How to Achieve Fast, Reliable Continuous Delivery with Microservices and DevOps

Introduction

The speaker, a senior technical expert at Alibaba, shares practical experiences on continuous delivery and microservices, focusing on how to make every code commit deployable while maintaining speed and reliability.

Continuous Delivery Concept

Continuous delivery is defined as a systematic approach to releasing reliable software. It distinguishes between the goal (what counts as continuous delivery) and the practices needed to achieve it. A typical development flow includes requirement analysis, coding, testing, and deployment, often illustrated with workflow diagrams.

Problems and Solutions

Key problems include lengthy regression testing before releases, which can take days and create bottlenecks. Solutions involve breaking development into short iterations, automating tests, and reducing manual steps. Automated testing (unit, integration, functional/end‑to‑end) is essential, with tools such as JUnit, Spring Test, Cucumber, Selenium, and language‑specific frameworks.

Testing Levels

Different test layers serve distinct purposes:

Unit tests run quickly, cover small code units, and enable safe refactoring.

Integration tests verify component interaction without test doubles, using tools like Moco and Pact.

Functional/acceptance tests exercise the system through its UI, often with Cucumber or Robot Framework.

End‑to‑end tests combine all services to validate overall behavior.

Maintaining test code quality and using IDE assistance are recommended to keep test maintenance costs low.

Continuous Integration Pipeline

Each code commit triggers a pipeline that compiles, runs static checks, executes tests, packages artifacts, and optionally runs acceptance tests before deployment. Tools such as Jenkins or Alibaba's CRP platform can orchestrate these stages. Important practices include frequent commits, versioned build artifacts, and staged execution where later stages depend on earlier success.

DevOps and Infrastructure as Code

DevOps encourages developers to deploy without waiting for a separate operations team. Infrastructure‑as‑code (IaC) ensures environments are reproducible and versioned. Ansible is highlighted as an agent‑less IaC tool offering inventory management, idempotent modules (e.g., lineinfile, service), and playbooks that describe desired states.

Immutable Servers and Docker

Immutable server concepts avoid in‑place changes; instead, new instances are provisioned from a known image. Docker containers provide lightweight, layered images that speed up builds and deployments. A typical Docker workflow builds a base image (e.g., Java + Tomcat) and adds the application WAR via a Dockerfile, then pushes the image to a registry.

Microservices Interaction

Microservices are small, autonomous services that complement continuous delivery by allowing isolated releases. However, they increase deployment complexity, making robust CI/CD pipelines and automated testing even more critical. Progressive microservice adoption, such as adding new features as separate services behind an Nginx proxy, is recommended.

Q&A Highlights

Q1: Docker deployment considerations – use environment variables for configuration and be aware of performance impacts when nesting Docker.

Q2: Key CI metrics – test coverage, code complexity, and static analysis; failures are displayed on a big screen to prompt immediate fixes.

Q3: Package contents – can be a simple WAR or a zip containing the WAR and install scripts.

Q4: Windows support – Ansible works on Windows; Docker on Windows is still maturing.

Conclusion

The talk emphasizes that achieving continuous delivery requires automated testing, a well‑designed CI pipeline, IaC tools like Ansible, and containerization with Docker. When these practices are in place, teams can safely adopt microservices and accelerate release cycles.