Portal: An Integrated Application Lifecycle Management Platform at Qunar

Background Introduction

Qunar is a technology‑focused company that, while pursuing rapid business growth, also emphasizes product quality and work efficiency, leading to the creation of many platform tools and management systems. These tools bring convenience but also introduce several critical problems.

1) High learning cost – Engineers must not only focus on their core tasks but also learn to use numerous platform tools, which reduces overall efficiency.

2) Switching overhead – Moving between tools (e.g., PMO for demand, Application Center for appcode, Application Tree for machines, Release System for jobs, Ops platform for permissions, Monitoring platform for metrics) increases time cost and error probability.

3) Data inconsistency – Dependencies between systems cause information duplication and mismatched data, such as manual copying of machine information from the Application Tree to other systems.

4) Fragmented operations – Different ops departments provide services on different dimensions, making end‑to‑end reliability hard to guarantee; a unified messaging mechanism is needed to stabilize upstream/downstream data flow.

Architecture Introduction

Analyzing the current state revealed that Ops provides rich machine‑centric capabilities, the Application Center manages apps via a unique appcode, and the Release System focuses on engineering units. By extending the appcode across Ops and Release, we created a single identifier that links all services. We also built an information center where each system publishes change events to a message hub, allowing downstream services to consume updates and maintain data consistency. A unified web portal now offers one‑stop registration, operation, release, and machine management, eliminating the need for users to switch between multiple systems.

The first‑phase architecture diagram shows integration of existing functions into a single user‑service platform. Although the initial step still relies on manual actions, future phases will move toward event‑driven automation and scenario‑driven intelligence as data accumulates.

Feature Introduction

1. Centralized Application Information Display

Each application’s details (admin, tree node, creation time, code path, deployment path, port, machine environment, logs, etc.) are now viewable in one place, eliminating the need to hop between Application Tree, Application Center, and other tools. Three modules are presented:

• Left side – static configuration and deployment parameters for quick overview and maintenance.

• Upper right – recent change events (FULL_GC, releases, config changes, scheduled tasks, machine starts, custom events) to help users track activity and diagnose issues.

• Lower right – machine environment information, allowing admins to edit ownership, grouping, batch, order, and view JDK/Tomcat versions; batch restart/shutdown actions reduce manual login risks.

2. Unified Application Management Entry

Portal provides a single entry to create or edit an application’s basic, release, and machine information, as well as to manage resources such as host provisioning, reclamation, scaling, and (in development) JDK upgrades. All actions are “what‑you‑see‑is‑what‑you‑get,” eliminating redundant data entry and reducing error rates.

3. Intelligent and Convenient Release Process

Previously, releases required multiple manual steps across different tools (schema creation, job definition, switching to the release system, etc.). Portal now lets users edit release parameters once, attach machines to the appropriate environment, and trigger a visualized release flow. A preview shows CI results, ports, paths, and machine batch details, enabling double‑check before deployment and simplifying troubleshooting.

4. Traceable Permission Management

Earlier, permission entry points were scattered across tools, each with its own workflow (e.g., job owner approval, Ops approval). Portal consolidates permission requests for configuration changes, application owners, members, and release roles, recording detailed approval logs for future audit. Qunar’s upcoming IAM system will further enable attribute‑, role‑, and group‑based access control.

Future Outlook

Portal’s release module now integrates CI (Cable), self‑service ops, and permission management, forming a basic DevOps loop. Future work will focus on three areas:

1) Functionality – Stabilize existing features while integrating additional tools such as Noah (environment management) and Odin (requirement management) to complete the DevOps toolchain.

2) Automation – Automate manual lifecycle steps (e.g., machine authorization, component upgrades, auto‑rollback on alerts, auto‑scaling based on traffic) by formalizing end‑to‑end processes.

3) Intelligence – Leverage accumulated, traceable data for business analysis, root‑cause mining, and scenario‑driven intelligent services.