JD Logistics Automated Testing Platform “Linglong”: Architecture, Features, and Enhancements

JD Logistics’ quality department, one of the earliest teams in JD Group to implement comprehensive automated testing, has been responsible for quality assurance across all logistics product lines. Over years of technical accumulation, the team identified pain points in quality, efficiency, process improvement, and technology, leading to the development of various automation, performance, and CI tools.

Daily practice relies on continuous integration (unit, API, UI) for regression testing, with large‑scale automated regression runs before each Tuesday and Thursday release. Test scripts are also shared with developers for self‑testing, raising overall quality awareness and reducing defect‑fix costs.

Background : Rapid growth of JD Logistics’ order fulfillment, warehousing, delivery, and customer service creates frequent, fast‑changing business requirements. Manual testing cannot keep pace, necessitating automation to reduce repetitive work and improve testing efficiency.

Current Platform Limitations (after 4 years of use): Task assignment to execution agents often times out, making troubleshooting labor‑intensive and under‑utilizing resources due to browser/JDK version mismatches. Single‑task execution hampers efficiency; the platform cannot run a collection of tasks together, forcing all tasks under a plan to execute sequentially. Support is limited to Java, while many lightweight testing frameworks in other languages are unavailable.

To address these issues, the team upgraded the platform, naming the new system “Linglong” (meaning delicate and agile).

New Features of Linglong : Improved task scheduling that eliminates slow agent dispatch and version‑binding problems. Ability to combine multiple test tasks and run them concurrently, boosting execution efficiency. Data‑driven testing for richer scenario coverage. Support for Python and mobile testing (in development), enabling multi‑language and diverse test types. End‑to‑end logistics business flow testing (in development) using pipeline‑based execution of full‑process scenarios.

System Architecture includes: WebUI layer : retains test plan, task, and case management, adds permission control per department, and provides statistical reports by department, system, and personnel. Task Scheduling : replaces STAF with Jenkins for dispatch; MongoDB serves as the document‑oriented database, and CI triggers are exposed via APIs. Test Execution Agents : replaces unreliable physical agents with KVM‑based virtual machines, upgrades JDK to 1.8 and drivers to the latest versions, and adds Python support for test frameworks.

Core Functionalities :

Test Cases : automatically scanned from annotated classes; metadata (name, scenario key, author, system, etc.) is extracted and stored in a collection.

Test Plans : configure related settings; both cases and plans are stored as tree structures in separate collections, allowing nested modules and tasks.

Test Scenes : combine selected cases and data to generate automated tasks; scene data is stored as a list of maps within the task document.

Multi‑Task Execution : agents allocate tasks based on availability or specific IP binding, with a waiting queue for busy agents.

Agent Program : assembles test cases, executes them via shell scripts, collects logs, extracts checkpoints, writes results back to the database, and integrates with Jenkins build logs.

Test Email : aggregates results from all tasks in a group and sends a summary email with detailed information for traceability.

After the first phase of enhancements, Linglong provides richer, more stable, and higher‑efficiency automated testing services for JD Logistics’ fast‑growing business. The second phase will focus on end‑to‑end business‑chain automation and mobile testing.

The team invites peers to exchange technical experiences, share successful cases, and collaborate.