Rust Async Patterns Skill

This skill provides guidance on writing correct, efficient, and safe asynchronous Rust code using tokio and async/await.

When to use this skill

•Implementing async functions or services
•Converting blocking code to async
•Managing concurrent operations
•Working with tokio runtime
•Handling I/O-bound or CPU-bound work in async context
•Implementing timeouts and cancellation

Core Principles

1. Never Block the Async Runtime

Rule: CPU-intensive or blocking operations must run in spawn_blocking, never directly in async functions.

Why: Blocking the async executor starves other tasks and degrades performance.

Bad:

rust

// BAD: Blocks the async runtime
pub async fn process_image(data: Vec<u8>) -> Result<Vec<u8>, Error> {
    // Heavy CPU work on async thread!
    let img = image::load_from_memory(&data)?;
    let resized = img.resize(100, 100, FilterType::Lanczos3);
    Ok(resized.to_bytes())
}

Good:

rust

// GOOD: CPU work in blocking thread pool
pub async fn process_image(data: Vec<u8>) -> Result<Vec<u8>, Error> {
    tokio::task::spawn_blocking(move || {
        let img = image::load_from_memory(&data)?;
        let resized = img.resize(100, 100, FilterType::Lanczos3);
        Ok(resized.to_bytes())
    })
    .await
    .map_err(|e| Error::JoinError(e))?
}

2. Always Set Timeouts on I/O Operations

Rule: All network I/O, file I/O, and external calls must have explicit timeouts.

Why: Prevents indefinite hangs and resource exhaustion.

rust

use tokio::time::{timeout, Duration};

pub async fn fetch_data(url: &str) -> Result<String, Error> {
    let request = reqwest::get(url);
    
    match timeout(Duration::from_secs(10), request).await {
        Ok(Ok(response)) => Ok(response.text().await?),
        Ok(Err(e)) => Err(Error::RequestFailed(e)),
        Err(_) => Err(Error::Timeout),
    }
}

3. Bound Concurrent Operations

Rule: Never spawn unbounded concurrent tasks. Use semaphores or buffered channels.

Why: Prevents resource exhaustion and OOM conditions.

rust

use tokio::sync::Semaphore;
use std::sync::Arc;

const MAX_CONCURRENT: usize = 10;

pub async fn process_batch(items: Vec<Item>) -> Result<Vec<Output>, Error> {
    let semaphore = Arc::new(Semaphore::new(MAX_CONCURRENT));
    let mut tasks = Vec::new();
    
    for item in items {
        let permit = semaphore.clone().acquire_owned().await?;
        
        let task = tokio::spawn(async move {
            let result = process_item(item).await;
            drop(permit); // Release permit
            result
        });
        
        tasks.push(task);
    }
    
    // Collect results
    let mut results = Vec::new();
    for task in tasks {
        results.push(task.await??);
    }
    
    Ok(results)
}

4. Use Structured Concurrency

Rule: Prefer tokio::select!, join!, and try_join! over manual task spawning when possible.

Why: Ensures proper cancellation and resource cleanup.

rust

use tokio::try_join;

pub async fn load_chart_with_audio(
    chart_path: &Path,
    audio_path: &Path,
) -> Result<(RoxChart, AudioData), Error> {
    // Both load concurrently, both must succeed
    let (chart, audio) = try_join!(
        load_chart(chart_path),
        load_audio(audio_path)
    )?;
    
    Ok((chart, audio))
}

Common Async Patterns

Pattern 1: Async Function with Timeout

rust

use tokio::time::{timeout, Duration};

pub async fn decode_with_timeout(
    data: &[u8],
    max_duration: Duration,
) -> Result<RoxChart, CodecError> {
    timeout(max_duration, async {
        // Actual decode work
        decode_chart(data)
    })
    .await
    .map_err(|_| CodecError::Timeout)?
}

