People over Process over GitOps
summary
Argues that organizational habits beat tooling: Kubernetes and GitOps help but don’t fix people or process, and ExternalDNS, Flux and scripted “robots” only solve plumbing if workflows match goals. Recommends measuring DORA metrics — deployment frequency, lead time, restore time, change failure rate, reliability — notes reliability only improves outcomes when the other four are strong. Advises decoupling CI/CD from deployment, using rolling updates and prestop hooks for zero downtime, and avoiding mandatory staging on every change because staging adds minutes and increases lead time.
AI-written summary · about
chapters (4)
- 0:00Start
- 8:15Deployment frequency Lead time for changes Time to restore service Change failure rate Reliability
- 11:14Decouple Integration, Delivery, and Deployment
- 13:53Maybe you don't need that Stage environment for every change?
related talks
27mFrom “No Time For GitOps” To Enterprise Adoption: Selling Flux The Human Way
36mHuman Factors and DevOps
34mThe Return of the Dull Stack Engineer
14mA DevOps Mindset: Transforming Organizational Culture
05mThe Road to DevOps Success: It’s Human
09mWhat Would Taylor Swift Do? How to Convince Yourself to Use GitOps
24mWhen Git Push Meets Human Pushback: The Culture of GitOpS
23mTrue DevSecOps with People, Processes, and Technologies