Design and Implementation of a Traffic Replay Platform for Interface and UI Automation at Ctrip

When Ctrip’s product releases grew in complexity, manual regression testing became time‑consuming and slowed release cycles, prompting the development of a traffic replay platform that minimizes user intervention.

The platform extracts production logs from ElasticSearch, using its powerful query and aggregation capabilities to treat log entries as traffic data for testing.

Interface Automation is divided into three stages: (1) building request messages by normalizing ES records with tokens and business tags; (2) executing requests by replaying logged responses, supporting concurrent mock testing and dual‑environment comparison; and (3) analyzing results by aggregating responses, generating reports with diff highlighting, and linking to original logs. The process is integrated into CI, delivering analysis reports within about three minutes per commit.

To address the maintenance burden of growing business tags, the platform aggregates ES query results, presenting the most frequent tag combinations to users, reducing manual query composition.

Image Comparison Testing (UI Automation) replaces fragile element‑based scripts with a log‑driven approach: server‑side request/response pairs are mocked for the UI, results are shown as side‑by‑side screenshots, pixel‑level diff matrices are overlaid to highlight changes, and Puppeteer replaces Selenium for headless Chrome control. Additional features include OCR‑based text diff, modular screenshot slicing, and multi‑server concurrent execution.

Since deployment, the platform has executed over 50,000 runs, uncovered more than 3,000 issues (averaging 10 per day), and increased release frequency from once a week to three times a week, demonstrating significant efficiency gains while acknowledging that no tool can guarantee 100% quality.