Tracing Observability

Spans

use tracing::{span, Level, info_span};

fn process_order(order_id: u64) {
    let span = info_span!("process_order", order_id);
    let _guard = span.enter();

    // All logs within this scope include order_id
    tracing::info!("Starting order processing");
    validate_order();
    charge_payment();
    tracing::info!("Order completed");
}

// Async spans
async fn fetch_data() {
    let span = info_span!("fetch_data");
    async {
        // async work
    }.instrument(span).await;
}

#[instrument] Macro

use tracing::instrument;

#[instrument]
fn calculate(x: i32, y: i32) -> i32 {
    x + y
}

// Skip sensitive fields
#[instrument(skip(password))]
fn login(username: &str, password: &str) -> bool {
    true
}

// Custom span name and level
#[instrument(name = "db_query", level = "debug")]
async fn query_database(sql: &str) -> Result<Vec<Row>, Error> {
    // ...
}

// Add fields
#[instrument(fields(request_id = %uuid::Uuid::new_v4()))]
async fn handle_request() {
    // ...
}

// Skip all, add specific
#[instrument(skip_all, fields(user_id))]
fn process_user(user_id: u64, sensitive_data: &str) {
    // ...
}

Subscribers

use tracing_subscriber::{fmt, prelude::*, EnvFilter, Layer};

fn setup_tracing() {
    // JSON output for production
    let json_layer = fmt::layer()
        .json()
        .with_target(true)
        .with_current_span(true);

    // Pretty output for development
    let pretty_layer = fmt::layer()
        .pretty()
        .with_file(true)
        .with_line_number(true);

    // Filter by RUST_LOG env var
    let filter = EnvFilter::try_from_default_env()
        .unwrap_or_else(|_| EnvFilter::new("info"));

    tracing_subscriber::registry()
        .with(filter)
        .with(if cfg!(debug_assertions) {
            pretty_layer.boxed()
        } else {
            json_layer.boxed()
        })
        .init();
}

OpenTelemetry Integration

use opentelemetry::sdk::trace::TracerProvider;
use tracing_opentelemetry::OpenTelemetryLayer;

fn setup_otel() {
    let provider = TracerProvider::builder()
        .with_simple_exporter(opentelemetry_jaeger::new_agent_pipeline())
        .build();

    let tracer = provider.tracer("my-service");

    tracing_subscriber::registry()
        .with(tracing_subscriber::fmt::layer())
        .with(OpenTelemetryLayer::new(tracer))
        .init();
}

Error Handling

use tracing::{error, instrument};

#[instrument(err)]
fn fallible_operation() -> Result<(), Error> {
    // Automatically logs error if Err is returned
    Err(Error::new("something failed"))
}

// With error level
#[instrument(err(level = Level::WARN))]
fn maybe_warn() -> Result<(), Error> {
    // ...
}

// Manual error logging
fn handle_error() {
    if let Err(e) = operation() {
        error!(error = %e, "Operation failed");
    }
}

Best Practices

1. Structured logging: Use fields, not string interpolation
2. Span context: Wrap related operations in spans
3. Skip sensitive: Use skip for passwords, tokens
4. Env filter: Configure via RUST_LOG env var
5. Async: Use .instrument(span) for async code