Crates.io | tracing-forest |
lib.rs | tracing-forest |
version | 0.1.6 |
source | src |
created_at | 2022-01-01 18:48:12.237049 |
updated_at | 2023-06-19 17:27:52.502599 |
description | Preserving contextual coherence among trace data from concurrent tasks |
homepage | |
repository | https://github.com/QnnOkabayashi/tracing-forest |
max_upload_size | |
id | 506267 |
size | 114,984 |
Preserve contextual coherence among trace data from concurrent tasks.
tracing
is a framework for instrumenting programs to collect structured
and async-aware diagnostics via the Subscriber
trait. The
tracing-subscriber
crate provides tools for composing Subscriber
s
from smaller units. This crate extends tracing-subscriber
by providing
ForestLayer
, a Layer
that preserves contextual coherence of trace
data from concurrent tasks when logging.
This crate is intended for programs running many nontrivial and disjoint
tasks concurrently, like server backends. Unlike other Subscriber
s
which simply keep track of the context of an event, tracing-forest
preserves
the contextual coherence when writing logs even in parallel contexts, allowing
readers to easily trace a sequence of events from the same task.
tracing-forest
is intended for authoring applications.
The easiest way to get started is to enable all features. Do this by
adding the following to your Cargo.toml
file:
tracing-forest = { version = "0.1.6", features = ["full"] }
Then, add tracing_forest::init
to your main function:
fn main() {
tracing_forest::init();
// ...
}
Similar to this crate, the tracing-tree
crate collects and writes trace
data as a tree. Unlike this crate, it doesn't maintain contextual coherence
in parallel contexts.
Observe the below program, which simulates serving multiple clients concurrently.
use tracing::info;
use tracing_subscriber::{layer::SubscriberExt, util::SubscriberInitExt, Registry};
use tracing_tree::HierarchicalLayer;
#[tracing::instrument]
async fn conn(id: u32) {
for i in 0..3 {
some_expensive_operation().await;
info!(id, "step {}", i);
}
}
#[tokio::main(flavor = "multi_thread")]
async fn main() {
// Use a `tracing-tree` subscriber
Registry::default()
.with(HierarchicalLayer::default())
.init();
let connections: Vec<_> = (0..3)
.map(|id| tokio::spawn(conn(id)))
.collect();
for conn in connections {
conn.await.unwrap();
}
}
tracing-tree
isn't intended for concurrent use, and this is demonstrated
by the output of the program:
conn id=2
conn id=0
conn id=1
23ms INFO step 0, id=2
84ms INFO step 0, id=1
94ms INFO step 1, id=2
118ms INFO step 0, id=0
130ms INFO step 1, id=1
193ms INFO step 2, id=2
217ms INFO step 1, id=0
301ms INFO step 2, id=1
326ms INFO step 2, id=0
We can instead use tracing-forest
as a drop-in replacement for tracing-tree
.
use tracing::info;
use tracing_subscriber::{layer::SubscriberExt, util::SubscriberInitExt, Registry};
use tracing_forest::ForestLayer;
#[tracing::instrument]
async fn conn(id: u32) {
// -- snip --
}
#[tokio::main(flavor = "multi_thread")]
async fn main() {
// Use a `tracing-forest` subscriber
Registry::default()
.with(ForestLayer::default())
.init();
// -- snip --
}
Now we can easily trace what happened:
INFO conn [ 150µs | 100.00% ] id: 1
INFO ┝━ i [info]: step 0 | id: 1
INFO ┝━ i [info]: step 1 | id: 1
INFO ┕━ i [info]: step 2 | id: 1
INFO conn [ 343µs | 100.00% ] id: 0
INFO ┝━ i [info]: step 0 | id: 0
INFO ┝━ i [info]: step 1 | id: 0
INFO ┕━ i [info]: step 2 | id: 0
INFO conn [ 233µs | 100.00% ] id: 2
INFO ┝━ i [info]: step 0 | id: 2
INFO ┝━ i [info]: step 1 | id: 2
INFO ┕━ i [info]: step 2 | id: 2
tracing-forest
is open-source software, distributed under the MIT license.