salvo-logging（齐射测井）

Salvo Logging and Tracing

[dependencies]
salvo = { version = "0.89.3", features = ["logging"] }
tracing = "0.1"
tracing-subscriber = { version = "0.3", features = ["env-filter"] }

Logger middleware

salvo::logging::Logger emits a tracing span per request with remote_addr, version, method, path, and response status + duration. It must be installed on a Service (not a Router) so it wraps the catcher and sees the final status:

use salvo::logging::Logger;
use salvo::prelude::*;

#[tokio::main]
async fn main() {
    tracing_subscriber::fmt().init();

    let router = Router::new().get(hello);
    let service = Service::new(router).hoop(Logger::new());

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

Logger::new().log_status_error(false) disables auto-logging of StatusError bodies (on by default).

Custom request logger

When Logger is not enough, write a handler that measures time around call_next:

use salvo::prelude::*;
use std::time::Instant;
use tracing::info;

#[handler]
async fn request_logger(
    req: &mut Request, depot: &mut Depot, res: &mut Response, ctrl: &mut FlowCtrl,
) {
    let start = Instant::now();
    let method = req.method().clone();
    let path = req.uri().path().to_string();

    ctrl.call_next(req, depot, res).await;

    info!(
        %method,
        %path,
        status = res.status_code.map(|s| s.as_u16()).unwrap_or(0),
        duration_ms = start.elapsed().as_millis() as u64,
        "request",
    );
}

Gotcha: inside a handler res.status_code is a Option<StatusCode> field (not a method). If the handler hasn't set it, it's None until the catcher runs.

Structured logging in handlers

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

#[handler]
#[instrument(skip_all, fields(user_id))]
async fn get_user(req: &mut Request, res: &mut Response) {
    let user_id: u32 = req.param("id").unwrap_or(0);
    tracing::Span::current().record("user_id", user_id);

    debug!("fetching user");
    match fetch_user(user_id).await {
        Ok(user) => { info!("found"); res.render(Json(user)); }
        Err(e) => { warn!(error = %e, "not found"); res.status_code(StatusCode::NOT_FOUND); }
    }
}

Log filtering

Configure levels via RUST_LOG (e.g. RUST_LOG=info,salvo=debug,hyper=warn):

use tracing_subscriber::{EnvFilter, fmt, layer::SubscriberExt, util::SubscriberInitExt};

fn init_logging() {
    let filter = EnvFilter::try_from_default_env()
        .unwrap_or_else(|_| EnvFilter::new("info,salvo=debug,hyper=warn"));

    tracing_subscriber::registry()
        .with(filter)
        .with(fmt::layer())
        .init();
}

JSON logging

tracing_subscriber::registry()
    .with(EnvFilter::try_from_default_env().unwrap_or_else(|_| EnvFilter::new("info")))
    .with(fmt::layer().json().with_current_span(true).with_span_list(true))
    .init();

File logging with rotation

tracing-appender = "0.2"

use tracing_appender::{rolling, non_blocking};

let file_appender = rolling::daily("logs", "app.log");
let (writer, _guard) = non_blocking(file_appender);

tracing_subscriber::registry()
    .with(EnvFilter::new("info"))
    .with(fmt::layer().with_writer(writer).with_ansi(false))
    .init();
// _guard MUST be held for the process lifetime or buffered logs are dropped.

Request ID

use uuid::Uuid;

#[handler]
async fn add_request_id(
    req: &mut Request, depot: &mut Depot, res: &mut Response, ctrl: &mut FlowCtrl,
) {
    let id = req.header::<String>("x-request-id")
        .unwrap_or_else(|| Uuid::new_v4().to_string());
    depot.insert("request_id", id.clone());
    res.headers_mut().insert("x-request-id", id.parse().unwrap());

    let span = tracing::info_span!("request", request_id = %id, method = %req.method(), path = %req.uri().path());
    let _enter = span.enter();
    ctrl.call_next(req, depot, res).await;
}

Salvo also ships salvo::request_id::RequestId middleware (feature request-id) that does this automatically.

Slow request alerting

#[handler]
async fn timing(req: &mut Request, depot: &mut Depot, res: &mut Response, ctrl: &mut FlowCtrl) {
    let start = Instant::now();
    ctrl.call_next(req, depot, res).await;
    let ms = start.elapsed().as_millis();
    if ms > 1000 {
        tracing::warn!(path = %req.uri().path(), duration_ms = ms as u64, "slow request");
    }
}

OpenTelemetry

Salvo provides salvo_otel (feature otel) for trace propagation. For raw tracing-otel:

opentelemetry = "0.22"
opentelemetry-otlp = "0.15"
tracing-opentelemetry = "0.23"

let tracer = opentelemetry_otlp::new_pipeline()
    .tracing()
    .with_exporter(opentelemetry_otlp::new_exporter().tonic().with_endpoint("http://localhost:4317"))
    .install_batch(opentelemetry_sdk::runtime::Tokio)
    .unwrap();

tracing_subscriber::registry()
    .with(tracing_subscriber::fmt::layer())
    .with(tracing_opentelemetry::layer().with_tracer(tracer))
    .init();

// On shutdown:
opentelemetry::global::shutdown_tracer_provider();

Gotchas

• Install Logger on Service, not Router — otherwise it logs before the catcher rewrites the status and body.
• Logger requires a tracing subscriber; without one, nothing is emitted.
• tracing-appender writers must keep their WorkerGuard alive or pending log lines are discarded at shutdown.
• Never log request bodies or headers like Authorization / Cookie unredacted.

Related Skills

• salvo-error-handling: Logger auto-logs StatusError bodies.
• salvo-middleware: Logger is just a hoop.
• salvo-testing: tracing-test crate verifies log output.