Continuous Delivery for Legacy Code
summary
Five million lines across Classic ASP, .NET, VBA, T‑SQL and more, no VCS, copy‑to‑prod deployments and 4,000 tables with ~2,000 huge stored procedures become a measured problem: put everything into Git, add a CI pipeline that compiles, parameterized infra scripts, automated smoke and performance tests, code cartter hotspot visuals, randomized categorization tests (approval tests) and Red production metrics in Grafana. Cycle time drops from six months to about 60 minutes, duplicate installs and a million hidden lines get pruned, and automated monitoring gives enough confidence for regular deployments.
AI-written summary · about
chapters (26)
- 0:00Introduction
- 1:10Legacy vs low-quality software
- 3:19Case study: broken legacy system
- 9:41Legacy paradox and business impact
- 10:46Fixing the foundation with version control
- 12:56Automated deployment pipeline
- 13:40Hotspots mindset for refactoring
- 14:21CodeCutter: visualizing quality
- 17:36Quality views for management
- 19:41Architecture tests for legacy
- 20:25Broken Window Theory focus
- 21:32Start with critical tests
- 23:17Sprouting tested code
- 24:20Testing untestable code
- 26:27Refactoring: coupling and cohesion
- 27:51Seams and feature flags
- 28:56Strangler Fig pattern
- 30:00RED metrics in production
- 34:13Continuous delivery pipeline
- 35:39Cycle time reduction
- 36:24Why legacy happens
- 39:18Changing behavior with Atomic Habits
- 42:03Summary and resources
- 42:37Q&A: Duplicated legacy code
- 45:13Q&A: Untestable code side effects
- 47:21Q&A: Behavioral change
related talks
49mContinuous Delivery for Legacy Code
50mContinuously delivering software to big brands
01h 01mA team's transition to Continuous Delivery
47mContinuous Delivery Warstories
32mContinuous Delivery Sounds Great But It Won’t Work Here
20mLessons Learned from Building a Deployment Process for 100+ Devs at a Large Bank
01h 10mContinuous Integration for the BI Developer
48mContinuous Delivery: Tales from Windowsland