Les objets durables expliqués

Serverless functions, like Cloudflare Workers, are typically stateless. This means each request is processed independently, without remembering past interactions.

While great for scalability and cost-efficiency, what if you need to maintain a consistent piece of information across multiple requests or users globally? Think about a live chat room, a game lobby, or a shared counter.

Cloudflare Durable Objects solve this fundamental challenge. They are a powerful primitive for building stateful serverless applications directly at the edge.

• They provide a single-instance, globally consistent object.
• Each Durable Object instance has a unique ID and its own private storage.
• They act like tiny, persistent servers that live at the edge.

The core concept behind Durable Objects is their unique identity. When you create or retrieve a Durable Object using a specific ID, Cloudflare ensures that only one instance of that object exists and runs globally.

No matter where a request originates from (e.g., Tokyo, New York, London), if it targets the same Durable Object ID, it will be routed to that single, consistent instance. This eliminates the need for complex distributed synchronization.

A Durable Object is defined as a standard JavaScript class. It must have:

• A constructor(state, env) method, where state provides access to its storage and env holds environment variables.
• An async fetch(request) method, which processes incoming HTTP requests, similar to a regular Worker.

It's essentially a self-contained, stateful mini-Worker.

Each Durable Object instance has its own private, durable key-value storage, accessible via this.state.storage. This allows the object to persist data across invocations and even across restarts.

• await this.state.storage.put('key', value): Stores data.
• await this.state.storage.get('key'): Retrieves data.
• await this.state.storage.delete('key'): Removes data.

Values can be any JSON-serializable type, making it versatile for various data needs.

This example shows a Cloudflare Worker interacting with a simple Counter Durable Object. The DO maintains a count that persists across requests and is globally consistent.

To run this, you'd define the Counter Durable Object binding in your wrangler.toml file.

1. The Counter class defines the Durable Object's behavior, handling /increment, /decrement, and /get routes and persisting its value.

2. The main Worker's fetch handler:

• It uses env.COUNTER.idFromName("global-counter") to obtain a deterministic ID for our counter.
• env.COUNTER.get(id) returns a stub, which is a network proxy to the Durable Object.
• stub.fetch(request) forwards the original request to the Durable Object instance, allowing it to process the request and update its state.

Imagine users from different locations (e.g., Europe, Asia, Americas) all interacting with the /increment endpoint of your Worker. Because they all target the "global-counter" ID, Cloudflare guarantees they will all interact with the exact same instance of the Counter Durable Object.

This ensures that the count is always consistent, regardless of the user's location, simplifying complex distributed state management that would otherwise be very challenging.

Durable Objects are ideal for scenarios requiring strong consistency and shared, mutable state at the edge:

• Real-time Collaboration: Shared whiteboards, document editing.
• Gaming: Game lobbies, player state, matchmaking.
• Live Chat: Chat rooms, message queues for real-time communication.
• Stateful APIs: Managing user sessions, shopping carts, or unique resource locks.

They bring traditional stateful patterns to the serverless edge, making complex applications simpler to build.

In this lesson, we explored Cloudflare Durable Objects, a powerful tool for stateful serverless applications.

• They provide globally consistent, single-instance state at the edge.
• Each DO has a unique ID and private, durable storage (state.storage).
• Workers interact with DOs by obtaining stubs and forwarding requests.
• They are ideal for real-time, collaborative, and stateful applications that demand strong consistency.

Durable Objects bridge the gap between stateless serverless and traditional stateful services, simplifying complex distributed systems.