[BUG] Intermittent error executing cncli sync on 1.32.1

[feqifei]

Hi, executing cncli sync on my public testnet node I run into the following error:
thread 'main' panicked at 'called Result::unwrap() on an Err value: Os { code: 104, kind: ConnectionReset, message: "Connection reset by peer" }', /home/ubuntu/.cargo/registry/src/github.com-1ecc6299db9ec823/cardano_ouroboros_network-0.2.8/src/mux/connection.rs:202:67note: run with RUST_BACKTRACE=1 environment variable to display a backtrace

I've not been able to replicate it, so not able to provide the backtrace log.
Node version 1.32.1, p2p feature activated

I informed Andrew Westberg, he tested on his node that faced the same issue on same code line. The second attempt went fine.
Also his node is on 1.32.1 with p2p activated.

I didn't face the issue on my mainnet nodes where the p2p feature is off.
I also updated the testnet node to 1.33.0-rc1 tag and I tried dozens of times to replicate the issue without success.

[mark-stopka]

@AndrewWestberg any chance you've got the backtrace?

[AndrewWestberg]

@mark-stopka The screenshot above is mine. I couldn't reproduce it when I tried to run it with backtrace. The line of code it fails on is provided so maybe we can figure out the issue by just looking at the code.

[mark-stopka]

@AndrewWestberg it is the logging of the message size;

rust-cardano-ouroboros-network/src/mux/connection.rs

Line 202 in 0e07f0c

let len = self.stream.write(&msg).unwrap();

The TODO for that code section does include a note to handle errors.

[AndrewWestberg]

I guess typically that code never throws any errors, but on rare occasion, some network communication glitch happens.

[pavlix]

I guess typically that code never throws any errors, but on rare occasion, some network communication glitch happens.

What is your ideal idea of recovery in that case? I would like to refactor the library to avoid most panic!()s but I'm curious how the tool would respond to an actual connection error.