salvo-timeout（齐射超时）

Salvo Request Timeout

Timeout lives in salvo_extra behind a feature flag — it is NOT on by default.

[dependencies]
salvo = { version = "0.89.3", features = ["timeout"] }

Timeout is re-exported via salvo::prelude::Timeout.

Basic usage

use std::time::Duration;
use salvo::prelude::*;

#[handler]
async fn slow() -> &'static str {
    tokio::time::sleep(Duration::from_secs(10)).await;
    "done"
}

#[tokio::main]
async fn main() {
    let router = Router::new()
        .hoop(Timeout::new(Duration::from_secs(5)))
        .push(Router::with_path("slow").get(slow));

    let acceptor = TcpListener::new("0.0.0.0:8080").bind().await;
    Server::new(acceptor).serve(router).await;
}

GOTCHA: on timeout Salvo returns 503 Service Unavailable, not 408 Request Timeout. This is intentional — some browsers retry 408 automatically. The response also includes Connection: close.

Custom error

Override the error via .error(|| StatusError):

use salvo::http::StatusError;

let timeout = Timeout::new(Duration::from_secs(5))
    .error(|| StatusError::gateway_timeout().brief("Upstream too slow."));

Per-route timeouts

let router = Router::new()
    .push(Router::with_path("quick")
        .hoop(Timeout::new(Duration::from_secs(2)))
        .get(quick_handler))
    .push(Router::with_path("upload")
        .hoop(Timeout::new(Duration::from_secs(120)))
        .post(upload_handler))
    .push(Router::with_path("report")
        .hoop(Timeout::new(Duration::from_secs(300)))
        .post(report_handler));

Global default + per-route override

Inner hoop runs after the outer one; a longer timeout on an inner route effectively overrides the shorter global timeout because the first handler to complete (or expire) wins the tokio::select! race.

let router = Router::new()
    .hoop(Timeout::new(Duration::from_secs(30)))  // default
    .push(Router::with_path("health").get(health))
    .push(
        Router::with_path("reports/generate")
            .hoop(Timeout::new(Duration::from_secs(300)))
            .post(generate_report),
    );

Handling timeouts in a Catcher

A Catcher can customize the 503 body:

use salvo::catcher::Catcher;

#[handler]
async fn on_timeout(res: &mut Response, ctrl: &mut FlowCtrl) {
    if res.status_code == Some(StatusCode::SERVICE_UNAVAILABLE) {
        res.render(Json(serde_json::json!({
            "error": "timeout",
            "message": "The request took too long to process",
        })));
        ctrl.skip_rest();
    }
}

let service = Service::new(router).catcher(Catcher::default().hoop(on_timeout));

Combining with rate limiting

let router = Router::new()
    .hoop(rate_limiter)                              // 429 on abuse
    .hoop(Timeout::new(Duration::from_secs(30)))     // 503 on stall
    .push(Router::with_path("api/{**rest}").get(api_handler));

WebSocket / SSE note

Do NOT wrap long-lived endpoints (WebSocket upgrade, SSE) with Timeout — the middleware aborts any handler that runs longer than the duration, which will terminate these connections.

Related Skills

• salvo-concurrency-limiter: cap in-flight requests
• salvo-rate-limiter: rate limiting
• salvo-graceful-shutdown: handle in-flight requests on shutdown