multi_thread Doesn't work

[zbcms]

Bug Report

Version

axum = "0.4"
axum = "0.5"

Platform

64-bit (Windows)

Description

Step 1： Firefox request -> [ sleep = 2s ]
Step 2：Fast switching Chrome request -> [ sleep > 3s ]

add multi_thread browser or phone Testing Doesn't work
but apache-jmeter Testing is OK

actix-web use sleep Testing is OK

// #[tokio::main]
 #[tokio::main(flavor = "multi_thread", worker_threads = 10)]
 async fn main() {
 
    // Build our application by creating our router.
    let app = axum::Router::new()
    .route("/",get(root));

    // Run our application as a hyper server on http://localhost:3000.
    axum::Server::bind(&"0.0.0.0:3000".parse().unwrap())
        .serve(app.into_make_service())
        .await
        .unwrap();
 }
// basic handler that responds with a static string
async fn root( ) ->String{

    let ten_millis = time::Duration::from_millis(2000);
    println!("\r\nsleep start-----------" );
    thread::sleep(ten_millis);
    println!("sleep stop-----------\r\n" );
   
    let time = SystemTime::now();
 
    format!("Hello, World! {:?}",time)
}

[davidpdrsn]

Don't use std::thread::sleep, since that blocks the async runtime. Use https://docs.rs/tokio/latest/tokio/time/fn.sleep.html instead.

[matthewgapp]

I'm not able to reproduce the blocking behavior here. Using the example code above (specifically using std::thread::sleep()), if I increase the sleep time to 20 secs, I'm able to simultaneously hit the / endpoint with three different curl commands from three different terminals. The requests start and complete in rapid succession as expected. This seems to indicate that each request is given it's own 'thread', as I'd hope so that heavy workloads in a given endpoint don't block all other endpoints. But I was able to reproduce blocking behavior when the Server future is wrapped in tokio::spawn. See here

Would you mind explaining what's going on @davidpdrsn? Why are requests not blocked by default using .await.unwrap() as you suggested and why are they blocked when wrapping the Server in tokio::spawn()?

[davidpdrsn]

I recommend reading https://ryhl.io/blog/async-what-is-blocking/