Fix unexpected Marking peer disconnected in DHT

[ackintosh]

Issue Addressed

We get the log below when LH fails to dial a peer:

2024-07-19T02:30:14.4401703Z Jul 19 02:30:14.439 DEBG Marking peer disconnected in DHT        error: Connection denied: ConnectionDenied { inner: Exceeded { limit: 1, kind: EstablishedPerPeer } }, peer_id: 

I noticed that the disconnect happens unexpectedly in the following case:

Lighthouse dials to a peer twice using TCP and QUIC (if QUIC is not disabled). Usually, one establishes a connection, and the other fails because the peer allows only one connection per peer. I think we shouldn't Marking peer disconnected in DHT since a TCP (in this case) connection has been established between the nodes.

Proposed Changes

I added a check to see if there’s an active connection before the disconnect.

[jxs]

LGTM Akihito, thanks! Left a comment

[jxs]

after yesterday's call where @jimmygchen and @AgeManning mentioned that this was occurring on a testnet with only Ligthouse clients I went to research this further.

It turns out that The connection limits NetworkBehaviour currently forbids us to have more than 1 connection per peer.
The error returned from the Swarm when a NetworkBehaviour rejects an outbound connection is DialError::Denied, so this PR addresses the issue it aims to.

[AgeManning]

This looks good to me, in that is masks the error.

However the deeper problem is that we are trying to connect to a peer twice.

It seems to me the easiest solution is that mask this error and force sequential connections when dialing. So this PR and a future one together is the easiest solution I think

[AgeManning]

@ackintosh - I think you just need to merge lastest unstable to fix CI

[jxs]

@Mergifyio queue

[mergify]

queue

✅ The pull request has been merged automatically

The pull request has been merged automatically at 62a39af