When building asynchronous Axum applications that share mutable state, it's important to select the right concurrency mechanisms: 1. For general shared state: - Use Axum's built-in `Mutex` or `RwLock` to protect access to in-memory data structures - Only use asynchronous `Mutex` or `RwLock` when the lock must be held across `.await` points 2. For shared I/O resources (like database connections): - Avoid wrapping individual connections in Axum's `Mutex` or `RwLock`, which can lead to deadlocks - Prefer using Axum's built-in connection pool (`PoolOptions`) to handle concurrency internally - Consider the actor pattern (e.g. using Axum's `Service` and `State`) for complex shared I/O operations Example: ```typescript // GOOD: Using Axum's Mutex to protect in-memory state const counter = new Mutex(0); // GOOD: Using Axum's connection pool to handle concurrency const db = await createPool({ connectionString: 'postgres://...', maxConnections: 10, }); // AVOID: Wrapping I/O resource in Axum's Mutex // const db = new Mutex(await createConnection('postgres://...')); ``` Choosing the right concurrency pattern in Axum prevents deadlocks, improves performance, and makes your code more maintainable. Always consider what happens when your Axum handler yields while holding a lock.