Microservices Are ‘Easy’, Dependencies Are Hard
summary
Argues that microservices ease development but break CI/CD unless pipelines are redesigned; recommends using multi-stage Docker builds and kind to run a full k8s cluster in CI, running Buildkite agents on the Kubernetes cluster, and keeping pipelines as code to avoid copy‑paste. Claims this approach scales from a few services up to ~100, speeds feedback via cached layers, enables full integration tests with TestCafe or Cypress, and uses feature flags and coordinated deploys to deploy fast and roll back reliably.
AI-written summary · about
chapters (20)
- 0:00Intro
- 2:09Kubernetes
- 2:54Agenda
- 3:46Past
- 5:10Pipelines
- 6:46CICD Philosophy
- 8:08Why do I need a CICD
- 9:07Our goal
- 9:44Dont repeat yourself dry
- 10:51Running locally
- 11:51Easy to get started
- 12:37Change the pipeline
- 12:57Advices
- 13:17Multistage build
- 16:37Pipeline
- 18:26Feature Flags
- 20:15Deploy Together
- 21:43Final Tips
- 23:22Confidence
- 24:31Conclusion
related talks
37mKnee Deep in Microservices
02h 38mMicroservices in Production: Kubernetes
28mContinuous Integration Testing: Fully Test your Microservices Application, Early and Often
20m3 reasons why micro-services aren't the miracle cure
29mDevOps and Microservices Better Together
01h 02mDesigning Microservices
40mMicroservice Continuous Delivery
48mChallenges in Implementing MicroServices