-
Notifications
You must be signed in to change notification settings - Fork 750
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Dynamically add more log layers at run-time #2499
Comments
Hayden helped me on Discord to get it to compile: fn tracing_init() {
use tracing_subscriber::prelude::*;
use tracing_subscriber::{filter, fmt, registry, reload};
let stdout_layer = fmt::Layer::default()
.with_filter(filter::LevelFilter::INFO)
.boxed();
let tracing_layers = vec![stdout_layer];
let (tracing_layers, reload_handle) = reload::Layer::new(tracing_layers);
registry().with(tracing_layers).init();
tracing::info!("before reload");
let reload_result = reload_handle.modify(|layers| {
let json_layer = fmt::Layer::default()
.json()
.with_filter(filter::LevelFilter::INFO)
.boxed();
(*layers).push(json_layer);
});
match reload_result {
Ok(_) => {} // Great!
Err(err) => tracing::warn!("Unable to add new layer: {}", err),
}
tracing::info!("after reload");
} But according to Hayden, it panics at runtime, so this is still an issue. |
+1, I've run into this when adding https://github.com/tokio-rs/tracing-opentelemetry to an existing app. It has an early |
type SimplifiedHandle = reload::Handle<
LevelFilter,
fmt::Layer<Registry, DefaultFields, Format<Full, SystemTime>, Stdout>,
>;
lazy_static! {
static ref RELOAD_HANDLE: Option<SimplifiedHandle> = None;
}
pub fn initialize_logging(level: i32) {
// Create a LevelFilter and a reload::Layer
// let (level_filter, reload_handle) = reload::Layer::new(level_to_filter(level));
// Update the static handle.
// *RELOAD_HANDLE = Some(reload_handle);
let file_fmt_layer = fmt::Layer::default()
.with_level(false)
.with_writer(CallbackWriter)
.with_timer(WithoutTimeFormat)
.with_filter(level_to_filter(level)); // Use the same level filter
let console_fmt_layer = fmt::Layer::default()
.with_file(true)
.with_line_number(true)
.with_writer(std::io::stdout)
.with_timer(CustomTimeFormat)
.with_filter(level_to_filter(level)); // Use the same level filter
// let subscriber = Registry::default().with(level_filter);
let subscriber = Registry::default()
.with(file_fmt_layer)
.with(console_fmt_layer);
tracing::subscriber::set_global_default(subscriber).expect("Failed to set subscriber");
}
pub fn update_logging_level(level: i32) {
let level_filter = level_to_filter(level);
if let Some(reload_handle) = &*RELOAD_HANDLE {
reload_handle
.modify(|filter| *filter = level_filter)
.expect("Failed to update logging level");
}
}
fn level_to_filter(level: i32) -> LevelFilter {
match level {
0 => LevelFilter::ERROR,
1 => LevelFilter::WARN,
2 => LevelFilter::INFO,
3 => LevelFilter::DEBUG,
4 => LevelFilter::TRACE,
_ => LevelFilter::INFO,
}
} how to save a reload_handle, need help |
found correct usage 😸 lazy_static! {
static ref IS_INITIALIZED: AtomicBool = AtomicBool::new(false);
static ref RELOAD_HANDLE: Mutex<Option<reload::Handle<filter::LevelFilter, Registry>>> =
Mutex::new(None);
}
pub fn setup_logging(level: i32) {
if IS_INITIALIZED.load(Ordering::Acquire) {
update_logging_level(level);
return;
}
let result = IS_INITIALIZED.compare_exchange(false, true, Ordering::Acquire, Ordering::Relaxed);
match result {
Ok(_) => {
let (level_filter, reload_handle) = reload::Layer::new(level_to_filter(level));
set_reload_handle(reload_handle);
tracing_subscriber::registry()
.with(level_filter)
.with(
fmt::Layer::default()
.with_level(false)
.with_writer(CStringWriter)
.with_timer(WithoutTimeFormat),
)
.with(
fmt::Layer::default()
.with_file(true)
.with_line_number(true)
.with_timer(CustomTimeFormat),
)
.init();
}
Err(_) => {
// if the logging has already been initialized, just update the logging level
update_logging_level(level);
}
}
}
pub fn update_logging_level(level: i32) {
if let Some(reload_handle) = &*RELOAD_HANDLE.lock().unwrap() {
reload_handle
.modify(|filter| *filter = level_to_filter(level))
.expect("Failed to update logging level");
}
}
fn set_reload_handle(reload_handle: reload::Handle<filter::LevelFilter, Registry>) {
let mut handle = RELOAD_HANDLE.lock().unwrap();
*handle = Some(reload_handle);
} |
Overall, I think I have a solution and figured i'd share in case anyone else is googling how to dynamically add layers. My use case was just initializing the console layer first and then adding a file system layer based off a tauri's app config and CLI options. Here's my main function: #[cfg_attr(mobile, tauri::mobile_entry_point)]
pub fn run() {
// this is where I init console layer and return handle
