Continuous Deployment Practice Using DevOps: CI/CD Pipeline with Jenkins, Docker, and Kubernetes

The department project required managing multiple services with complex underlying environments, prompting the adoption of a DevOps approach that automates code building, testing, image packaging, and deployment.

Project management is divided into code management (including branch control), image management (versioning), and configuration management (deployment servers and parameters). The DevOps workflow implements continuous integration with Bitbucket, continuous testing with Jenkins and JUnit, continuous building with Docker and Harbor, and continuous deployment using configuration management and Kubernetes.

The implementation process uses Jenkins to orchestrate each stage: continuous integration pulls code and configuration from Bitbucket, continuous testing runs Go unit tests with the beego framework, continuous building creates Docker images pushed to Harbor, and continuous deployment pushes containers via SSH to target servers and starts pods with Kubernetes YAML files.

The logical architecture consists of four layers: code management (branch development and config repository), image management (base and application images), deployment management (Dockerfile, YAML, scripts), and deployment routing (environment‑specific IP routing, with future routing‑based selection).

Technically, image management separates base images (OS and dependencies) from four application images (deployment platform, management platform, browser platform, portal). All environment and configuration files are centrally maintained.

The deployment architecture relies on Jenkins scheduling, two Kubernetes master clusters, a Harbor server, and SSH plugins to push images and execute deployment scripts, enabling one‑click deployment of the latest code to development or test environments.

In conclusion, the current setup supports one‑click deployment of updated pods to both development and test environments, reducing the code‑to‑runtime cycle to about two minutes, with ongoing improvements and an invitation for interested peers to collaborate.