HTTPRoute Traffic Splitting

  5 minute read  

The HTTPRoute resource allows one or more backendRefs to be provided. Requests will be routed to these upstreams if they match the rules of the HTTPRoute. If an invalid backendRef is configured, then HTTP responses will be returned with status code 500 for all requests that would have been sent to that backend.

Installation

Follow the steps from the Quickstart to install Envoy Gateway and the example manifest. Before proceeding, you should be able to query the example backend using HTTP.

Single backendRef

When a single backendRef is configured in a HTTPRoute, it will receive 100% of the traffic.

cat <<EOF | kubectl apply -f -
apiVersion: gateway.networking.k8s.io/v1
kind: HTTPRoute
metadata:
  name: http-headers
spec:
  parentRefs:
  - name: eg
  hostnames:
  - backends.example
  rules:
  - matches:
    - path:
        type: PathPrefix
        value: /
    backendRefs:
    - group: ""
      kind: Service
      name: backend
      port: 3000
EOF

The HTTPRoute status should indicate that it has been accepted and is bound to the example Gateway.

kubectl get httproute/http-headers -o yaml

Get the Gateway’s address:

export GATEWAY_HOST=$(kubectl get gateway/eg -o jsonpath='{.status.addresses[0].value}')

Querying backends.example/get should result in a 200 response from the example Gateway and the output from the example app should indicate which pod handled the request. There is only one pod in the deployment for the example app from the quickstart, so it will be the same on all subsequent requests.

$ curl -vvv --header "Host: backends.example" "http://${GATEWAY_HOST}/get"
...
> GET /get HTTP/1.1
> Host: backends.example
> User-Agent: curl/7.81.0
> Accept: */*
> add-header: something
>
* Mark bundle as not supporting multiuse
< HTTP/1.1 200 OK
< content-type: application/json
< x-content-type-options: nosniff
< content-length: 474
< x-envoy-upstream-service-time: 0
< server: envoy
<
...
 "namespace": "default",
 "ingress": "",
 "service": "",
 "pod": "backend-79665566f5-s589f"
...

Multiple backendRefs

If multiple backendRefs are configured, then traffic will be split between the backendRefs equally unless a weight is configured.

First, create a second instance of the example app from the quickstart:

cat <<EOF | kubectl apply -f -
---
apiVersion: v1
kind: ServiceAccount
metadata:
  name: backend-2
---
apiVersion: v1
kind: Service
metadata:
  name: backend-2
  labels:
    app: backend-2
    service: backend-2
spec:
  ports:
    - name: http
      port: 3000
      targetPort: 3000
  selector:
    app: backend-2
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: backend-2
spec:
  replicas: 1
  selector:
    matchLabels:
      app: backend-2
      version: v1
  template:
    metadata:
      labels:
        app: backend-2
        version: v1
    spec:
      serviceAccountName: backend-2
      containers:
        - image: gcr.io/k8s-staging-gateway-api/echo-basic:v20231214-v1.0.0-140-gf544a46e
          imagePullPolicy: IfNotPresent
          name: backend-2
          ports:
            - containerPort: 3000
          env:
            - name: POD_NAME
              valueFrom:
                fieldRef:
                  fieldPath: metadata.name
            - name: NAMESPACE
              valueFrom:
                fieldRef:
                  fieldPath: metadata.namespace
EOF

Then create an HTTPRoute that uses both the app from the quickstart and the second instance that was just created

cat <<EOF | kubectl apply -f -
apiVersion: gateway.networking.k8s.io/v1
kind: HTTPRoute
metadata:
  name: http-headers
spec:
  parentRefs:
  - name: eg
  hostnames:
  - backends.example
  rules:
  - matches:
    - path:
        type: PathPrefix
        value: /
    backendRefs:
    - group: ""
      kind: Service
      name: backend
      port: 3000
    - group: ""
      kind: Service
      name: backend-2
      port: 3000
EOF

Querying backends.example/get should result in 200 responses from the example Gateway and the output from the example app that indicates which pod handled the request should switch between the first pod and the second one from the new deployment on subsequent requests.

$ curl -vvv --header "Host: backends.example" "http://${GATEWAY_HOST}/get"
...
> GET /get HTTP/1.1
> Host: backends.example
> User-Agent: curl/7.81.0
> Accept: */*
> add-header: something
>
* Mark bundle as not supporting multiuse
< HTTP/1.1 200 OK
< content-type: application/json
< x-content-type-options: nosniff
< content-length: 474
< x-envoy-upstream-service-time: 0
< server: envoy
<
...
 "namespace": "default",
 "ingress": "",
 "service": "",
 "pod": "backend-75bcd4c969-lsxpz"
...

Weighted backendRefs

If multiple backendRefs are configured and an un-even traffic split between the backends is desired, then the weight field can be used to control the weight of requests to each backend. If weight is not configured for a backendRef it is assumed to be 1.

The weight field in a backendRef controls the distribution of the traffic split. The proportion of requests to a single backendRef is calculated by dividing its weight by the sum of all backendRef weights in the HTTPRoute. The weight is not a percentage and the sum of all weights does not need to add up to 100.

The HTTPRoute below will configure the gateway to send 80% of the traffic to the backend service, and 20% to the backend-2 service.

cat <<EOF | kubectl apply -f -
apiVersion: gateway.networking.k8s.io/v1
kind: HTTPRoute
metadata:
  name: http-headers
spec:
  parentRefs:
  - name: eg
  hostnames:
  - backends.example
  rules:
  - matches:
    - path:
        type: PathPrefix
        value: /
    backendRefs:
    - group: ""
      kind: Service
      name: backend
      port: 3000
      weight: 8
    - group: ""
      kind: Service
      name: backend-2
      port: 3000
      weight: 2
EOF

Invalid backendRefs

backendRefs can be considered invalid for the following reasons:

  • The group field is configured to something other than "". Currently, only the core API group (specified by omitting the group field or setting it to an empty string) is supported
  • The kind field is configured to anything other than Service. Envoy Gateway currently only supports Kubernetes Service backendRefs
  • The backendRef configures a service with a namespace not permitted by any existing ReferenceGrants
  • The port field is not configured or is configured to a port that does not exist on the Service
  • The named Service configured by the backendRef cannot be found

Modifying the above example to make the backend-2 backendRef invalid by using a port that does not exist on the Service will result in 80% of the traffic being sent to the backend service, and 20% of the traffic receiving an HTTP response with status code 500.

cat <<EOF | kubectl apply -f -
apiVersion: gateway.networking.k8s.io/v1
kind: HTTPRoute
metadata:
  name: http-headers
spec:
  parentRefs:
  - name: eg
  hostnames:
  - backends.example
  rules:
  - matches:
    - path:
        type: PathPrefix
        value: /
    backendRefs:
    - group: ""
      kind: Service
      name: backend
      port: 3000
      weight: 8
    - group: ""
      kind: Service
      name: backend-2
      port: 9000
      weight: 2
EOF

Querying backends.example/get should result in 200 responses 80% of the time, and 500 responses 20% of the time.

$ curl -vvv --header "Host: backends.example" "http://${GATEWAY_HOST}/get"
> GET /get HTTP/1.1
> Host: backends.example
> User-Agent: curl/7.81.0
> Accept: */*
>
* Mark bundle as not supporting multiuse
< HTTP/1.1 500 Internal Server Error
< server: envoy
< content-length: 0
<

Last modified November 21, 2024: docs: fix wrong description on ALSEnvoyProxyAccessLog (#4751) (2def6a4)