta1
kind: Gateway
metadata:
  name: gateway
spec:
  gatewayClassName: istio
  listeners:
  - name: default
    hostname: "*.example.com"
    port: 80
    protocol: HTTP
    allowedRoutes:
      namespaces:
        from: All
---
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: gateway
spec:
  # Match the generated Deployment by reference
  # Note: Do not use `kind: Gateway`.
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: gateway-istio
  minReplicas: 2
  maxReplicas: 5
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 50
---
apiVersion: policy/v1
kind: PodDisruptionBudget
metadata:
  name: gateway
spec:
  minAvailable: 1
  selector:
    # Match the generated Deployment by label
    matchLabels:
      istio.io/gateway-name: gateway

Manual Deployment

If you do not want to have an automated deployment, a Deployment and Service can be configured manually.

When this option is done, you will need to manually link the Gateway to the Service, as well as keep their port configuration in sync.

To link a Gateway to a Service, configure the addresses field to point to a single Hostname.

apiVersion: gateway.networking.k8s.io/v1beta1
kind: Gateway
metadata:
  name: gateway
spec:
  addresses:
  - value: ingress.istio-gateways.svc.cluster.local
    type: Hostname
...

Mesh Traffic

The Gateway API can also be used to configure mesh traffic. This is done by configuring the parentRef to point to a service, instead of a gateway.

For example, to add a header on all calls to an in-cluster Service named example:

apiVersion: gateway.networking.k8s.io/v1beta1
kind: HTTPRoute
metadata:
  name: mesh
spec:
  parentRefs:
  - kind: Service
    name: example
  rules:
  - filters:
    - type: RequestHeaderModifier
      requestHeaderModifier:
        add:
        - name: my-added-header
          value: added-value
  - backendRefs:
    - name: example
      port: 80

More details and examples can be found in other traffic management tasks.

Cleanup

Uninstall Istio and the httpbin sample:

$ kubectl delete -f @samples/httpbin/httpbin.yaml@
$ kubectl delete httproute http
$ kubectl delete gateways.gateway.networking.k8s.io gateway -n istio-ingress
$ istioctl uninstall -y --purge
$ kubectl delete ns istio-system
$ kubectl delete ns istio-ingress

Remove the Gateway API CRDs if they are no longer needed:

$ kubectl kustomize "github.com/kubernetes-sigs/gateway-api/config/crd?ref=v0.6.1" | kubectl delete -f -