Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Serverless Microservices are the New Black

Serverless Microservices are the New Black

Looking at a simple example of a microservice (nanoservice?) built with API Gateway and Apache OpenWhisk

Lorna Mitchell

June 14, 2018
Tweet

More Decks by Lorna Mitchell

Other Decks in Technology

Transcript

  1. Serverless FaaS: Functions as a Service • write a function

    (many languages supported) • deploy it to the cloud (Lambda, Cloud Functions, etc) • only pay while the function is running (charged per GBsec) • your platform scales on demand @lornajane
  2. When To Go Serverless • To create a small, scalable

    application (focussed API, microservices) • For occasional server needs (contact form on static site) • To provide compute power (processing quantities of data) • To move heavy lifting off web platform (classic example: PDF generation) @lornajane
  3. Microservices Microservices are: • small and modular • loosely coupled

    • independently developed and deployed • great for building components • decentralised ... they are basically small HTTP APIs @lornajane
  4. Microservice Design Points I prefer RESTful-ish APIs • Status codes

    are important • Headers are for metadata • URLs and verbs together define what happens • All endpoints are stateless • SSL is required @lornajane
  5. Lorna's Plans Service Keep a list of my travel plans,

    with locations and dates. Use a serverless platform (IBM Cloud Functions) and PostgreSQL GET /plans list all plans GET /plans/42 show one plan POST /plans create a new plan DELETE /plans/42 delete a plan @lornajane
  6. Writing Serverless Functions • Small, self-contained units of functionality •

    Run on-demand in response to an event • Incoming parameters can include • event information • parameters set at deploy time • connection info for other services @lornajane
  7. Make Plans: the Code 1 const pgp = require('pg-promise')(); 2

    function main(params) { 3 var postgres_url = params['__bx_creds']['compose-for-postgresql']['uri']; 4 var base_url = params['__ow_headers']['x-forwarded-url']; 5 return new Promise(function(resolve, reject) { 6 db = pgp(postgres_url, []); 7 8 db.one("INSERT INTO plans (location, travel_date) VALUES 9 ($1, $2) RETURNING plan_id", 10 [location, travel_date]) 11 .then(function(data) { 12 var redirect_to = base_url + "/" + data.plan_id; 13 resolve({headers: {"Location": redirect_to}, 14 statusCode: 303}) 15 }) @lornajane
  8. Prepare to Deploy: package In OpenWhisk, there are "packages". These

    let us: • group actions together • add parameters to a package that will be available to all actions From the deployment script, the line to create plans-api: ibmcloud wsk package update plans-api @lornajane
  9. Prepare to Deploy: services The function needs to connect to

    the database. We can bind the database to the package to achieve this: ibmcloud wsk service bind compose-for-postgresql plans-api @lornajane
  10. Prepare to Deploy: libraries To include extra libraries, we can:

    • create package.json and run npm install • zip up index.js and node_modules into a zip file • deploy the zip file, and include runtime instructions cd write-plan zip -rq write-plan.zip index.js node_modules @lornajane
  11. Make Plans: Deploy We're ready! Push the action to the

    cloud: ibmcloud wsk action update --kind nodejs:8 --web raw \ plans-api/write-plan write-plan.zip @lornajane
  12. Make Plans: API Gateway To have this action repsond to

    our HTTP request, set up the API Gateway: • create the API • set up the path, verb and action to link together ibmcloud wsk api create /plans GET plans-api/get-plans \ --response-type http @lornajane
  13. Calling The Endpoint Quick example with cURL (other clients available):

    $ curl -L -H "Content-Type: application/json" \ https://service.eu.apiconnect.ibmcloud.com/.../plans \ -d '{"location": "Turin", "travel_date": "2018-04-11"}' { "plans": [{ "plan_id": 3, "travel_date": "2018-04-11T00:00:00.000Z", "location": "Turin" }] } @lornajane
  14. Project Structure Many possible approaches, this is mine: . ├──

    deploy.sh ├── get-plans │ ├── index.js │ ├── node_modules │ ├── package-lock.json │ └── package.json └── write-plan ├── index.js ├── node_modules ├── package-lock.json └── package.json @lornajane
  15. Deployment Using https://travis-ci.com/ • deploy with a script • script

    downloads ibmcloud tool and cloud-functions plugin • set an API key as an environment variable • then run commands (see deploy.sh in GitHub project) @lornajane
  16. Serverless Ideal for working with many small parts Apache OpenWhisk

    paired with API Gateway: perfect candidate for microservices @lornajane
  17. Microservices Service Oriented Architecture is alive and well • microservices

    expose endpoints • they share reusable components • specific components guard access to services/datastores • each component can be separately developed, tested and deployed @lornajane
  18. Resources • Code: https://github.com/lornajane/plans-microservice • Apache OpenWhisk: http://openwhisk.apache.org • IBM

    Cloud Functions: https://www.ibm.com/cloud/functions • My blog: https://lornajane.net @lornajane