My notes from the DevOps Handbook

by Gene Kim, Jez Humble, Patrick Debois, John Willis

Downward spiral continued

• product delivery cycles -> slower
• fewer projects undertaken and less ambitious

• slower and weaker feedback

• downward spiral - everywhere

• in countless organizations of all types and sizes

• projects are critical - mechanism of change in the organizations

  • approval
  • budget
  • accountability

• business leaders are far more reliant on IT than they think A. firms with IT weaknesses B. firms with other than IT weakness C. clean firms

  • A -> 8x CEO turnover of C
  • A -> 4x CFO turnover of C

• human and economic costs

• burnout
• fatigue
• cynicism
• powerless

• afraid to do the right thing because of punishment

• learned helplessness - unwillingness or inability of people to act in a way that avoids the same problem in the future

• for employees

  • long hours
  • weekends at work
  • reduced quality of life
  • losing the best people

• economic

  • $2.6 trillion of value creation = annual GDP of France
  • 2011 -> world's 5% GDP -> IT
  • 50% of that -> operation/maintenance costs
  • 1/3 of that 50% -> urgent, unplanned work

• ethics of DevOps

DevOps enables us to improve organizational goals, performance and improve the human condition

• breaking the downward spiral - ideally

  • small teams of developers
  • independent implementation of features
  • validate feature correctness
  • production-like environments
  • quick, safe and secure deployment
  • deployment - routine and predictable
  • deployments during the day, not at night
  • fast feedback loops at every step of the process
  • fast and automated tests

• automated testing

  • quick mistake discovery
  • fast fixes
  • genuine learning
  • problems fixed as they are found
• pervasive telemetry in production - monitoring
• everyone feels productive
  • architecture allows small teams to work safely and independently
  • architecturally decoupled
  • small batches quickly delivered to customers
• dark launch
  • all the code for the feature in production
  • visible only to internal employees
  • small batch of users to test
  • feature toggle in the configuration - rollout
  • controlled, predictable, reversible, low stress
• organizational learning
  • prevent in the future
  • fix faster in the future
• constant learning
• hypothesis-driven culture
• scientific method - nothing is taken for granted
• measuring in experiments
• value everyone's time
  • don't build features customers don't want
  • deploy code that works
  • fix something that actually is the cause of the problem
• teams that:
  • are responsible for their goals
  • intact
  • iterating and improving
  • learning
• high-trust, collaborative culture where people are rewarded for taking risks
  • open about problems
• full ownership of the quality of the work
  • automated testing
  • peer review
• in case something goes wrong
  • blameless post-mortem
  • what caused the accident
  • how to prevent it
• planned fails to grow system resilience
• everybody has ownership in their work
  • work that matters
  • organization that is winning in the marketplace