mirror of
https://github.com/sigp/lighthouse.git
synced 2026-05-07 16:55:46 +00:00
Upgrade discovery and restructure task execution (#1693)
* Initial rebase * Remove old code * Correct release tests * Rebase commit * Remove eth2-testnet dep on eth2libp2p * Remove crates lost in rebase * Remove unused dep
This commit is contained in:
@@ -10,6 +10,7 @@ slog = { version = "2.5.2", features = ["max_level_trace"] }
|
||||
sloggers = "1.0.0"
|
||||
types = { "path" = "../../consensus/types" }
|
||||
eth2_config = { "path" = "../../common/eth2_config" }
|
||||
task_executor = { "path" = "../../common/task_executor" }
|
||||
eth2_testnet_config = { path = "../../common/eth2_testnet_config" }
|
||||
logging = { path = "../../common/logging" }
|
||||
slog-term = "2.5.0"
|
||||
@@ -19,6 +20,3 @@ futures = "0.3.5"
|
||||
parking_lot = "0.11.0"
|
||||
slog-json = "2.3.0"
|
||||
exit-future = "0.2.0"
|
||||
lazy_static = "1.4.0"
|
||||
lighthouse_metrics = { path = "../../common/lighthouse_metrics" }
|
||||
discv5 = { version = "0.1.0-alpha.12", features = ["libp2p"] }
|
||||
|
||||
@@ -1,156 +0,0 @@
|
||||
use crate::metrics;
|
||||
use futures::channel::mpsc::Sender;
|
||||
use futures::prelude::*;
|
||||
use slog::{debug, trace};
|
||||
use tokio::runtime::Handle;
|
||||
|
||||
/// A wrapper over a runtime handle which can spawn async and blocking tasks.
|
||||
#[derive(Clone)]
|
||||
pub struct TaskExecutor {
|
||||
/// The handle to the runtime on which tasks are spawned
|
||||
pub handle: Handle,
|
||||
/// The receiver exit future which on receiving shuts down the task
|
||||
pub(crate) exit: exit_future::Exit,
|
||||
/// Sender given to tasks, so that if they encounter a state in which execution cannot
|
||||
/// continue they can request that everything shuts down.
|
||||
///
|
||||
/// The task must provide a reason for shutting down.
|
||||
pub(crate) signal_tx: Sender<&'static str>,
|
||||
|
||||
pub(crate) log: slog::Logger,
|
||||
}
|
||||
|
||||
impl TaskExecutor {
|
||||
/// Create a new task executor.
|
||||
///
|
||||
/// Note: this function is mainly useful in tests. A `TaskExecutor` should be normally obtained from
|
||||
/// a [`RuntimeContext`](struct.RuntimeContext.html)
|
||||
pub fn new(
|
||||
handle: Handle,
|
||||
exit: exit_future::Exit,
|
||||
log: slog::Logger,
|
||||
signal_tx: Sender<&'static str>,
|
||||
) -> Self {
|
||||
Self {
|
||||
handle,
|
||||
exit,
|
||||
signal_tx,
|
||||
log,
|
||||
}
|
||||
}
|
||||
|
||||
/// Spawn a future on the tokio runtime wrapped in an `exit_future::Exit`. The task is canceled
|
||||
/// when the corresponding exit_future `Signal` is fired/dropped.
|
||||
///
|
||||
/// This function generates prometheus metrics on number of tasks and task duration.
|
||||
pub fn spawn(&self, task: impl Future<Output = ()> + Send + 'static, name: &'static str) {
|
||||
let exit = self.exit.clone();
|
||||
let log = self.log.clone();
|
||||
|
||||
if let Some(int_gauge) = metrics::get_int_gauge(&metrics::ASYNC_TASKS_COUNT, &[name]) {
|
||||
// Task is shutdown before it completes if `exit` receives
|
||||
let int_gauge_1 = int_gauge.clone();
|
||||
let future = future::select(Box::pin(task), exit).then(move |either| {
|
||||
match either {
|
||||
future::Either::Left(_) => trace!(log, "Async task completed"; "task" => name),
|
||||
future::Either::Right(_) => {
|
||||
debug!(log, "Async task shutdown, exit received"; "task" => name)
|
||||
}
|
||||
}
|
||||
int_gauge_1.dec();
|
||||
futures::future::ready(())
|
||||
});
|
||||
|
||||
int_gauge.inc();
|
||||
self.handle.spawn(future);
|
||||
}
|
||||
}
|
||||
|
||||
/// Spawn a future on the tokio runtime. This function does not wrap the task in an `exit_future::Exit`
|
||||
/// like [spawn](#method.spawn).
