PostOps: A Non-Surgical Tale of Software, Fragility, and Reliability
summary
Argues that ops should shrink into software: adopt PostOps by automating toil, defining SLAs, and treating services (not servers) as the unit of work. Describes concrete systems and tools—GFS, Omega, Zookeeper, canaryed rollouts, Puppet/Chef—and a failure case where a singly‑homed global key‑value in RAM plus kernel updates makes epic syncs fail (per‑20‑minute crash probability ≈ 0.85). Recommends intent/constraint‑based config, automated canarying, and fewer humans doing repetitive fixes via site reliability engineering.
AI-written summary · about
chapters (21)
- 0:00Intro
- 2:43Machines Eating People
- 3:18The Human Brain
- 3:28Who am I
- 6:58Google Scale
- 11:57Everyone has the same problems
- 12:13I Hate Computers
- 13:16Failures
- 15:55Site Reliability Engineering
- 16:28Leather Jackets
- 22:30The Basics
- 29:06What weve learned
- 32:35Opportunity cost
- 36:11Evolution of systems
- 42:39People
- 49:20DevOps
- 50:19Organizational Structure
- 50:47Traditional Organizational Structure
- 53:54DevOps as a Set of Functions
- 55:16Embedding DevOps
- 56:06The Scientific Revolution
related talks
34mDevops or: How I Learned to Stop Worrying and Love the Ops Team
33mOperations: the last Mile for DevOps
49mOps and Operability
33mThe End of the Beginning
47mDevOps is for Product Engineers, too.
53mGround Truths of a Rugged DevOps Practitioner
36mPostgreSQL: Ups, DevOps...
26mMission Critical Mindset Building a Culture of Resilience in DevOps Teams