request-response behaviour on relayed connection

[fgimenez]

I have a setup composed of 2 peers running on browsers and a relay server, the browsers configuration is similar to the file sharing example. I'm trying to request a value from one of the browser peers to the other browser peer through a relayed connection, the request is received correctly but the response is not received, I'm getting an eof error that seems to be originated when trying to read the response length here https://github.com/libp2p/rust-libp2p/blob/master/core/src/upgrade/transfer.rs#L69

This is the sequence of actions (all based on the file sharing example above):

• Both browser peers connect to the relay server
• browser peer 1 starts providing a key in kademlia
• browser peer 2 discovers browser peer 1 relayed multiaddr by getting the key providers and dials it
• browser peer 2 requests the value associated with the key to browser peer 1
• on browser peer 1 logs all looks ok, request received and response sent
• on browser peer 2 and relay server logs an eof is shown when receiving the response. This is how the relay server logs look like:

[2022-10-17T16:23:06Z DEBUG wasm_net::network::relay_server] received event: Behaviour(Relay(CircuitReqAccepted { src_peer_id: PeerId("12D3KooWMpeKAbMK4BTPsQY3rG7XwtdstseHGcq7kffY8LToYYKK"), dst_peer_id: PeerId("12D3KooWP285Hw3CSTdr9oU6Ezz4hDoi6XS5vfDjjNeTJ1uFMGvp") }))
[2022-10-17T16:23:06Z DEBUG wasm_net::network::relay_server] Behaviour(Relay(CircuitReqAccepted { src_peer_id: PeerId("12D3KooWMpeKAbMK4BTPsQY3rG7XwtdstseHGcq7kffY8LToYYKK"), dst_peer_id: PeerId("12D3KooWP285Hw3CSTdr9oU6Ezz4hDoi6XS5vfDjjNeTJ1uFMGvp") }))
[2022-10-17T16:23:06Z DEBUG netlink_proto::connection] reading incoming messages
[2022-10-17T16:23:06Z DEBUG netlink_proto::codecs] NetlinkCodec: decoding next message
[2022-10-17T16:23:06Z DEBUG netlink_proto::connection] forwarding unsolicited messages to the connection handle
[2022-10-17T16:23:06Z DEBUG netlink_proto::connection] forwaring responses to previous requests to the connection handle
[2022-10-17T16:23:06Z DEBUG netlink_proto::connection] handling requests
[2022-10-17T16:23:06Z DEBUG netlink_proto::connection] sending messages
[2022-10-17T16:23:07Z DEBUG soketto::connection] 2a799733: cannot receive, connection is closed
[2022-10-17T16:23:07Z ERROR yamux::connection] a46441cc: socket error: decode error: i/o error: unexpected end of file
[2022-10-17T16:23:07Z DEBUG libp2p_swarm] Connection closed with error IO(Custom { kind: Other, error: A(YamuxError(Decode(Io(Kind(UnexpectedEof))))) }): Connected { endpoint: Listener { local_addr: "/ip4/127.0.0.1/tcp/38615/ws", send_back_addr: "/ip4/127.0.0.1/tcp/50392/ws" }, peer_id: PeerId("12D3KooWP285Hw3CSTdr9oU6Ezz4hDoi6XS5vfDjjNeTJ1uFMGvp") }; Total (peer): 0.

and this is how the logs look like on browser peer 2:

%cDEBUG%c /home/fgimenez/.cargo/registry/src/github.com-1ecc6299db9ec823/yamux-0.10.2/src/connection/stream.rs:317%c 1d9ae702/9: eof
    color: white; padding: 0 3px; background: blue;
    font-weight: bold; color: inherit
    background: inherit; color: inherit
    %cDEBUG%c /home/fgimenez/.cargo/registry/src/github.com-1ecc6299db9ec823/yamux-0.10.2/src/connection.rs:698%c 735dd40a: socket eof
    color: white; padding: 0 3px; background: blue;
    font-weight: bold; color: inherit
    background: inherit; color: inherit
    %cDEBUG%c /home/fgimenez/workspace/nillion/libp2p-poc/libp2p/swarm/src/lib.rs:767%c Connection closed with error IO(Custom { kind: Other, error: A(YamuxError(Closed)) }): Connected { endpoint: Dialer { address: "/ip4/127.0.0.1/tcp/38615/p2p/12D3KooWDpJ7As7BWAwRMfu1VU2WCqNjvq387JEYKDBj4kx6nXTN/ws/p2p-circuit/p2p/12D3KooWP285Hw3CSTdr9oU6Ezz4hDoi6XS5vfDjjNeTJ1uFMGvp", role_override: Dialer }, peer_id: PeerId("12D3KooWP285Hw3CSTdr9oU6Ezz4hDoi6XS5vfDjjNeTJ1uFMGvp") }; Total (peer): 0.
    color: white; padding: 0 3px; background: blue;

Complete logs of the 3 nodes are attached.
browser1.txt
browser2.txt
relay-server.txt

So my question, is it possible to use the request-response behaviour in this scenario? If not, is there any other existing behaviour that could be used, or would it be possible to develop one?

Thanks!

[fgimenez]

Sorry for the noise, there was a problem with the test, the response sender was being closed too soon and the stream was suddenly closed, all is working ok 👍

[gelfand]

Hello, how do you fixed it in your code? Because I'm facing the same issue right now using libp2p's request-response protocol, and have no idea how to fix it

[fgimenez]

how do you fixed it in your code?

In my case it was just an issue in the wasm-pack based test that runs the browser that receives the request and sends the response, just adding a small delay at the end was enough to fix it:

use futures_timer::Delay
use instant::Duration;
.....

network_client
    .respond_value(your_response, channel)
    .await;

Delay::new(Duration::from_secs(1)).await;

[youngjoon-lee]

Hi @fgimenez

I’m trying the same thing as you had done. File-sharing in browser using relay. But I’m struggling to make the relayed connection not closed immediately after being established.
Could you share your example code if possible? It would be super helpful to understand how you configured transport and relay. Thank you.

[mxinden]

For debugging purposes, does adding the KeepAlive NetworkBehaviour help?

[fgimenez]

Could you share your example code if possible?

@youngjoon-lee sorry, I can't share this code, but it really doesn't have anything special other than what was mentioned in #3034 (reply in thread), did you try @mxinden suggestion about using KeepAlive?

[youngjoon-lee]

Thank you @mxinden @fgimenez , It works with keep-alive!