Release Early, Often, Stable
summary
A practical approach to doing very frequent, safe releases for many independent microservices is described: achieving about 500 deploys/day by making every commit a potential release candidate. The setup uses Kafka as an event bus, TeamCity for builds, an artifact repository, dynamic feature flags, automated tests, monitoring, fast rollbacks, and an internal pipeline system to reduce risk and accelerate feedback, assuming proper testing and metrics are in place.
AI-written summary · about
chapters (26)
- 0:00Early release philosophy
- 1:02Two release models
- 2:26Optimal web release frequency
- 3:53Choosing release frequency
- 4:37Risks of frequent releases
- 6:24Why frequent releases matter
- 7:29Dev optimization via releases
- 8:39Continuous Integration
- 10:26From CI to Continuous Delivery
- 11:12Feature flags for safe deploy
- 12:59Key release ingredients
- 15:31Simpler version control
- 16:35Frequent releases and DevOps
- 18:26Microservices foundation
- 19:09Wix CI/CD architecture
- 25:23Wix CI/CD workflow
- 27:31Unified dev/test environment
- 28:11Gradual feature rollouts
- 29:38Flag management with Petric and A/B
- 30:21Bug monitoring and response
- 33:33Why it works and motivation
- 34:18Q&A: code review
- 36:52Q&A: linting local vs CI
- 38:17Q&A: Git Flow issues
- 40:52Q&A: Git Flow evolution
- 42:36Q&A: CI/CD in other firms
related talks
44mI Deploy On Fridays
38mBuilding a high quality product
41m100 releases per day
28mHow to ship apps better, faster, stronger
46mContinuous Deployment Workflows
05mRelease management as a catalyst for CI/CD
51mBetter software, faster: principles of Continuous Delivery and DevOps
34mEnabling a Continuous Deployment Mindset