The 12 Factor App

Deploying is my business... And business is Good!

# BIO

ROMAN SACHENKO

BEER
MUSIC INSTRUMENTS
- bass
- drums
CAMPING
WEIRD JOKES
HOT SAUCES
SOFTWARE ENGINEERING

# BRIEF HISTORY

Reduce on-boarding time
Follow DevOps standards
Increase portability
Reduce communication needs
Adapt for CI/CD usage
Improve scalability

# WHY

DISCO - Deploy Infrastructure Safely, Consistently, and Optimally

# Coverage

DEVELOPMENT

CONFIGURATION

SCALING

ADMINISTRATION

DEPLOYMENT

ERROR HANDLING

DEPENDENCIES MANAGEMENT

THE 12 FACTOR APP

# (I) CODEBASE

"One codebase tracked in revision control, many deploys"

# (I) CODEBASE

Same codebase for everyone and everywhere
Shared code in the same repository
Packed and exported shared code
- NPM
- Artifactory

# (II) DEPENDENCIES

"Explicitly declare and isolate dependencies"

// .nvmrc
v18.12.0

// dockerfile
FROM node:18-alpine

in .nvmrc
in dockerfile
always commit package-lock.json/yarn.lock

# (II) DEPENDENCIES

Fix dependencies versions

# (II) DEPENDENCIES

Explicit dependencies declaration
Dockerfile and Package Manager - FRIENDS
Fix dependencies versions
No global dependencies
Check dependencies during service startup
Clean up
- npm prune --production

# (III) CONFIG

"Store config in the environment"

# (III) CONFIG

Do not hardcode config variables
Use environment variables everywhere
- process.env (NodeJS)
- use config service for access
.env for local development is not a shame
Validate config variables during the service startup

# (III) CONFIG

const config = () => ({
  dbConnectionString: process.env.DB_CONNECTION_STRING,
  customConf: {
    auth: process.env.CUSTOM_AUTH,
    options: process.env.CUSTOM_OPTIONS,
  }
})

const validate = (config) => {
  const errors = validateConfig(config)
  
  if (errors?.length) {
    throw new Error('Invalid config error')
  }
}

# (IV) BACKING SERVICES

"Treat backing services as attached resources"

# (IV) BACKING SERVICES

Databases, cache, queue (etc.) providers are resources
Local, development, staging, production parity
Use URL-based connection strings
Easy to attach and detach

[DEV]

[STAGING]

[PROD]

URL

# (V) BUILD, RELEASE, RUN

"Strictly separate build and run stages"

# (V) BUILD, RELEASE, RUN

Define build, release and run steps - split them
Tag artifacts and releases
Apply "Config" factor

# (VI) PROCESSES

"Execute the app as one or more stateless processes"

# (VI) PROCESSES

LOAD BALANCER

PROXY

# (VI) PROCESSES

Race conditions
Inaccurate calculations
Security issues
Reliability issues

Oops...

# (VI) PROCESSES

Stateless service, stateless process
Share nothing
Do not rely on in-memory, disk cached data
Rely on "Backing services"
Keep in mind horizontal scaling

# (VII) PORT BINDING

"Export services via port binding"

# (VII) PORT BINDING

Do not hardcode ports
Runtime ports injection
"Config" factor
Dynamic port biding

app.listen(process.env.PORT, () => {
  log(`What's up?`)
})

FROM nginx
EXPOSE 80 443

# (VIII) CONCURRENCY

"Scale out via the process model"

# (VIII) CONCURRENCY

Keep in mind horizontal scaling
Share nothing
Containerize
"Backing services" factor
Unix process model
Use process manager

# (IX) DISPOSABILITY

"Maximize robustness with fast startup and graceful shutdown"

# (IX) DISPOSABILITY

Start up fast
Shutdown gracefully
Aim to catch what is uncaugh

# (IX) DISPOSABILITY

/**
 * finish operations
 * release resources
 * log
 * re-route requests/handlers if needed
 */
const stopHandler = async () => {
  logger.info('Shutting down server...')
  
  setShutDownStatus() // respond with 503 
  
  await releaseResources()
  
  await closeServer()
}

 ['SIGTERM', 'SIGINT', 'SIGHUP']
   .forEach((signal) => process.on(signal, stopHandler))

# (IX) DISPOSABILITY

process.on('uncaughtException', async (err) => {
  logger.error('Uncaught Exception', err)
  await releaseResources()
  process.exit(1)
})

process.on('unhandledRejection', async (err) => {
  logger.error('Unhandled Rejection', err)
  await releaseResources()
  process.exit(1)
})
  
process.on('exit', (code) => {
  logger.info(`Process exited with code ${code}`)
  process.exit(code)
})

# (X) DEV/PROD PARITY

"Keep development, staging, and production as similar as possible"

# (X) DEV/PROD PARITY

Same code
Same dependencies
Same backing services
Same configurations method
Minimize the gaps

Aim to

# (XI) LOGS

"Treat logs as event streams"

# (XI) LOGS

Write logs to STDOUT / STDERR
Use structured output (JSON)
Aggregate logs outside of your app

What

# (XI) LOGS

> STDOUT

LOGS AGENT

> STDOUT

# (XII) ADMIN PROCESSES

"Run admin/management tasks as one-off processes"

# (XII) ADMIN PROCESSES

Define app logic and admin tasks - split
- same repository, different folders
- db migrations, seeding, hot changes
- manual re-do operations
Don't run admin tasks from the app startup script
Re-use your code
Learn built-in instruments and libraries

What