Building a DevOps Toolchain for R&D Efficiency at Yanxuan

Yanxuan has grown over several years, serving thousands of customers and expanding its R&D staff from dozens to hundreds. The increasing demand for faster project delivery has raised the challenge of improving product development efficiency, making DevOps an unavoidable focus.

DevOps is described as a cyclical, progressive process that combines culture and tools. By establishing a toolchain that fits the organization’s culture, standardizing collaboration processes, and automating workflows, teams can accelerate product development while maintaining quality.

Reasons for revamping the DevOps toolchain (since 2019):

Rapid requirement changes increase efficiency challenges.

Historical standards and processes decay, raising collaboration costs.

The architecture is shifting toward container‑based micro‑services, requiring the toolchain to adapt to containers and cloud environments.

Starting points – core DevOps phases:

Plan : Project management functions.

Build (Code) : Development functions.

Test : Quality assurance functions.

Release (Deploy) : QA and operations functions.

Monitor : Involves development, QA, and operations.

In Yanxuan, building the DevOps toolchain means empowering roles such as PM, QA, and SRE through platforms, making the project team lighter and more agile while preserving quality.

CMDB : The CMDB manages the relationships among products, resources, and personnel, providing a unified model for various tools and aligning terminology across development, QA, and SRE.

Monitoring system is divided into three layers:

Resource monitoring : Uses open‑falcon and Grafana to visualize resource usage.

Application monitoring : Based on Yanxuan’s Caesar system (a customized Pinpoint), covering end‑to‑end traces for apps, web front‑ends, and back‑end services.

Business monitoring : Real‑time monitoring built on a data‑lake concept, supporting massive log/binlog ingestion, high‑concurrency dashboards, and alerting during peak traffic.

CI/CD challenges identified:

Inconsistent branch management strategies across teams.

Multiple CI/CD tools (GitLab‑CI, Jenkins) leading to fragmented pipelines.

Insufficient automated test coverage, requiring manual intervention.

The solution centers on unifying the concept of “artifact.” Artifacts are produced early in the CI stage (GitLab‑CI), ensuring that QA validates the same package that will be deployed to production. The CI pipeline handles code compilation, packaging, linting, and basic testing, while QA runs automated validation on the artifact. Deployment is managed by Yanxuan’s self‑developed Opera release platform.

R&D Efficiency Platform – Overmind:

One‑stop : End‑to‑end continuous delivery pipeline, reducing platform switching costs.

Visualization : Full‑process visibility to identify bottlenecks and ensure compliance.

Full‑process metrics : Measures and correlates efficiency data across domains, enabling data‑driven improvement.

Future work includes risk control during changes, environment governance to avoid conflicts, and standardizing plugins for easier capability iteration.

Author : Saga, Technical Director of Yanxuan’s Basic Technology Department, focusing on business middle‑platform and DevOps efficiency improvements.