Applying APIServer to Baidu Waimai CMDB for Machine Resource Lifecycle Management

This paper starts from the practical scenario of Baidu Waimai’s CMDB and describes the lifecycle of machine‑related resources as well as how each stage is managed. It then introduces APIServer, borrowed from Kubernetes, as a core component of the CMDB.

APIServer treats every resource as an object stored in etcd and provides a unified RESTful API for create, read, update, delete, and watch operations. By abstracting machines, IPs, DNS records, and other assets as objects, the CMDB can perform lifecycle actions through standard API calls.

The Watch mechanism, inherited from Kubernetes, enables asynchronous, publish‑subscribe style notifications. Components subscribe to changes via APIServer, receiving updates when resources are created, modified, or deleted, which improves performance and data consistency.

The CMDB architecture uses APIServer to store machine objects in etcd, exposing CRUD and watch APIs to other services such as agents, workflow engines, DNS, and Puppet. The workflow for machine lifecycle includes five stages: synchronization, allocation, initialization, change, and repair, each implemented via specific API calls.

For the complex hostname‑change process, the workflow proceeds through order submission, approval, DNS/etcd update, local host update, and order completion, with each step triggering watch events and recording events for traceability.

In summary, APIServer provides a unified, object‑oriented interface that simplifies resource management, enables reliable watch‑based communication, and supports the current CMDB features (machine sync, repair, queries, hostname changes) while outlining future work on broader machine‑change operations.