why four times thread are getting created for a single server?

[101mirsha]

Could you please a bit elaborate?
for _ in 0..4 {
let server = server.clone();
handles.push(thread::spawn(move || {
...

https://github.com/tiny-http/tiny-http/blob/master/examples/hello-world.rs#L7

[dheijl]

Each thread can handle a single request at a time. Until it has finished responding to the request, it can handle no other requests. So in the example code you can handle 4 requests simultaneously, unless of course each serving thread spawns a new thread itself to handle the request, in which case you can serve many more requests simultaneously at the expense of having an uncontrollable number of threads active at the same time.

[brandonros]

To clarify, is that to say under the hood the default behavior if you just have 1 instance of server in a loop handling incoming_requests() is to process 1 HTTP request at a time and block all other requests until that is finished?

[dheijl]

Incoming http requests are still accepted and queued, but not handled until a server thread is available.

[brandonros]

Incoming http requests are still accepted and queued, but not handled until a server thread is available.

Interesting. I read the various conversations around #221

I'm sure it's something I probably did (and I'm sure this probably isn't the right place to discuss this) but I figured that

under the hood the default behavior if you just have 1 instance of server in a loop handling incoming_requests() is to process 1 HTTP request at a time and block all other requests until that is finished?

was the case because I wrote a small HTTP server that is handling a handful of requests at a time and it typically needs to be restarted because /ping takes over 5 seconds/fails.

Then I stumbled across this and applied it:

let server = Arc::new(Server::http("0.0.0.0:9000").expect("failed to start http server"));
    let num_threads = 16;
    let pool = threadpool::ThreadPool::new(num_threads);
    for _ in 0..num_threads {
        let server = server.clone();
        let router = Router::new(route_handlers_map.clone());
        pool.execute(move || {
            for request in server.incoming_requests() {
                match router.handle_request(request) {
                    Ok(_) => (),
                    Err(err) => log::error!("{err:?}")
                }
            }
        });
    }

The issue went away. But I guess it makes sense. Giving "16x" the amount of processing power probably just hid whatever the original issue was.