It's Not Continuous Delivery if You Can't Deploy Right Now
summary
Claim: continuous delivery only counts if the team can deploy right now to production. Explains daily trunk commits (no long-lived feature branches), CI servers running all tests, and parallel fan-out/fan-in pipelines that run SAST, dependency scans, load and spike tests, plus canary and dark launches. Gives Heartbleed fixes in hours, a GitHub experiment that improved merge performance orders of magnitude, a study finding ~120 libs with ~21% vulnerable, and a Night Capital manual deploy causing a $440M loss.
AI-written summary · about
chapters (21)
- 0:00Introduction
- 8:25Continuous Integration
- 14:30Pipeline
- 15:38Security
- 18:24Performance
- 20:42Parallel Processing
- 24:58Livery
- 28:39User Stories
- 29:11Feature Toggles
- 31:27Deploy vs Release
- 33:16Backend Upgrades
- 34:41Managing Risk
- 37:03Optimize for MTTR
- 39:00Its Not Continuous Delivery
- 39:15Canary Release
- 40:06Dark Launching
- 41:38Feedback
- 42:25Alerts
- 43:25Gartner Hype Curve
- 44:10Night Capital
- 45:36Summary
related talks
55mThe Rationale for Continuous Delivery
33mContinuous Delivery - What does good really look like?
32mContinuous Delivery Sounds Great But It Won’t Work Here
28mMinimum Viable Continuous Delivery
47mContinuous Delivery Warstories
11mCreate unfair competitive advantage w/CI/CD Pipelines
30mModern Continuous Delivery
41mAll Things Continuous