Reliability and speed by 'push to main' mentality - Håvard Høiby - NDC Oslo 2024

Maintain only production environment, no staging/development environments - this reduces complexity and forces better practices

Push directly to main/master branch from day one - enables faster integration and learning from production workloads

Break changes into small, atomic increments rather than large PRs:

• Reduces risk
• Makes rollbacks easier
• Enables faster feedback
• Simplifies code review

Use feature flags extensively:

• Test new features in production
• Control rollout
• Easy rollback capability
• Decouple deployment from release

Employ expand/contract pattern for breaking changes:

• Add new version alongside old
• Gradually migrate traffic
• Remove old version when safe

Embrace incidents and errors:

• Build blameless culture
• Learn from mistakes
• Monitor application metrics over infrastructure
• Have clear incident response process

Write “dumb code”:

• Keep it simple
• Minimize abstractions
• Make it easy to understand and modify
• Focus on explicit over clever

Enable developers to be autonomous:

• Trust teams to deploy
• Give access to production tools
• Allow direct fixes without process overhead
• Empower decision making

Test based on impact and criticality:

• More testing for critical paths
• Less ceremony for low-risk changes
• Use production data for testing
• Monitor real usage over synthetic tests

Optimize for speed of integration:

• Minimize time from idea to production
• Remove artificial process barriers
• Learn from production quickly
• Continuously deliver small changes