Ensuring API Idempotency: Primary Keys, Optimistic Locks, and Token Strategies

Using Unique Database Primary Keys

Database primary keys enforce uniqueness, guaranteeing that each insert operation is idempotent because only one record with a given key can exist. To make this work in distributed systems, a distributed ID (instead of an auto‑increment key) is used as the primary key, ensuring global uniqueness. This approach is most suitable for create and delete operations.

Applying Optimistic Locking

An optimistic‑lock scheme adds a version field to the target table. Each update includes the current version value as a match condition; the database only updates the row if the version matches the stored one, then increments the version. This method is appropriate for update operations but requires an extra column in the business table.

Enabling Anti‑Repeat Token (Idempotency Token)

For scenarios such as rapid user clicks or retry after timeout, the client first requests a global token from the backend and includes it in subsequent API calls. The backend stores the token as a key and the user information as the value in Redis. When a request arrives, the backend checks whether the token exists and whether the stored value matches; if so, the operation proceeds, otherwise the request is rejected as a duplicate. This token‑based solution works for create, update, and delete operations.

These three techniques together provide a comprehensive toolkit for implementing reliable API idempotency.