Heroku

Heroku is a fully managed cloud Platform as a Service (PaaS) that enables developers to build, run, and scale applications without infrastructure management overhead.

When to Use

• Rapid Application Deployment & MVPs: Deploying full-stack web applications with git push heroku main and zero server management.
• Managed Data Add-ons: Instantly provisioning production-grade Heroku Postgres, Redis, and Apache Kafka.
• Procfile-Driven Workloads: Declaring web, background worker, and cron scheduler processes clearly.
• Enterprise Review Apps: Automatically spinning up ephemeral preview environments for every GitHub pull request.

Quick Start

heroku create
git push heroku main

# Add Postgres
heroku addons:create heroku-postgresql:standard-0

Core Concepts

Declarative Process Definition with Procfile

Declaring web servers, queue workers, and database release migrations:

# Procfile in repository root
release: python manage.py migrate --no-input
web: gunicorn myproject.wsgi:application --workers 4 --bind 0.0.0.0:$PORT
worker: celery -A myproject worker -l info --concurrency 2
clock: python clock.py

Ephemeral Review Apps Configuration (app.json)

Automating branch preview environments:

{
  "name": "SaaS Platform",
  "description": "Production SaaS application with Heroku Postgres and Redis",
  "scripts": {
    "postdeploy": "python manage.py setup_test_fixtures"
  },
  "env": {
    "DJANGO_SETTINGS_MODULE": "myproject.settings.review",
    "SECRET_KEY": { "generator": "secret" }
  },
  "addons": [
    { "plan": "heroku-postgresql:essential-0" },
    { "plan": "heroku-redis:mini" }
  ]
}

Heroku CLI Operations

Scaling dynos, managing config, and running one-off processes:

# Set production environment variables
heroku config:set NODE_ENV=production DATABASE_POOL_SIZE=20 -a my-prod-app

# Scale web and worker dynos
heroku ps:scale web=2:standard-2x worker=1:standard-1x -a my-prod-app

# Run one-off interactive console or migration
heroku run bash -a my-prod-app
heroku logs --tail -a my-prod-app

Common Patterns

Multi-Process Procfile with Web and Background Worker

Problem: Heavy asynchronous tasks block the web dyno and trigger H12 request timeouts.

Solution: Separate HTTP web serving from background worker processes in Procfile:

web: node dist/server.js
worker: node dist/worker.js
release: npx prisma migrate deploy

Scale dynos: heroku ps:scale web=2 worker=1

Best Practices

Do:

• Use the release: phase in Procfile to run database migrations before routing traffic to new dynos.
• Configure WEB_CONCURRENCY to match dyno memory capacity and prevent R14 (Memory Quota Exceeded) errors.
• Bind to $PORT provided by Heroku; never hardcode HTTP port numbers in web applications.
• Use Heroku Review Apps in GitHub pull request workflows for stakeholder review.

Don't:

• Store uploaded user files on dyno local filesystems; dynos are ephemeral—use AWS S3 or Cloudflare R2.
• Run long-running CPU tasks in web dynos; offload to background worker dynos.
• Leave development add-ons on production apps; upgrade to production-tier Postgres with automated failover.

Troubleshooting

References

• Heroku Documentation