Improving Front-End Service Availability in Meituan Financial Payments

This article is based on Meituan senior engineer Tian Yang’s talk at Meituan Tech Salon #31 titled “Offline Payment Million‑Order Service – Front‑and‑Back End Architecture Practice”.

Introduction

In 2017 Meituan Finance’s front‑end team faced many generic problems that affected service reliability. The team shared lessons from a technical salon and aims to help developers better understand front‑end availability.

Business Overview

Mobile payment has become the primary choice for users, and Meituan Finance’s payment service grew to a ten‑million‑orders‑per‑day scale within a year – a “rocket‑shaped” growth with an “ice‑berg” user experience. Although users only see a simple QR‑code page, the backend involves many merchant, supply‑chain, and platform systems, demanding high stability.

Definition of Front‑End Availability

Availability is usually calculated as %availability = (Total Elapsed Time – Sum of Inoperative Times) / Total Elapsed Time, i.e., MTBF / (MTBF + MTTR). Front‑end availability differs from back‑end because user environments vary widely (different phones, OS versions, networks).

Key Factors Affecting Availability Client upgrade compatibility issues. Code optimization and service migration side‑effects. External dependency failures (APIs, infrastructure). Incomplete execution of development processes. These can be grouped into internal problems (“I hit a pig”) and external problems (“pig hit me”). Improving Internal Node Availability

Standardize processes (SOP, release criteria).

Choose reasonable solutions; prefer simple designs for quick fallback.

Enforce code standards to avoid trivial bugs.

Additional suggestions include unit and automated testing, especially on Android containers. External Link Availability Static resources may suffer from load failures, network hijacking, or execution errors. Mitigations:

Retry failed CSS loads; fallback to minimal styling.

When core JS fails, downgrade to an MVP version.

Use HTTPS to prevent hijacking; monitor at provincial/city level.

Collect user environment data via Meituan’s “120” incident‑response mode.

For interface services, front‑end should collaborate with back‑end to resolve issues quickly, using end‑to‑end monitoring and degradation. End‑to‑End Monitoring and Degradation Meituan uses the open‑source CAT system to define coverage per page, drill down by region, and trigger automatic fallback when CDN or services are unavailable. Fault‑Drill Necessity Regularly simulate failures (e.g., kill back‑end APIs, block static resources) to verify monitoring and degradation mechanisms. Standard Process for Assurance

Pre‑release: enforce code, branch, test, and release standards.

During release: implement monitoring and degradation plans; rehearse frequently.

Post‑release: conduct case studies and enforce actionable follow‑ups.

Conclusion Improving front‑end service availability relies on two pillars: disciplined process execution and comprehensive engineering automation. Teams should build generic solutions covering the whole lifecycle—from front‑end to back‑end, development, testing, and deployment—to reduce repetitive work and focus on core business.