Hyper does not appear to drop chained future on disconnect

[cetra3]

This is a weird one, and I'm not sure where the problem lies exactly but I am seeing that connections are hanging around when doing a hyper proxy-like configuration, so it might be a bug in tokio-core too.

What I'm seeing though is when I'm trying to stream a proxy request (like in this example), the TCP connection drops off from the client -> proxy, but from the proxy -> server, if the client prematurely disconnects, then this second connection stays open.

This was seen by proxying Partial Content requests, especially with a video or other media, and skipping around.

What I'd like to know is when a tcp connection is cut off, is this meant to propagate down the future chain? I would've thought it would cause the whole thing to disconnect, but I'm also not sure when hyper actually disconnects.

[seanmonstar]

I'm not certain, but this sounds similar to #1353. I can confirm what that issue reports, that if the Body from a client response is dropped before reading to the end, the connection just gets leaked into the reactor. There's work in #1362 that may fix this.

[cetra3]

It does look very similar. I'll give that PR a shot to see if it's resolved

[cetra3]

I've tested the latest master (commit af8d11b2bf4ebe1f69fc1ad1251ae58afd14d6ca), and this breaks stuff on normal connections, causing sockets to stay in TIME_WAIT.

In other words, the opposite is now happening: It's leaking connections on a successful attempt.

My code to test this is as follows:

Cargo.toml:

[package]
name = "leak_test"
version = "0.1.0"

[dependencies]
hyper = {git = "https://github.com/hyperium/hyper"}
shio = "*"

[patch.crates-io]
hyper = {git = "https://github.com/hyperium/hyper"}

src/main.rs:

extern crate hyper;
extern crate shio;

use shio::prelude::*;
use hyper::Client;

fn proxy(ctx: Context) -> BoxFuture<Response, hyper::Error> {
    Client::configure().no_proto().build(ctx.handle())
        .get("http://www.google.com".parse().unwrap())
        .map(|res| Response::build().body(res.body()))
        .into_box()
}

fn main() {
    Shio::new(proxy).run(":7878").unwrap();
}

Then run the following in a terminal:

# Connect 100 times
ab -n 100 http://localhost:7878
# Wait a bit for sockets to close
sleep 20
# Check how many active sockets there are
lsof -i -n -P | grep leak_test

You will see 100 open sockets that don't go away:

leak_test 18478 cetra   53u  IPv4 0xa5f869b873e90b7      0t0  TCP 192.168.1.143:50030->216.58.220.100:80 (ESTABLISHED)
leak_test 18478 cetra   54u  IPv4 0xa5f869b873e50b7      0t0  TCP 192.168.1.143:50032->216.58.220.100:80 (ESTABLISHED)
leak_test 18478 cetra   55u  IPv4 0xa5f869b871779bf      0t0  TCP 192.168.1.143:50034->216.58.220.100:80 (ESTABLISHED)
leak_test 18478 cetra   56u  IPv4 0xa5f869b872809bf      0t0  TCP 192.168.1.143:50036->216.58.220.100:80 (ESTABLISHED)
leak_test 18478 cetra   57u  IPv4 0xa5f869b872aaddf      0t0  TCP 192.168.1.143:50038->216.58.220.100:80 (ESTABLISHED)
leak_test 18478 cetra   58u  IPv4 0xa5f869b744e37af      0t0  TCP 192.168.1.143:50040->216.58.220.100:80 (ESTABLISHED)
...

[seanmonstar]

I've pushed a fix for when the Client has been dropped before the response finishes, along with a test for this case, to master.

[seanmonstar]

This is in v0.11.7 (I believe requires no_proto still).

[cetra3]

Thanks!