Task pool for Tokio Runtime
https://crates.io/crates/tokio-task-pool
A typical pattern
loop {
let (socket, _) = listener.accept().await.unwrap();
tokio::spawn(async move {
process(socket).await;
});
}
is actually an anti-pattern which may break your production.
Why? Because this pattern behaves equally to an unbounded channel. If the producer has higher rate than consumers, it floods runtime with tasks and sooner or later causes memory overflow.
-
Use a pool of workers instead
-
Use task spawning but manually limit the number of active tasks with a semaphore
-
Use this crate which does the job out-of-the-box
-
Pool objects with safe spawn methods, which automatically limit number of tasks
-
Tasks can be automatically aborted if run timeout is set, global or per task
Simple spawning is pretty similar to tokio::spawn, but async because the producer must be blocked until there is an empty task slot in the pool:
use std::time::Duration;
use tokio_task_pool::Pool;
#[tokio::main]
async fn main() {
let pool = Pool::bounded(5)
.with_spawn_timeout(Duration::from_secs(5))
.with_run_timeout(Duration::from_secs(10));
pool.spawn(async move {
// do some job
}).await.unwrap();
}
Refer to the crate documentation.
- A "log" feature enables automatic error logging for timed out tasks (via the log crate)