Stuck at redis #NA2

[pgtuk]

I’m stuck on Stage #NA2.

I get it that we don’t receive ACK from some replicas. If replica doesn’t respond to GETACK i get connection reset by peer error, which is handled as if connection has been closed. But tests expect command propagation for this kind of replicas, which i don’t understand how is it possible.
I can’t propagate commands via this exact connection, replica should’ve reconnected.

I would appreciate any help, stuck on this one for too long.

Here are my logs:

#[derive(Debug)]
pub struct Connection {
    stream: BufWriter<TcpStream>,
    buffer: BytesMut,
    pub(crate) is_repl_conn: bool,
}

impl Connection {
    pub fn new(stream: TcpStream) -> Connection {
        // let (reader, writer) = stream.split()

        Connection {
            stream: BufWriter::new(stream),
            buffer: BytesMut::with_capacity(4096),
            is_repl_conn: false,
        }
    }

    pub async fn read_frame(&mut self) -> Result<Option<Frame>, FrameError> {
        loop {
            if let Some(frame) = self.parse_frame()? {
                return Ok(Some(frame));
            };

            if 0 == self.stream.read_buf(&mut self.buffer).await? {
                if self.buffer.is_empty() {
                    return Ok(None);
                } else {
                    return Err("connection reset by peer".into());
                }
            }
        }
    }
}

[rohitpaulk]

@pgtuk the connection reset seems to happen after the test fails and everything is being closed down (note the “Terminating program” log).

The tester doesn’t make spawn replicas with broken connections - some of them might not reply to ACKs but their connections should stay alive.