|
||||
/// The caller of this function is responsible for wrapping up the task with an `exit_future::Exit` to
|
||||
/// ensure that the task gets canceled appropriately.
|
||||
/// This function generates prometheus metrics on number of tasks and task duration.
|
||||
///
|
||||
/// This is useful in cases where the future to be spawned needs to do additional cleanup work when
|
||||
/// the task is completed/canceled (e.g. writing local variables to disk) or the task is created from
|
||||
/// some framework which does its own cleanup (e.g. a hyper server).
|
||||
pub fn spawn_without_exit(
|
||||
&self,
|
||||
task: impl Future<Output = ()> + Send + 'static,
|
||||
name: &'static str,
|
||||
) {
|
||||
if let Some(int_gauge) = metrics::get_int_gauge(&metrics::ASYNC_TASKS_COUNT, &[name]) {
|
||||
let int_gauge_1 = int_gauge.clone();
|
||||
let future = task.then(move |_| {
|
||||
int_gauge_1.dec();
|
||||
futures::future::ready(())
|
||||
});
|
||||
|
||||
int_gauge.inc();
|
||||
self.handle.spawn(future);
|
||||
}
|
||||
}
|
||||
|
||||
/// Spawn a blocking task on a dedicated tokio thread pool wrapped in an exit future.
|
||||
/// This function generates prometheus metrics on number of tasks and task duration.
|
||||
pub fn spawn_blocking<F>(&self, task: F, name: &'static str)
|
||||
where
|
||||
F: FnOnce() + Send + 'static,
|
||||
{
|
||||
let exit = self.exit.clone();
|
||||
let log = self.log.clone();
|
||||
|
||||
if let Some(metric) = metrics::get_histogram(&metrics::BLOCKING_TASKS_HISTOGRAM, &[name]) {
|
||||
if let Some(int_gauge) = metrics::get_int_gauge(&metrics::BLOCKING_TASKS_COUNT, &[name])
|
||||
{
|
||||
let int_gauge_1 = int_gauge.clone();
|
||||
let timer = metric.start_timer();
|
||||
let join_handle = self.handle.spawn_blocking(task);
|
||||
|
||||
let future = future::select(join_handle, exit).then(move |either| {
|
||||
match either {
|
||||
future::Either::Left(_) => {
|
||||
trace!(log, "Blocking task completed"; "task" => name)
|
||||
}
|
||||
future::Either::Right(_) => {
|
||||
debug!(log, "Blocking task shutdown, exit received"; "task" => name)
|
||||
}
|
||||
}
|
||||
timer.observe_duration();
|
||||
int_gauge_1.dec();
|
||||
futures::future::ready(())
|
||||
});
|
||||
|
||||
int_gauge.inc();
|
||||
self.handle.spawn(future);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// Returns the underlying runtime handle.
|
||||
pub fn runtime_handle(&self) -> Handle {
|
||||
self.handle.clone()
|
||||
}
|
||||
|
||||
/// Returns a copy of the `exit_future::Exit`.
|
||||
pub fn exit(&self) -> exit_future::Exit {
|
||||
self.exit.clone()
|
||||
}
|
||||
|
||||
/// Get a channel to request shutting down.
|
||||
pub fn shutdown_sender(&self) -> Sender<&'static str> {
|
||||
self.signal_tx.clone()
|
||||
}
|
||||
|
||||
/// Returns a reference to the logger.
|
||||
pub fn log(&self) -> &slog::Logger {
|
||||
&self.log
|
||||
}
|
||||
}
|
||||
|
||||
impl discv5::Executor for TaskExecutor {
|
||||
fn spawn(&self, future: std::pin::Pin<Box<dyn Future<Output = ()> + Send>>) {
|
||||
self.spawn(future, "discv5")
|
||||
}
|
||||
}
|
||||
@@ -15,7 +15,6 @@ use futures::channel::{
|
||||
};
|
||||
use futures::{future, StreamExt};
|
||||
|
||||
pub use executor::TaskExecutor;
|
||||
use slog::{info, o, Drain, Level, Logger};
|
||||
use sloggers::{null::NullLoggerBuilder, Build};
|
||||
use std::cell::RefCell;
|
||||
@@ -23,10 +22,9 @@ use std::ffi::OsStr;
|
||||
use std::fs::{rename as FsRename, OpenOptions};
|
||||
use std::path::PathBuf;
|
||||
use std::time::{SystemTime, UNIX_EPOCH};
|
||||
use task_executor::TaskExecutor;
|
||||
use tokio::runtime::{Builder as RuntimeBuilder, Runtime};
|
||||
use types::{EthSpec, InteropEthSpec, MainnetEthSpec, MinimalEthSpec};
|
||||
mod executor;
|
||||
mod metrics;
|
||||
|
||||
pub const ETH2_CONFIG_FILENAME: &str = "eth2-spec.toml";
|
||||
const LOG_CHANNEL_SIZE: usize = 2048;
|
||||
@@ -311,12 +309,7 @@ impl<E: EthSpec> RuntimeContext<E> {
|
||||
/// The generated service will have the `service_name` in all it's logs.