Pattern 2: CPU-Bound Work in Async Context

rust

pub async fn compute_hash(data: Vec<u8>) -> Result<Hash, Error> {
    tokio::task::spawn_blocking(move || {
        // CPU-intensive hashing
        blake3::hash(&data)
    })
    .await
    .map_err(|e| Error::TaskFailed(e.to_string()))
}

Pattern 3: Concurrent Stream Processing

rust

use futures::stream::{self, StreamExt};

pub async fn process_charts_concurrent(
    paths: Vec<PathBuf>,
    concurrency: usize,
) -> Vec<Result<RoxChart, CodecError>> {
    stream::iter(paths)
        .map(|path| async move {
            decode_chart(&path).await
        })
        .buffer_unordered(concurrency)
        .collect()
        .await
}

Pattern 4: Graceful Shutdown with Cancellation

rust

use tokio::sync::watch;
use tokio::select;

pub async fn worker_with_shutdown(
    mut shutdown_rx: watch::Receiver<bool>,
) -> Result<(), Error> {
    loop {
        select! {
            _ = shutdown_rx.changed() => {
                info!("Shutdown signal received");
                break;
            }
            result = do_work() => {
                result?;
            }
        }
    }
    
    Ok(())
}

Pattern 5: Retry with Exponential Backoff

rust

use tokio::time::{sleep, Duration};

pub async fn retry_with_backoff<F, Fut, T, E>(
    mut operation: F,
    max_retries: u32,
) -> Result<T, E>
where
    F: FnMut() -> Fut,
    Fut: std::future::Future<Output = Result<T, E>>,
{
    let mut attempt = 0;
    
    loop {
        match operation().await {
            Ok(result) => return Ok(result),
            Err(e) if attempt >= max_retries => return Err(e),
            Err(_) => {
                attempt += 1;
                let delay = Duration::from_millis(100 * 2_u64.pow(attempt));
                sleep(delay).await;
            }
        }
    }
}

Decision Tree

When to use async vs sync?

•
Is this I/O-bound work? (network, file, database)
- •Yes → Use async
- •No → Continue to next question
•
Is this CPU-bound work?
- •Yes → Use spawn_blocking if called from async context
- •No → Continue to next question
•
Do I need concurrency?
- •Yes → Use async
- •No → Sync is simpler

When to use `spawn` vs `spawn_blocking`?

•
Is the work CPU-intensive? (image processing, hashing, compression)
- •Yes → Use spawn_blocking
- •No → Continue
•
Does it call blocking APIs? (std::fs, blocking network)
- •Yes → Use spawn_blocking
- •No → Use spawn for async work

How to handle multiple concurrent operations?

•
Must all succeed?
- •Yes → Use try_join!
- •No → Continue
•
Do I need first result?
- •Yes → Use select!
- •No → Use join! or manual task collection

Resource Budgets

Always define explicit limits:

rust

// Configuration for async operations
pub struct AsyncConfig {
    /// Maximum concurrent blocking tasks
    pub max_blocking_tasks: usize,
    
    /// Maximum concurrent I/O operations
    pub max_io_concurrent: usize,
    
    /// Default timeout for I/O operations
    pub io_timeout: Duration,
    
    /// Maximum retry attempts
    pub max_retries: u32,
}

impl Default for AsyncConfig {
    fn default() -> Self {
        Self {
            max_blocking_tasks: num_cpus::get(),
            max_io_concurrent: 100,
            io_timeout: Duration::from_secs(30),
            max_retries: 3,
        }
    }
}

Common Mistakes to Avoid

❌ Blocking in async functions

rust

// BAD
pub async fn process(data: Vec<u8>) -> Result<Vec<u8>, Error> {
    std::thread::sleep(Duration::from_secs(1)); // Blocks executor!
    Ok(data)
}

rust

// GOOD
pub async fn process(data: Vec<u8>) -> Result<Vec<u8>, Error> {
    tokio::time::sleep(Duration::from_secs(1)).await;
    Ok(data)
}

❌ No timeout on I/O

rust

// BAD
pub async fn fetch(url: &str) -> Result<String, Error> {
    Ok(reqwest::get(url).await?.text().await?)
}

rust

// GOOD
pub async fn fetch(url: &str) -> Result<String, Error> {
    let response = timeout(
        Duration::from_secs(10),
        reqwest::get(url)
    ).await??;
    
    Ok(response.text().await?)
}

❌ Unbounded task spawning

rust

// BAD
pub async fn process_all(items: Vec<Item>) -> Vec<Output> {
    let mut tasks = Vec::new();
    for item in items {
        tasks.push(tokio::spawn(process(item))); // Could spawn millions!
    }
    // ...
}

rust

// GOOD
pub async fn process_all(items: Vec<Item>) -> Vec<Output> {
    stream::iter(items)
        .map(|item| process(item))
        .buffer_unordered(10) // Bounded concurrency
        .collect()
        .await
}

❌ Ignoring cancellation

rust

// BAD
pub async fn long_running_task() {
    loop {
        do_work().await;
        // No way to cancel!
    }
}

rust

// GOOD
pub async fn long_running_task(mut shutdown: watch::Receiver<bool>) {
    loop {
        select! {
            _ = shutdown.changed() => break,
            _ = do_work() => {}
        }
    }
}

Testing Async Code

rust

#[cfg(test)]
mod tests {
    use super::*;
    use tokio::time::{pause, advance, Duration};

    #[tokio::test]
    async fn test_timeout_works() {
        let result = timeout(
            Duration::from_millis(100),
            async {
                tokio::time::sleep(Duration::from_secs(10)).await;
                "done"
            }
        ).await;
        
        assert!(result.is_err());
    }

    #[tokio::test]
    async fn test_concurrent_processing() {
        let items = vec![1, 2, 3, 4, 5];
        let results = process_batch(items).await.unwrap();
        assert_eq!(results.len(), 5);
    }
}

Observability

Instrument async functions for tracing:

rust

use tracing::{instrument, info, warn};

#[instrument(skip(data))]
pub async fn process_chart(
    chart_id: &str,
    data: Vec<u8>,
) -> Result<RoxChart, Error> {
    info!("Starting chart processing");
    
    let chart = tokio::task::spawn_blocking(move || {
        decode_chart(&data)
    })
    .await
    .map_err(|e| {
        warn!(error = %e, "Blocking task failed");
        Error::DecodeFailed
    })??;
    
    info!(note_count = chart.notes.len(), "Chart processed");
    Ok(chart)
}

Checklist

When implementing async code:

• CPU-bound work uses spawn_blocking
• All I/O operations have timeouts
• Concurrent operations are bounded
• Proper error handling with ? and Result
• Cancellation is supported where needed
• Resource limits are explicit and documented
• Async functions are instrumented for observability
• Tests cover timeout and cancellation scenarios

References

•User rule: rule-memory-and-ressources.md (Resource & Memory Budget)
•User rule: rule-observability.md (Observability)
•Tokio documentation
•Async Rust Book

Rust Async Patterns Skill

When to use this skill

Core Principles

1. Never Block the Async Runtime

2. Always Set Timeouts on I/O Operations

3. Bound Concurrent Operations

4. Use Structured Concurrency

Common Async Patterns

Pattern 1: Async Function with Timeout

Pattern 2: CPU-Bound Work in Async Context

Pattern 3: Concurrent Stream Processing

Pattern 4: Graceful Shutdown with Cancellation

Pattern 5: Retry with Exponential Backoff

Decision Tree

When to use async vs sync?

When to use spawn vs spawn_blocking?

How to handle multiple concurrent operations?

Resource Budgets

Common Mistakes to Avoid

❌ Blocking in async functions

❌ No timeout on I/O

❌ Unbounded task spawning

❌ Ignoring cancellation

Testing Async Code

Observability

Checklist

References

When to use `spawn` vs `spawn_blocking`?