Building a Culture of Continuous Delivery
summary
Netflix centralizes delivery into a control plane that runs Spinnaker ~95% of deployments, bakes artifacts with Jenkins/Gradle, and pushes canaries that take live traffic; the container runtime Titus launches 500,000 containers per day (on ~8,000 VMs, ~450,000 vCPUs). Feature flags and config rollouts match code pipelines, chaos testing is opt-out, and region evacuations now complete in under 10 minutes. Result: thousands of production changes per day with fewer large rollbacks.
AI-written summary · about
chapters (41)
- 0:00Introduction
- 1:28Safety Reliability
- 3:11Producing Good Content
- 5:36Strong Technology
- 17:01Centralized Teams
- 17:23Central Platform Teams
- 17:53Outreach
- 18:45Tools
- 19:12Spinnaker
- 19:22Atlas
- 22:07Ansible
- 23:22Terraform
- 24:30Deployment
- 26:29Chaos Test
- 27:44Recap
- 27:59Common Problems
- 28:23Additional Languages
- 29:00Chaos Monkey
- 32:40Dont Wait Forever
- 33:46Best Practices
- 35:05Pipelines
- 35:40Canaries
- 36:36Metrics
- 37:12Newt
- 39:21Jenkins is a workhorse
- 39:57Centralized Jenkins team
- 40:27Whats the alternative
- 42:24Control Planes
- 44:12Contact Information
- 44:37DB Schema
- 45:11Deploy New Code
- 46:01Additional Controls
- 47:03Metatron
- 48:20Mobile Apps
- 49:32iOS
- 51:31Tamper detection
- 52:34We do everything
- 52:51Load balancers
- 53:34Eureka
- 53:56Biggest Failure
- 56:01Incident Reviews
related talks
29mFast, Safe, and Continuous -- Picking all 3 at Netflix
29mHow Netflix Thinks About DevOps
31mHow Netflix Thinks About DevOps
53mContinuous Deployment with Spinnaker — Automate All Your Deployments!
52mContinuous Deployment to the Cloud using Spinnaker
36mSpinnaker - Land of a 1000 Builds
27mDeploying Docker Containers with Netflix Spinnaker on Kubernetes and Titus
48mContinuous Delivery with Spinnaker