Skip to content

Developer Guide



Install these tools before proceeding:

  1. AWS CLI,
  2. kubectl - the Kubernetes CLI,
  3. helm - the package manager for Kubernetes,
  4. eksctl- the CLI for Amazon EKS,
  5. go v1.20.x - language,
  6. yq - CLI to manipulate yaml files,
  7. jq - CLI to manipulate json files,
  8. make- build automation tool.

Cluster creation and setup

Before proceeding to the next sections, you need to:

  1. Create a and set up a cluster dev-cluster with the controller following the AWS Gateway API Controller installation guide on Amazon EKS.


    You can either install the Controller and CRDs following the steps in the installation guide or using the steps below if you prefer to create the individual CRDs.

  2. Clone the AWS Gateway API Controller repository.

    git clone
    cd aws-application-networking-k8s

  3. Install dependencies with script:
    make toolchain


Once cluster is ready, we need to apply CRDs for gateway-api resources. First install core gateway-api CRDs:

Install the latest v1 CRDs:

kubectl apply -f config/crds/bases/k8s-gateway-v1.0.0.yaml


Note that v1 CRDs are not included in deploy-*.yaml and helm chart by default.

Install gateway-api v1beta1 CRDs.

kubectl apply -f config/crds/bases/k8s-gateway-v0.6.1.yaml

And install additional CRDs for the controller:

kubectl apply -f config/crds/bases/externaldns.k8s.io_dnsendpoints.yaml
kubectl apply -f config/crds/bases/gateway.networking.k8s.io_tlsroutes.yaml
kubectl apply -f config/crds/bases/application-networking.k8s.aws_serviceexports.yaml
kubectl apply -f config/crds/bases/application-networking.k8s.aws_serviceimports.yaml
kubectl apply -f config/crds/bases/application-networking.k8s.aws_targetgrouppolicies.yaml
kubectl apply -f config/crds/bases/application-networking.k8s.aws_vpcassociationpolicies.yaml
kubectl apply -f config/crds/bases/application-networking.k8s.aws_accesslogpolicies.yaml
kubectl apply -f config/crds/bases/application-networking.k8s.aws_iamauthpolicies.yaml

When e2e tests are terminated during execution, it might break clean-up stage and resources will leak. To delete dangling resources manually use cleanup script:

make e2e-clean

Local Development

A minimal test of changes can be done with make presubmit. This command will also run on PR.

make presubmit

Start controller in development mode, that will point to cluster (see setup above).

// should be region of the cluster
REGION=us-west-2 make run

You can explore a collection of different yaml configurations in the examples folder that can be applied to cluster.

To run it against specific lattice service endpoint.


To easier load environment variables, if you hope to run the controller by GoLand IDE locally, you could run the ./scripts/ And use "EnvFile" GoLand plugin to read the env variables from the generated .env file.

End-to-End Testing

For larger changes it's recommended to run e2e suites on your local cluster. E2E tests require a service network named test-gateway with cluster VPC associated to run. You can either set up service network manually or use DEFAULT_SERVICE_NETWORK option when running controller locally. (e.g. DEFAULT_SERVICE_NETWORK=test-gateway make run)

REGION=us-west-2 make e2e-test

For the RAM Share test suite, which runs cross-account e2e tests, you will need a secondary AWS Account with a role that can be assumed by the primary account during test execution. You can create an IAM Role, with a Trust Policy allowing the primary account to assume it, via the AWS IAM Console.

export SECONDARY_ACCOUNT_TEST_ROLE_ARN=arn:aws:iam::000000000000:role/MyRole
export FOCUS="RAM Share"
REGION=us-west-2 make e2e-test

You can use the FOCUS environment variable to run some specific test cases based on filter condition. You could assign the string in the Describe("xxxxxx") or It("xxxxxx") to the FOCUS environment variable to run the specific test cases.

var _ = Describe("HTTPRoute path matches", func() {
    It("HTTPRoute should support multiple path matches", func() {
        // test case body

For example, to run the test case "HTTPRoute should support multiple path matches", you could run the following command:

export FOCUS="HTTPRoute should support multiple path matches"
export REGION=us-west-2
make e2e-test

Conversely, you can use the SKIP environment variable to skip specific test cases.

For example, to skip the same test as above, you would run the following command:

export SKIP="HTTPRoute should support multiple path matches"

For more detail on filter condition for ginkgo

After all test cases running finished, in the AfterSuite() function, it will clean up k8s and vpc lattice resource created by current test cases running.


The controller documentation is managed in docs/ directory, and built with mkdocs. It uses mike to manage versioning. To build and verify your changes locally:

pip install -r requirements.txt
make docs
The website will be located in site/ directory. You can also run a local dev-server by running mike serve or mkdocs serve.


Before sending a Pull Request, you should run unit tests:

make presubmit

For larger, functional changes, run e2e tests:

make e2e-test

Make Docker Image

make docker-build

Deploy Controller inside a Kubernetes Cluster

Generate deploy.yaml

make build-deploy

Then follow Deploying the AWS Gateway API Controller to configure and deploy the docker image.