init4 bin-base

Common functionality for binaries produced by the init4 project. This crate provides:

• Environment parsing utilities
• Standard tracing setup with otlp support
• Standard server for Prometheus metrics
• Standard environment variables to configure these features

Note: This crate is intended as a base for all binaries in the init4 project. It is not intended for outside consumption.

Installation

[dependencies]
init4-bin-base = "0.18"

Quick Start

use init4_bin_base::init4;

fn main() {
    init4();
    // your code here
}

Build the crate docs with cargo doc --open for more details.

init4 Tracing Best Practices

Carefully Consider Level

Event and span level should correspond to the significance of the event as follows:

By default, the OTLP exporter captures DEBUG and higher. Configure with OTEL_LEVEL env var. The log formatter logs at INFO level. Configure with RUST_LOG env var.

// ❌ Avoid
warn!("Connected to database");

// ✅ Instead
info!("Connected to database");

Import from bin-base

Re-export all necessary crates from init4-bin-base rather than adding them to your Cargo.toml:

// ❌ Avoid
use tracing::info;

// ✅ Instead
use init4_bin_base::deps::tracing::info;

Spans

Spans represent the duration of a unit of work. They should be:

• Time-limited — at most a few seconds
• Work-associated — tied to a specific action
• Informative — have useful data, not over-verbose

Inheritance

Spans inherit the currently-entered span as their parent. Avoid spurious span relationships:

// ❌ Avoid — accidental parent-child relationship
let span = info_span!("outer_function").entered();
let my_closure = || {
    let span = info_span!("accidental_child").entered();
    // do some work
};
do_work(closure);

// ✅ Instead — closure span created before outer span
let my_closure = || {
    let span = info_span!("not_a_child").entered();
    // do some work
};
let span = info_span!("outer_function").entered();
do_work(closure);

Avoid Over-Verbose Spans

When instrumenting methods, skip self and add only needed fields:

// ❌ Avoid — self will be Debug-printed (verbose)
#[instrument]
async fn my_method(&self) { }

// ✅ Instead — skip self, add specific fields
#[instrument(skip(self), fields(self.id = self.id))]
async fn my_method(&self) { }

For multiple arguments, skip all and add back what you need:

// ❌ Avoid
#[instrument]
async fn my_method(&self, arg1: i32, arg2: String) { }

// ✅ Instead
#[instrument(skip_all, fields(arg1))]
async fn my_method(&self, arg1: i32, arg2: String) { }

Instrument Futures, Not JoinHandles

// ❌ Avoid — span won't propagate to the future
tokio::spawn(fut).instrument(span);

// ✅ Instead
tokio::spawn(fut.instrument(span));

Instrument Work, Not Tasks

Avoid adding spans to long-running tasks. Create spans in the internal loop instead:

// ❌ Avoid — span open for entire task lifetime
let span = info_span!("task");
tokio::spawn(async {
    loop {
        // work
        tokio::time::sleep(Duration::from_secs(1)).await;
    }
}.instrument(span));

// ✅ Instead — span per iteration
tokio::spawn(async {
    loop {
        let span = info_span!("loop_iteration").entered();
        // work
        drop(span);
        tokio::time::sleep(Duration::from_secs(1)).await;
    }
});

Root Spans

Root spans are top-level spans in a trace. Ensure they correspond to a SINGLE UNIT OF WORK:

// ❌ Avoid — nested work units under one span
let span = info_span!("task");
for item in my_vec {
    let work_span = info_span!("work_unit").entered();
    // work
}

// ✅ Instead — each work unit is a root span
let work_loop = info_span!("work_loop").entered();
for item in my_vec {
    let span = info_span!(parent: None, "work_unit").entered();
    // work
    drop(span);
}

With #[instrument]:

// ✅ Create root span
#[instrument(parent = None)]
async fn a_unit_of_work() { }

Be Careful with instrument(err)

Using #[instrument(err)] emits errors at EACH span level. Only root spans should have instrument(err):

// ❌ Avoid — error emitted multiple times
#[instrument(err)]
async fn one() -> Result<(), ()> { }

#[instrument(err)]
async fn two() -> Result<(), ()> {
    one().await?;
}

// ✅ Instead — only root span has err
#[instrument]
async fn one() -> Result<(), ()> { }

#[instrument(parent = None, err)]
async fn two() -> Result<(), ()> {
    one().await?;
}

To track error bubbling, record additional info:

#[instrument(err)]
async fn do_thing() -> std::io::Result<()> {
    do_inner().await.inspect_err(|_| {
        tracing::span::Span::current().record("err_source", "do_inner");
    })
}

Managing Events

Events represent state at a single point in time. They should be:

• Informative — useful data, not over-verbose
• Descriptive — clear, concise messages
• Lifecycle-aware — record lifecycle of a unit of work
• Non-repetitive — fire ONCE in a span's lifetime

Avoid String Interpolation

Events are structured data. String interpolation loses type information:

// ❌ Avoid
info!("Value calculated: {}", x);

// ✅ Instead
info!(x, "Value calculated");

Lifecycle Events

Events should capture significant lifecycle steps, not every step:

// ❌ Avoid — using events for start/end
info!("Parsing input");
let parsed = parse_input(input);
info!("Input parsed");

// ✅ Instead — use spans
let span = info_span!("parse_input").entered();
let parsed = parse_input(input);
drop(span);

// ✅ Even better — use #[instrument]
#[instrument(skip(input), fields(input_size = input.len()))]
fn parse_input(input: String) -> Option<ParsedInput> { }

DRY: Don't Repeat Yourself (at INFO and DEBUG)

If firing the same event many times, you're violating span rules or verbosity rules:

// ❌ Avoid — same event many times
for i in my_vec {
    info!(i, "processing");
    do_work(i);
}

// ✅ Instead — trace per item, info for summary
for i in my_vec {
    do_work(i);
    trace!(i, "processed vec item");
}
info!(my_vec.len(), "processed my vec");

License

This project is licensed under the MIT License.