Deploy production Deploy to Production Deploy ទៅ production

• Input validation on all user-provided data
• Per-user sandbox isolation (never a shared sandbox)
• Secrets stay in the Worker — not in sandbox env vars
• Authentication gates access to code execution
• Rate limiting configured (Workers rate limiting or a KV counter)

• Sandboxes are destroyed with destroy() when no longer needed
• setKeepAlive(true) used for long-running operations, disabled when done
• Error handling and timeouts on all exec() calls
• Logging enabled in wrangler.jsonc

Bước 1: Deploy ứng dụng Step 1: Deploy Your Application ជំហាន 1: Deploy កម្មវិធី

Chúng ta đang xây What we're building អ្វីដែលយើងកំពុងសង់

Your sandbox application running on Cloudflare's global network.

Vì sao quan trọng Why this matters ហេតុអ្វីសំខាន់

Production deployment makes your app available worldwide via Cloudflare's edge, with automatic scaling and no infrastructure to manage.

npx wrangler login

npm run deploy

⛅️ wrangler 4.x.x Uploading worker script… Uploading container image…
(this takes 1-2 minutes on first deploy) ✨ Success! Published my-sandbox
(3.14s) https://my-sandbox.your-subdomain.workers.dev

curl https://my-sandbox.your-subdomain.workers.dev/run

Bước 2: Cleanup sandbox Step 2: Implement Sandbox Cleanup ជំហាន 2: Cleanup sandbox

Chúng ta đang xây What we're building អ្វីដែលយើងកំពុងសង់

A pattern that guarantees sandbox resources are freed after every request.

Vì sao quan trọng Why this matters ហេតុអ្វីសំខាន់

Sandboxes consume CPU, memory, and disk space. Failing to destroy them causes resource leaks, higher costs, and eventually hitting capacity limits.

export default {
  async fetch(request: Request, env: Env): Promise<Response> {
    const sessionId = crypto.randomUUID();
    const sandbox = getSandbox(env.Sandbox, sessionId);

    try {
      const result = await sandbox.exec("npm test");

      return Response.json({
        success: result.success,
        output: result.stdout,
      });
    } finally {
      // Always runs — even if exec() threw or the response already started
      await sandbox.destroy();
    }
  },
};

Bước 3: Quản lý lifecycle sandbox Step 3: Manage Sandbox Lifecycle ជំហាន 3: គ្រប់គ្រង lifecycle sandbox

Chúng ta đang xây What we're building អ្វីដែលយើងកំពុងសង់

Understanding and controlling every stage of the sandbox lifecycle.

Vì sao quan trọng Why this matters ហេតុអ្វីសំខាន់

Proper lifecycle management prevents timeouts on long tasks and resource leaks on abandoned sandboxes.

getSandbox()     →    use (exec, files, runCode, ...)    →    destroy()
    ↓                            ↓
Lazy init on              Reuse the same instance
first use                 across multiple calls

const sandbox = getSandbox(env.Sandbox, "my-sandbox");

// Prevent the sandbox from automatically sleeping (sends heartbeat pings)
await sandbox.setKeepAlive(true);

// Re-enable automatic sleep when you're done with long operations
await sandbox.setKeepAlive(false);

// Terminate the container and free all resources
await sandbox.destroy();

const sandbox = getSandbox(env.Sandbox, "my-sandbox", {
  sleepAfter: "30m", // sleep after 30 minutes of inactivity (default: "10m")
});

if (url.pathname === "/build" && request.method === "POST") {
  const sandbox = getSandbox(env.Sandbox, `build-${crypto.randomUUID()}`);

  try {
    // Prevent automatic sleep during the long build process
    await sandbox.setKeepAlive(true);

    // Clone, install, and build
    await sandbox.gitCheckout("https://github.com/my-org/my-app.git");
    await sandbox.exec("npm ci", {
      cwd: "/workspace/my-app",
      timeout: 120_000,
    });

    const buildResult = await sandbox.exec("npm run build", {
      cwd: "/workspace/my-app",
      timeout: 180_000,
    });

    return Response.json({
      success: buildResult.success,
      output: buildResult.stdout,
    });
  } finally {
    await sandbox.destroy();
  }
}

Bước 4: Cấu hình production Step 4: Production Configuration ជំហាន 4: កំណត់ production

Chúng ta đang xây What we're building អ្វីដែលយើងកំពុងសង់

An optimized wrangler.jsonc tuned for production workloads.

Vì sao quan trọng Why this matters ហេតុអ្វីសំខាន់

Development defaults (lite instances, low max_instances) are not suitable for real traffic. Choose an instance type that matches your workload.

{
  "name": "my-sandbox",
  "main": "src/index.ts",
  "compatibility_date": "2026-01-06", // Update to your deploy date
  "compatibility_flags": ["nodejs_compat"],
  "observability": {
    "enabled": true,
  },
  "containers": [
    {
      "class_name": "Sandbox",
      "image": "./Dockerfile",
      "instance_type": "standard-1",
      "max_instances": 20,
    },
  ],
  "durable_objects": {
    "bindings": [
      {
        "class_name": "Sandbox",
        "name": "Sandbox",
      },
    ],
  },
  "migrations": [
    {
      "new_sqlite_classes": ["Sandbox"],
      "tag": "v1",
    },
  ],
}

Bước 5: Monitor bản deploy Step 5: Monitor Your Deployment ជំហាន 5: Monitor ការដាក់ឱ្យប្រើ

Chúng ta đang xây What we're building អ្វីដែលយើងកំពុងសង់

Visibility into your running production deployment via the Cloudflare dashboard and logs.

Vì sao quan trọng Why this matters ហេតុអ្វីសំខាន់

You can't fix what you can't see — monitoring tells you when things go wrong before your users do.

1. Go to Workers & Pages → your worker
2. Click Logs to see real-time invocation logs
3. Check Metrics for request counts, error rates, and CPU time
4. Open Settings → Observability to configure log retention

# Basic smoke test
curl https://my-sandbox.your-subdomain.workers.dev/run

# Code execution test
curl -X POST https://my-sandbox.your-subdomain.workers.dev/code \
  -H "Content-Type: application/json" \
  -d '{"code": "console.log(\"Production sandbox working!\", new Date().toISOString())"}'

Bạn đã xong lab Sandbox SDK! You completed the Sandbox SDK lab! អ្នកបានបញ្ចប់ lab Sandbox SDK!

1. 01 Set up a sandbox project and executed your first command
2. 02 Mastered exec(), writeFile(), readFile(), and listFiles()
3. 03 Built a multi-language code executor with runCode()
4. 04 Created an AI-powered code executor using Workers AI
5. 05 Built a live app preview with git, processes, and port proxying
6. 06 Applied production security: isolation, secrets, validation, auth
7. 07 Deployed to Cloudflare's global network with proper lifecycle management