Build Your Own

Continuous deployment

Infrastructure 

RootedCON, 2018-03-03

Your Host this Evening

@pinchito

Organizer at MadridJS

Backend developer at LogTrust

What We Will Cover

Testing

Integration

Deployment

Distributed Deployment

Monitoring

Tests

Who automatizes the automatizer?

Anonymous

types of Tests

Unit tests

Integration tests / system tests

Load tests

Automatic tests — are the bests!

Better than the rests!

Automate!

Design a control API

Start and stop the system using the API

Self-contained tests: clean-up

Three basic rules:

• One single button
• Fail fast and loudly
• No human intervention
  

Make your life easier

Make your systems easy to test

Integration

Dr Frankenstein

Integration Environment

Between Development and Production

Also called: preproduction, replica, build env

Enough data to run all tests

Populated with a script

Self-contained

Continuous Integration

Store unit and integration tests

Run the whole suite with every change

Check if something has broken

For any new error, introduce a new test

Why Integrate All the Time?

Close the gap between development and tests

Safety net for developers

Notice errors when the code is fresh

Shorten development cycles!

Tooling

Jenkins

Travis-CI

Strider CD

(CD > CI)

Various proprietary dumpster fires

Deployment

Anonymous

Deploying Code

Put your code into production

Different servers, different rules

With node:

git pull
npm install
sudo service [name] restart

Automate!

Do the manual way...

Or do the cool way: with code

DevOps for the win!

Three basic rules:

• One single button
• Fail fast and loudly
• No human intervention
  

Make your life easier

Make your systems easy to deploy

Continuous Deployment

Too much of a good thing?

Shorten deployment cycles

Don't let changes accumulate

Stability increases (paradoxically?)

Practice makes perfect!

Improving Your Life

Continuous Integration

Continuous Delivery

Continuous Deployment

Reject imitations!

Test Before Deploying

Run all test suites

Verify results

Certify quality

A small but essential addition:

git pull
npm install
npm test
sudo service [name] restart

Distributed

Deployment

From one Server to Many

Horizontal architecture

One deployment, many servers

Production: listen to signal

Integration: send signals

Overview

Dynamic List of Servers

Get from your cloud provider?

Use AWS/gcloud/Azure API

Integration finds all servers

Notify all production servers

Bonus Track

MONITORING

Proactive Monitoring

Let alarms, alerts and warnings come to you

Every alarm must have:

• A cause
  
• An alert system
  
• An action associated

Make your life easier

Make your systems easy to monitor

Practical Session 5

Send a notification if deployment fails

Where To Go from here

Automatic rollback

Staged deployments

Canary deployments

Feature flags

Testing in Production

Thanks! We're Hiring!

@pinchito