serverless-discovery-sdk-js

The last serverless micro-service you'll ever wonder how to find

The AWS Serverless Discovery SDK interacts with a discovery microservice to discover endpoints for micro-services written for a serverless architecture. This is similar to clustered services such as Consul or ZooKeeper, but without the concept of instances or nodes that must be monitored for online state. This library is designed to support use both on the server side (for service-to-service lookups) and on the browser/client side.

This project contains the Typescript/Javascript bindings for the discovery service; Other bindings can be found in the AdAstraDev organization on GitHub

Installation

npm install @adastradev/serverless-discovery-sdk

Usage

Service/Cloud Dependencies

Semver versioning is supported by the discovery service 1.1.x. Pass a semver compatible value in the lookupService call to receive the newest compatible matching version. Services and their desired versions can also be specified in the cloudDependencies field of package.json.

{
  "cloudDependencies": {
    "service1": "1.x",
    "service2": "^1.2.8-testbranch", // A pre-release version for development purposes
    "service3": "3.x.x"
  }
}

Version Postfix values

In some testing environments, it can be useful to modify the lookup version to avoid collision with a production environment. If the VERSION_POSTFIX environment variable at runtime, it will always append this to the version of a lookup call.

If you are looking up services which are highly coupled or are not well isolated, and using them for system tests, you should:

• Set the VERSION_POSTFIX environment variable set to -staging
• Pass the environment variable through to the runtime where lookups are happening (lambda, docker, etc.)

If there is a lookup for serviceA, version 1.1.0, it will instead only talk to 1.1.0-staging. All lookup calls will follow a similar pattern while the environment variable is present.

TL;DR: If you are looking up services which are not well isolated, and rely on a staging environment to avoid operations on prod databases/resources, add the following to your pipeline in a staging deployment/testing step.

bitbucket-pipelines.yml:

- export VERSION_POSTFIX='-staging'
# Deployment steps follow...

serverless.yml

provider:
  environment:
    VERSION_POSTFIX: ${env:VERSION_POSTFIX, ''}

Code Example

I recommend setting up a utility function to handle construction of the SDK, and the lookup call - see below example.

import { DiscoverySdk } from '@adastradev/serverless-discovery-sdk';

export default async function lookup(serviceName) {

  const sdk = new DiscoverySdk(
    process.env.DISCOVERY_SERVICE_URL,
    process.env.DISCOVERY_SERVICE_REGION,
    // Non-versioned services will default to lookup via this stage
    process.env.DEFAULT_STAGE,
    undefined,
    // Create map of cloudDependencies from package.json
    new Map(Object.entries(require('../path/to/package.json')['cloudDependencies'])),
  );

  const endpoints = await sdk.lookupService(
    serviceName
  );

  return endpoints[0];

}