let log_handle = match logging::setup() {
Ok(handle) => {
info!("Console tracing initialized");
handle
}
Err(e) => {
panic!("Error setting up logging: {:?}", e);
}
};
let mut builder = tauri::Builder::default();
// other stuff unreleated
// fs layer added in setup callback
builder = builder.setup(|app| Ok(setup(app, log_handle)?));
// build invokers and run tauri
});
}
// parent is None to just get rid of large tauri messages
#[tracing::instrument(skip_all, parent = None)]
fn setup(app: &mut App, log_handle: logging::LogHandle) -> Result<()> {
let rt = tauri::async_runtime::handle();
rt.block_on(async {
// other stuff, get config
logging::setup_fs(&config.logs_dir, log_handle)?;
info!("File tracing initialized");
// other stuff
})?;
info!("App setup complete");
Ok(())
} And then here is my logging functions. The LogHandle type will change depending on what's being setup. If you want to dynamically change the filter, you'll need two reload handlers probably. Or loop over the vec and change the EnvFilter. Doing this made my Types annoying so I didn't put too much effort into trying to get it to work. // imports
// type for return value
pub type LogHandle = Handle<
Vec<Box<dyn Layer<Layered<EnvFilter, Registry>> + Send + Sync>>,
Layered<EnvFilter, Registry>,
>;
// initialize console layer
pub fn setup() -> Result<LogHandle> {
let layers = default_layers()?;
// wrap the vec in a reload layer
let (layers, reload_handle) = reload::Layer::new(layers);
let env_filter_ = env_filter();
let subscriber = tracing_subscriber::registry()
.with(env_filter_)
.with(layers);
match tracing::subscriber::set_global_default(subscriber) {
Ok(_) => Ok(reload_handle),
Err(_e) => Err(Error::msg("Tracing subscriber already registered.")),
}
}
// later add dir layer via handle
pub fn setup_fs(log_dir: &PathBuf, handle: LogHandle) -> Result<()> {
handle.modify(|filter| {
(*filter).push(fs_layer(log_dir).unwrap());
})?;
Ok(())
}
// probably doesn't need to be it's own func, just an artifact of previous attempts
// Vec impls Layer so it's important to be a vec
fn default_layers<S>() -> Result<Vec<Box<dyn Layer<S> + Send + Sync + 'static>>>
where
S: Subscriber,
for<'a> S: LookupSpan<'a>,
{
let console_error_ = console_layer()?;
Ok(vec![console_error_])
}
fn env_filter() -> EnvFilter {
// build EnvFilter from RUST_LOG and add custom directives if you want, this is unrelated to adding a new layer
// if you want to dynamically change filter, you'll probably need to get a reload handle to this though
}
fn console_layer<S>() -> Result<Box<dyn Layer<S> + Send + Sync + 'static>>
where
S: Subscriber,
for<'a> S: LookupSpan<'a>,
{
let timer = UtcTime::rfc_3339();
Ok(fmt::layer()
.with_timer(timer)
.with_thread_ids(true)
.with_target(true)
.compact()
.boxed())
}
pub fn fs_layer<S>(log_dir: &PathBuf) -> Result<Box<dyn Layer<S> + Send + Sync + 'static>>
where
S: Subscriber,
for<'a> S: LookupSpan<'a>,
{
// create dir, build appender and timer, etc
Ok(fmt::layer()
.with_writer(file_appender)
.with_timer(timer.clone())
.with_thread_ids(true)
.with_thread_names(true)
.with_target(true)
.with_file(true)
.with_line_number(true)
.with_ansi(false)
.boxed())
} |
Feature Request
Crates
tracing-subscriber
Motivation
There are many places I want to pipe my log events:
stderr
These can all be implemented as
Layer
s.The first thing I want to add is the
stderr
and file layers. The GUI and the network layers I want to add later once the GUI and network connections are up and running. That way any problems with the GUI/network setup is properly logged to stdout and to file.In other words: I want to dynamically add layers as the program runs.
Proposal
I propose some sort of simple API along the lines of
tracing_subscriber::registry().add(layer);
Prior art
My own C++ logging Loguru has a simple method for adding new log sinks (https://emilk.github.io/loguru/index.html#callbacks):
This lets different sub-modules of an application can easily add new log sinks at will.
Alternatives
As I was trying to get this work, several helpful people on Discord pointed me towards
tracing_subscriber::reload
.I've tried to get it to work, but so far I've failed. My code thus far:
There are several issues with this approach:
reload_handle
Perhaps all the above issues could be solved by an
tracing_subscriber
expert plus a little bit of documentation. I don't know - I'm not an expert :)The error message for the above code:
The text was updated successfully, but these errors were encountered: