Developer Guide

This section tells how to develop Envoy Gateway.

Envoy Gateway is built using a make-based build system. Our CI is based on Github Actions using workflows.

Prerequisites

go

make

docker

python3

  • Need a python3 program
  • Must have a functioning venv module; this is part of the standard library, but some distributions (such as Debian and Ubuntu) replace it with a stub and require you to install a python3-venv package separately.

Quickstart

  • Run make help to see all the available targets to build, test and run Envoy Gateway.

Building

  • Run make build to build all the binaries.
  • Run make build BINS="envoy-gateway" to build the Envoy Gateway binary.
  • Run make build BINS="egctl" to build the egctl binary.

Note: The binaries get generated in the bin/$OS/$ARCH directory, for example, bin/linux/amd64/.

Testing

  • Run make test to run the golang tests.

  • Run make testdata to generate the golden YAML testdata files.

Running Linters

  • Run make lint to make sure your code passes all the linter checks. Note: The golangci-lint configuration resides here.

Building and Pushing the Image

  • Run IMAGE=docker.io/you/gateway-dev make image to build the docker image.
  • Run IMAGE=docker.io/you/gateway-dev make push-multiarch to build and push the multi-arch docker image.

Note: Replace IMAGE with your registry’s image name.

Deploying Envoy Gateway for Test/Dev

  • Run make create-cluster to create a Kind cluster.

Option 1: Use the Latest gateway-dev Image

  • Run TAG=latest make kube-deploy to deploy Envoy Gateway in the Kind cluster using the latest image. Replace latest to use a different image tag.

Option 2: Use a Custom Image

  • Run make kube-install-image to build an image from the tip of your current branch and load it in the Kind cluster.
  • Run IMAGE_PULL_POLICY=IfNotPresent make kube-deploy to install Envoy Gateway into the Kind cluster using your custom image.

Deploying Envoy Gateway in Kubernetes

  • Run TAG=latest make kube-deploy to deploy Envoy Gateway using the latest image into a Kubernetes cluster (linked to the current kube context). Preface the command with IMAGE or replace TAG to use a different Envoy Gateway image or tag.
  • Run make kube-undeploy to uninstall Envoy Gateway from the cluster.

Note: Envoy Gateway is tested against Kubernetes v1.24.0.

Demo Setup

  • Run make kube-demo to deploy a demo backend service, gatewayclass, gateway and httproute resource (similar to steps outlined in the Quickstart docs) and test the configuration.
  • Run make kube-demo-undeploy to delete the resources created by the make kube-demo command.

Run Gateway API Conformance Tests

The commands below deploy Envoy Gateway to a Kubernetes cluster and run the Gateway API conformance tests. Refer to the Gateway API conformance homepage to learn more about the tests. If Envoy Gateway is already installed, run TAG=latest make run-conformance to run the conformance tests.

On a Linux Host

  • Run TAG=latest make conformance to create a Kind cluster, install Envoy Gateway using the latest gateway-dev image, and run Gateway API conformance tests.

On a Mac Host

Since Mac doesn’t support directly exposing the Docker network to the Mac host, use one of the following workarounds to run conformance tests:

  • Deploy your own Kubernetes cluster or use Docker Desktop with Kubernetes support and then run TAG=latest make kube-deploy run-conformance. This will install Envoy Gateway using the latest gateway-dev image to the Kubernetes cluster using the current kubectl context and run the conformance tests. Use make kube-undeploy to uninstall Envoy Gateway.
  • Install and run Docker Mac Net Connect and then run TAG=latest make conformance.

Note: Preface commands with IMAGE or replace TAG to use a different Envoy Gateway image or tag. If TAG is unspecified, the short SHA of your current branch is used.

Debugging the Envoy Config

An easy way to view the envoy config that Envoy Gateway is using is to port-forward to the admin interface port (currently 19000) on the Envoy deployment that corresponds to a Gateway so that it can be accessed locally.

Get the name of the Envoy deployment. The following example is for Gateway eg in the default namespace:

export ENVOY_DEPLOYMENT=$(kubectl get deploy -n envoy-gateway-system --selector=gateway.envoyproxy.io/owning-gateway-namespace=default,gateway.envoyproxy.io/owning-gateway-name=eg -o jsonpath='{.items[0].metadata.name}')

Port forward the admin interface port:

kubectl port-forward deploy/${ENVOY_DEPLOYMENT} -n envoy-gateway-system 19000:19000

Now you are able to view the running Envoy configuration by navigating to 127.0.0.1:19000/config_dump.

There are many other endpoints on the Envoy admin interface that may be helpful when debugging.

JWT Testing

An example JSON Web Token (JWT) and JSON Web Key Set (JWKS) are used for the request authentication task. The JWT was created by the JWT Debugger, using the RS256 algorithm. The public key from the JWTs verify signature was copied to JWK Creator for generating the JWK. The JWK Creator was configured with matching settings, i.e. Signing public key use and the RS256 algorithm. The generated JWK was wrapped in a JWKS structure and is hosted in the repo.

Benchmarking

Envoy Gateway uses nighthawk for benchmarking, and mainly concerned with its performance and scalability as a control-plane.

The performance and scalability concerns come from several aspects for control-plane:

  • The consumption of memory and CPU.
  • The rate of configuration changes.

The benchmark test is running on a Kind cluster, you can start a Kind cluster and run benchmark test on it by executing make benchmark.

The benchmark report will be included in the release artifacts, you can learn more by downloading the detailed benchmark report, namely benchmark_report.zip.

Here are some brief benchmark reports about Envoy Gateway:

  • It will take up nearly 550MiB memory and 11s total CPU time for (1 GatewayClass + 1 Gateway + 500 HTTRoutes) settings

Last modified December 20, 2024: chore: fix typo (#4958) (2a10d47)