|
||||
pub fn service_context(&self, service_name: String) -> Self {
|
||||
Self {
|
||||
executor: TaskExecutor {
|
||||
handle: self.executor.handle.clone(),
|
||||
signal_tx: self.executor.signal_tx.clone(),
|
||||
exit: self.executor.exit.clone(),
|
||||
log: self.executor.log.new(o!("service" => service_name)),
|
||||
},
|
||||
executor: self.executor.clone_with_name(service_name),
|
||||
eth_spec_instance: self.eth_spec_instance.clone(),
|
||||
eth2_config: self.eth2_config.clone(),
|
||||
}
|
||||
@@ -361,12 +354,12 @@ impl<E: EthSpec> Environment<E> {
|
||||
/// Returns a `Context` where no "service" has been added to the logger output.
|
||||
pub fn core_context(&mut self) -> RuntimeContext<E> {
|
||||
RuntimeContext {
|
||||
executor: TaskExecutor {
|
||||
exit: self.exit.clone(),
|
||||
signal_tx: self.signal_tx.clone(),
|
||||
handle: self.runtime().handle().clone(),
|
||||
log: self.log.clone(),
|
||||
},
|
||||
executor: TaskExecutor::new(
|
||||
self.runtime().handle().clone(),
|
||||
self.exit.clone(),
|
||||
self.log.clone(),
|
||||
self.signal_tx.clone(),
|
||||
),
|
||||
eth_spec_instance: self.eth_spec_instance.clone(),
|
||||
eth2_config: self.eth2_config.clone(),
|
||||
}
|
||||
@@ -375,12 +368,12 @@ impl<E: EthSpec> Environment<E> {
|
||||
/// Returns a `Context` where the `service_name` is added to the logger output.
|
||||
pub fn service_context(&mut self, service_name: String) -> RuntimeContext<E> {
|
||||
RuntimeContext {
|
||||
executor: TaskExecutor {
|
||||
exit: self.exit.clone(),
|
||||
signal_tx: self.signal_tx.clone(),
|
||||
handle: self.runtime().handle().clone(),
|
||||
log: self.log.new(o!("service" => service_name)),
|
||||
},
|
||||
executor: TaskExecutor::new(
|
||||
self.runtime().handle().clone(),
|
||||
self.exit.clone(),
|
||||
self.log.new(o!("service" => service_name)),
|
||||
self.signal_tx.clone(),
|
||||
),
|
||||
eth_spec_instance: self.eth_spec_instance.clone(),
|
||||
eth2_config: self.eth2_config.clone(),
|
||||
}
|
||||
|
||||
@@ -1,21 +0,0 @@
|
||||
/// Handles async task metrics
|
||||
use lazy_static::lazy_static;
|
||||
pub use lighthouse_metrics::*;
|
||||
|
||||
lazy_static! {
|
||||
pub static ref ASYNC_TASKS_COUNT: Result<IntGaugeVec> = try_create_int_gauge_vec(
|
||||
"async_tasks_count",
|
||||
"Total number of async tasks spawned using spawn",
|
||||
&["async_task_count"]
|
||||
);
|
||||
pub static ref BLOCKING_TASKS_COUNT: Result<IntGaugeVec> = try_create_int_gauge_vec(
|
||||
"blocking_tasks_count",
|
||||
"Total number of async tasks spawned using spawn_blocking",
|
||||
&["blocking_task_count"]
|
||||
);
|
||||
pub static ref BLOCKING_TASKS_HISTOGRAM: Result<HistogramVec> = try_create_histogram_vec(
|
||||
"blocking_tasks_histogram",
|
||||
"Time taken by blocking tasks",
|
||||
&["blocking_task_hist"]
|
||||
);
|
||||
}
|
||||
Reference in New Issue
Block a user