Deprecated YamuxConfig::set_max_buffer_size

[MitchTurner]

It looks like you're moving to deprecate set_max_buffer_size on YamuxConfig. What is the recommended replacement for this method? If I don't set this value my peering breaks.

[mxinden]

Why does your peering (I am assuming connection) break when you don't set the value?

[thomaseizinger]

What does this test exactly do? We have yamux 0.13 working in all our integration tests so I am surprised it fails for you.

[MitchTurner]

The test spins up 3 nodes. Node A is connected to Node B, Node B is connected to Node C, but Node C is not connected to Node A.

The test is to see that a message can be published by A and relayed to C.

It fails when A tries to publish the message with the error InsufficientPeers.

[thomaseizinger]

Wait so your connection does succeed but you can't publish?

[MitchTurner]

Node A waits until it has a peer and then tries to publish.

https://github.com/FuelLabs/fuel-core/blob/6df2ff2748db059f9982706c18f07077ef9cc563/crates/services/p2p/src/p2p_service.rs#L1396

The assumption here is that if we get a PeerInfoUpdated event and peer is included in the peer_manager, then it is included in the gossipsub, but maybe there is some race condition in our implementation or perhaps elsewhere that doesn't guarantee this.

I seem to remember debugging something similar to this while upgrading. I'm not sure what the solution was, possilbly reordering some of the behaviors in the FuelBehaviour... I had do that for a couple bugs.

Doesn't really explain why the yamux buffer stuff breaks this.

[thomaseizinger]

• If you turn up the logs for libp2p_swarm, do you see the connection getting established?
• Can you compare logs with and without set_max_buffer_size (i.e. version 0.12 and version 0.13)?
• What is your connection timeout? More than 0?
• Are all peers subscribed to the same topic?