control-plane

This repository contains a Go-based implementation of an API server that implements the discovery service APIs defined in data-plane-api.

Scope

Due to the variety of platforms out there, there is no single control plane implementation that can satisfy everyone's needs. Hence this code base does not attempt to be a full scale control plane for a fleet of Envoy proxies. Instead, it provides infrastructure that is shared by multiple different control plane implementations. The components provided by this library are:

• API Server: A generic gRPC based API server that implements xDS APIs as defined in the data-plane-api. The API server is responsible for pushing configuration updates to Envoys. Consumers should be able to import this go library and use the API server as is, in production deployments.

• Configuration Cache: The library will cache Envoy configurations in memory in an attempt to provide fast response to consumer Envoys. It is the responsibility of the consumer of this library to populate the cache as well as invalidate it when necessary. The cache will be keyed based on a pre-defined hash function whose keys are based on the Node information.

At this moment, this repository will not tackle translating platform specific representation of resources (e.g., services, instances of services, etc.) into Envoy-style configuration. Based on usage and feedback, we might decided to revisit this aspect at a later point in time.

Requirements

1. Go 1.12+

Quick start

It's recommended to run the command with script ./build/run_docker.sh as it executes the command in the same environment as the circle ci. This makes sure to produce a consistent set of generated files.

1. Setup existing build:

   ./build/run_docker.sh make build test

2. Generate proto files (if you update the data-plane-api dependency)

   ./build/run_docker.sh make generate

   NOTE: you may need to apply a small patch to correct imports in the generate files:

   ./build/run_docker.sh make generate-patch

   Format the code:

   ./build/run_docker.sh make format

   Run build and unit tests again:

   ./build/run_docker.sh make build test

3. Run integration test against the latest Envoy binary:

   ./build/run_docker.sh make integration

Usage

Register services on the gRPC server as follows.

import (
	"net"

	api "github.com/envoyproxy/go-control-plane/envoy/api/v2"
	discovery "github.com/envoyproxy/go-control-plane/envoy/service/discovery/v2"
	"github.com/envoyproxy/go-control-plane/pkg/cache"
	xds "github.com/envoyproxy/go-control-plane/pkg/server"
)

func main() {
	snapshotCache := cache.NewSnapshotCache(false, hash{}, nil)
	server := xds.NewServer(snapshotCache, nil)
	grpcServer := grpc.NewServer()
	lis, _ := net.Listen("tcp", ":8080")

	discovery.RegisterAggregatedDiscoveryServiceServer(grpcServer, server)
	api.RegisterEndpointDiscoveryServiceServer(grpcServer, server)
	api.RegisterClusterDiscoveryServiceServer(grpcServer, server)
	api.RegisterRouteDiscoveryServiceServer(grpcServer, server)
	api.RegisterListenerDiscoveryServiceServer(grpcServer, server)
	go func() {
		if err := grpcServer.Serve(lis); err != nil {
			// error handling
		}
	}()
}

As mentioned in Scope, you need to cache Envoy configurations. Generate the key based on the node information as follows and cache the configurations.

import "github.com/envoyproxy/go-control-plane/pkg/cache"

var clusters, endpoints, routes, listeners []cache.Resource

snapshotCache := cache.NewSnapshotCache(false, hash{}, nil)
snapshot := cache.NewSnapshot("1.0", endpoints, clusters, routes, listeners)
_ = snapshotCache.SetSnapshot("node1", snapshot)